Lebenswissenschaften und Ethik
Refine
Year of publication
Document Type
Language
- English (119) (remove)
Is part of the Bibliography
- no (119)
Keywords
- Bildgebendes Verfahren (17)
- Deep Learning (15)
- Artificial Intelligence (12)
- Diagnose (11)
- Maschinelles Lernen (9)
- Gehirn (8)
- Kernspintomografie (8)
- Speiseröhrenkrankheit (8)
- Dreidimensionale Bildverarbeitung (7)
- Künstliche Intelligenz (7)
Institute
- Regensburg Medical Image Computing (ReMIC) (119) (remove)
Begutachtungsstatus
- peer-reviewed (69)
- begutachtet (1)
In 2015 we began a sub-challenge at the EndoVis workshop at MICCAI in Munich using endoscope images of exvivo tissue with automatically generated annotations from robot forward kinematics and instrument CAD models. However, the limited background variation and simple motion rendered the dataset uninformative in learning about which techniques would be suitable for segmentation in real surgery. In 2017, at the same workshop in Quebec we introduced the robotic instrument segmentation dataset with 10 teams participating in the challenge to perform binary, articulating parts and type segmentation of da Vinci instruments. This challenge included realistic instrument motion and more complex porcine tissue as background and was widely addressed with modfications on U-Nets and other popular CNN architectures [1].
In 2018 we added to the complexity by introducing a set of anatomical objects and medical devices to the segmented classes. To avoid over-complicating the challenge, we continued with porcine data which is dramatically simpler than human tissue due to the lack of fatty tissue occluding many organs.
Objective: Artificial intelligence (AI) may reduce underdiagnosed or overlooked upper GI (UGI) neoplastic and preneoplastic conditions, due to subtle appearance and low disease prevalence. Only disease-specific AI performances have been reported, generating uncertainty on its clinical value.
Design: We searched PubMed, Embase and Scopus until July 2020, for studies on the diagnostic performance of AI in detection and characterisation of UGI lesions. Primary outcomes were pooled diagnostic accuracy, sensitivity and specificity of AI. Secondary outcomes were pooled positive (PPV) and negative (NPV) predictive values. We calculated pooled proportion rates (%), designed summary receiving operating characteristic curves with respective area under the curves (AUCs) and performed metaregression and sensitivity analysis.
Results: Overall, 19 studies on detection of oesophageal squamous cell neoplasia (ESCN) or Barrett's esophagus-related neoplasia (BERN) or gastric adenocarcinoma (GCA) were included with 218, 445, 453 patients and 7976, 2340, 13 562 images, respectively. AI-sensitivity/specificity/PPV/NPV/positive likelihood ratio/negative likelihood ratio for UGI neoplasia detection were 90% (CI 85% to 94%)/89% (CI 85% to 92%)/87% (CI 83% to 91%)/91% (CI 87% to 94%)/8.2 (CI 5.7 to 11.7)/0.111 (CI 0.071 to 0.175), respectively, with an overall AUC of 0.95 (CI 0.93 to 0.97). No difference in AI performance across ESCN, BERN and GCA was found, AUC being 0.94 (CI 0.52 to 0.99), 0.96 (CI 0.95 to 0.98), 0.93 (CI 0.83 to 0.99), respectively. Overall, study quality was low, with high risk of selection bias. No significant publication bias was found.
Conclusion: We found a high overall AI accuracy for the diagnosis of any neoplastic lesion of the UGI tract that was independent of the underlying condition. This may be expected to substantially reduce the miss rate of precancerous lesions and early cancer when implemented in clinical practice.
Signal transmission between different brain regions requires connecting fiber tracts, the structural basis of the human connectome. In contrast to animal brains, where a multitude of tract tracing methods can be used, magnetic resonance (MR)-based diffusion imaging is presently the only promising approach to study fiber tracts between specific human brain regions. However, this procedure has various inherent restrictions caused by its relatively low spatial resolution. Here, we introduce 3D-polarized light imaging (3D-PLI) to map the three-dimensional course of fiber tracts in the human brain with a resolution at a submillimeter scale based on a voxel size of 100 μm isotropic or less. 3D-PLI demonstrates nerve fibers by utilizing their intrinsic birefringence of myelin sheaths surrounding axons. This optical method enables the demonstration of 3D fiber orientations in serial microtome sections of entire human brains. Examples for the feasibility of this novel approach are given here. 3D-PLI enables the study of brain regions of intense fiber crossing in unprecedented detail, and provides an independent evaluation of fiber tracts derived from diffusion imaging data.
Although it has long been presumed that d-amino acids are uncommon in mammalians, substantial amounts of free d-serine have been detected in the mammalian brain. d-Serine has been demonstrated to be an important modulator of glutamatergic neurotransmission and acts as an agonist at the strychnine-insensitive glycine site of N-methyl-d-aspartate receptors. The blood-to-brain transfer of d-serine is thought to be extremely low, and it is assumed that d-serine is generated by isomerization of l-serine in the brain. Stimulated by the observation of a preferred transport of the d-isomer of proline at the blood–brain barrier, we investigated the differential uptake of [3H]-d-serine and [3H]-l-serine in the rat brain 1 h after intravenous injection using quantitative autoradiography. Surprisingly, brain uptake of [3H]-d-serine was significantly higher than that of [3H]-l-serine, indicating a preferred transport of the d-enantiomer of serine at the blood–brain barrier. This finding indicates that exogenous d-serine may have a direct influence on glutamatergic neurotransmission and associated diseases.
Objectives: C-11-methionine (MET) is particularly useful in brain tumor diagnosis but unspecific uptake e.g. in cerebral ischemia has been reported (1). The F-18-labeled amino acid O-(2-[F-18]fluoroethyl)-L-tyrosine (FET) shows a similar clinical potential as MET in brain tumor diagnosis but is applicable on a wider clinical scale. The aim of this study was to evaluate the uptake of FET and H-3-MET in focal cortical ischemia in rats by dual tracer autoradiography.
Methods: Focal cortical ischemia was induced in 12 Fisher CDF rats using the photothrombosis model (PT). One day (n=3) , two days (n=5) and 7 days (n=4) after induction of the lesion FET and H-3-MET were injected intravenously. One hour after tracer injection animals were killed, the brains were removed immediately and frozen in 2-methylbutane at -50°C. Brains were cut in coronal sections (thickness: 20 µm) and exposed first to H-3 insensitive photoimager plates to measure FET distribution. After decay of F-18 the distribution of H-3-MET was determined. The autoradiograms were evaluated by regions of interest (ROIs) placed on areas with increased tracer uptake in the PT and the contralateral brain. Lesion to brain ratios (L/B) were calculated by dividing the mean uptake in the lesion and the brain. Based on previous studies in gliomas a L/B ratio > 1.6 was considered as pathological for FET.
Results: Variable increased uptake of both tracers was observed in the PT and its demarcation zone at all stages after PT. The cut-off level of 1.6 for FET was exceeded in 9/12 animals. One day after PT the L/B ratios were 2.0 ± 0.6 for FET vs. 2.1 ± 1.0 for MET (mean ± SD); two days after lesion 2.2 ± 0.7 for FET vs. 2.7 ± 1.0 for MET and 7 days after lesion 2.4 ± 0.4 for FET vs. 2.4 ± 0.1 for MET. In single cases discrepancies in the uptake pattern of FET and MET were observed.
Conclusions: FET like MET may exhibit significant uptake in infarcted areas or the immediate vincinity which has to be considered in the differential diagnosis of unkown brain lesions. The discrepancies in the uptake pattern of FET and MET in some cases indicates either differences in the transport mechanisms of both amino acids or a different affinity for certain cellular components.
Mass spectrometric imaging (MSI) is a young innovative analytical technique and combines different fields of advanced mass spectrometry and biomedical research with the aim to provide maps of elements and molecules, complexes or fragments. Especially essential metals such as zinc, copper, iron and manganese play a functional role in signaling, metabolism and homeostasis of the cell. Due to the high degree of spatial organization of metals in biological systems their distribution analysis is of key interest in life sciences. We have developed analytical techniques termed BrainMet using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) imaging to measure the distribution of trace metals in biological tissues for biomedical research and feasibility studies—including bioaccumulation and bioavailability studies, ecological risk assessment and toxicity studies in humans and other organisms. The analytical BrainMet techniques provide quantitative images of metal distributions in brain tissue slices which can be combined with other imaging modalities such as photomicrography of native or processed tissue (histochemistry, immunostaining) and autoradiography or with in vivo techniques such as positron emission tomography or magnetic resonance tomography.
Prospective and instrumental developments will be discussed concerning the development of the metalloprotein microscopy using a laser microdissection (LMD) apparatus for specific sample introduction into an inductively coupled plasma mass spectrometer (LMD-ICP-MS) or an application of the near field effect in LA-ICP-MS (NF-LA-ICP-MS). These nano-scale mass spectrometric techniques provide improved spatial resolution down to the single cell level.
Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) has been developed and established as an emerging technique in the generation of quantitative images of metal distributions in thin tissue sections of brain samples (such as human, rat and mouse brain), with applications in research related to neurodegenerative disorders. A new analytical protocol is described which includes sample preparation by cryo-cutting of thin tissue sections and matrix-matched laboratory standards, mass spectrometric measurements, data acquisition, and quantitative analysis. Specific examples of the bioimaging of metal distributions in normal rodent brains are provided. Differences to the normal were assessed in a Parkinson’s disease and a stroke brain model. Furthermore, changes during normal aging were studied. Powerful analytical techniques are also required for the determination and characterization of metal-containing proteins within a large pool of proteins, e.g., after denaturing or non-denaturing electrophoretic separation of proteins in one-dimensional and two-dimensional gels. LA-ICP-MS can be employed to detect metalloproteins in protein bands or spots separated after gel electrophoresis. MALDI-MS can then be used to identify specific metal-containing proteins in these bands or spots. The combination of these techniques is described in the second section.
Bioimaging of Metals by Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS)
(2010)
The distribution analysis of (essential, beneficial, or toxic) metals (e.g., Cu, Fe, Zn, Pb, and others), metalloids, and non‐metals in biological tissues is of key interest in life science. Over the past few years, the development and application of several imaging mass spectrometric techniques has been rapidly growing in biology and medicine. Especially, in brain research metalloproteins are in the focus of targeted therapy approaches of neurodegenerative diseases such as Alzheimer's and Parkinson's disease, or stroke, or tumor growth. Laser ablation inductively coupled plasma mass spectrometry (LA‐ICP‐MS) using double‐focusing sector field (LA‐ICP‐SFMS) or quadrupole‐based mass spectrometers (LA‐ICP‐QMS) has been successfully applied as a powerful imaging (mapping) technique to produce quantitative images of detailed regionally specific element distributions in thin tissue sections of human or rodent brain. Imaging LA‐ICP‐QMS was also applied to investigate metal distributions in plant and animal sections to study, for example, the uptake and transport of nutrient and toxic elements or environmental contamination. The combination of imaging LA‐ICP‐MS of metals with proteomic studies using biomolecular mass spectrometry identifies metal‐containing proteins and also phosphoproteins. Metal‐containing proteins were imaged in a two‐dimensional gel after electrophoretic separation of proteins (SDS or Blue Native PAGE). Recent progress in LA‐ICP‐MS imaging as a stand‐alone technique and in combination with MALDI/ESI‐MS for selected life science applications is summarized.
Purpose
MR-based attenuation correction (AC) will become an integral part of combined PET/MR systems. Here, we propose a toolbox to validate MR-AC of clinical PET/MRI data sets.
Methods
Torso scans of ten patients were acquired on a combined PET/CT and on a 1.5-T MRI system. MR-based attenuation data were derived from the CT following MR–CT image co-registration and subsequent histogram matching. PET images were reconstructed after CT- (PET/CT) and MR-based AC (PET/MRI). Lesion-to-background (L/B) ratios were estimated on PET/CT and PET/MRI.
Results
MR–CT histogram matching leads to a mean voxel intensity difference in the CT- and MR-based attenuation images of 12% (max). Mean differences between PET/MRI and PET/CT were 19% (max). L/B ratios were similar except for the lung where local misregistration and intensity transformation leads to a biased PET/MRI.
Conclusion
Our toolbox can be used to study pitfalls in MR-AC. We found that co-registration accuracy and pixel value transformation determine the accuracy of PET/MRI.
Document recommendation systems for locating relevant literature have mostly relied on methods developed a decade ago. This is largely due to the lack of a large offline gold-standard benchmark of relevant documents that cover a variety of research fields such that newly developed literature search techniques can be compared, improved and translated into practice. To overcome this bottleneck, we have established the RElevant LIterature SearcH consortium consisting of more than 1500 scientists from 84 countries, who have collectively annotated the relevance of over 180 000 PubMed-listed articles with regard to their respective seed (input) article/s. The majority of annotations were contributed by highly experienced, original authors of the seed articles. The collected data cover 76% of all unique PubMed Medical Subject Headings descriptors. No systematic biases were observed across different experience levels, research fields or time spent on annotations. More importantly, annotations of the same document pairs contributed by different scientists were highly concordant. We further show that the three representative baseline methods used to generate recommended articles for evaluation (Okapi Best Matching 25, Term Frequency–Inverse Document Frequency and PubMed Related Articles) had similar overall performances. Additionally, we found that these methods each tend to produce distinct collections of recommended articles, suggesting that a hybrid method may be required to completely capture all relevant articles. The established database server located at https://relishdb.ict.griffith.edu.au is freely available for the downloading of annotation data and the blind testing of new methods. We expect that this benchmark will be useful for stimulating the development of new powerful techniques for title and title/abstract-based search engines for relevant articles in biomedical research.
Aims
Congenital breast asymmetry represents a particular challenge to the classic techniques of plastic surgery given the young age of patients at presentation. This study reviews and compares the long-term results of traditional breast augmentation using silicone implants and the more innovative technique of lipografting.
Methods
To achieve this, we not only captured subjective parameters such as satisfaction with outcome and symmetry, but also objective parameters including breast vol-ume and anthropometric measurements. The objective examination was performed manually and by using the Vectra H2 photogrammetry scanning system.
Results
Differences between patients undergoing either implant augmentation or lipograft were revealed not to be significant with respect to patient satisfaction with surgical outcome (p= 0.55) and symmetry (p= 0.69). Furthermore, a breast symmetry of 93 % was reported in both groups. Likewise, no statistically significant volume difference between the left and right breasts was observed in both groups (p\0.41). However, lipograft patients needed on average 2.9 procedures to achieve the desired result, compared with 1.3 for implant augmentation. In contrast, patients treated with implant augmentation may require anumber of implant changes during their lifetime.
Conclusion
Both methods may be considered for patients presenting with congenital breast asymmetry.
Chest X-Ray (CXR) images as part of a non-invasive diagnosis method are commonly used in today’s medical workflow. In traditional methods, physicians usually use their experience to interpret CXR images, however, there is a large interobserver variance. Computer vision may be used as a standard for assisted diagnosis. In this study, we applied an encoder-decoder neural network architecture for automatic lung region detection. We compared a three-class approach (left lung, right lung, background) and a two-class approach (lung, background). The differentiation of left and right lungs as direct result of a semantic segmentation on basis of neural nets rather than post-processing a lung-background segmentation is done here for the first time. Our evaluation was done on the NIH Chest X-ray dataset, from which 1736 images were extracted and manually annotated. We achieved 94:9% mIoU and 92% mIoU as segmentation quality measures for the two-class-model and the three-class-model, respectively. This result is very promising for the segmentation of lung regions having the simultaneous classification of left and right lung in mind.
Polarized light imaging (PLI) enables the evaluation of fiber orientations in histological sections of human postmortem brains, with ultra-high spatial resolution. PLI is based on the birefringent properties of the myelin sheath of nerve fibers. As a result, the polarization state of light propagating through a rotating polarimeter is changed in such a way that the detected signal at each measurement unit of a charged-coupled device (CCD) camera describes a sinusoidal signal. Vectors of the fiber orientation defined by inclination and direction angles can then directly be derived from the optical signals employing PLI analysis. However, noise, light scatter and filter inhomogeneities interfere with the original sinusoidal PLI signals. We here introduce a novel method using independent component analysis (ICA) to decompose the PLI images into statistically independent component maps. After decomposition, gray and white matter structures can clearly be distinguished from noise and other artifacts. The signal enhancement after artifact rejection is quantitatively evaluated in 134 histological whole brain sections. Thus, the primary sinusoidal signals from polarized light imaging can be effectively restored after noise and artifact rejection utilizing ICA. Our method therefore contributes to the analysis of nerve fiber orientation in the human brain within a micrometer scale.
Computer-assisted analysis of endoscopic images can be helpful to the automatic diagnosis and classification of neoplastic lesions. Barrett's esophagus (BE) is a common type of reflux that is not straight forward to be detected by endoscopic surveillance, thus being way susceptible to erroneous diagnosis, which can cause cancer when not treated properly. In this work, we introduce the Optimum-Path Forest (OPF) classifier to the task of automatic identification of Barrett'sesophagus, with promising results and outperforming the well known Support Vector Machines (SVM) in the aforementioned context. We consider describing endoscopic images by means of feature extractors based on key point information, such as the Speeded up Robust Features (SURF) and Scale-Invariant Feature Transform (SIFT), for further designing a bag-of-visual-wordsthat is used to feed both OPF and SVM classifiers. The best results were obtained by means of the OPF classifier for both feature extractors, with values lying on 0.732 (SURF) - 0.735(SIFT) for sensitivity, 0.782 (SURF) - 0.806 (SIFT) for specificity, and 0.738 (SURF) - 0.732 (SIFT) for the accuracy.
In this work, we propose the use of single channel Color Co-occurrence Matrices for texture description of Barrett’sEsophagus (BE)and adenocarcinoma images. Further classification using supervised learning techniques, such as Optimum-Path Forest (OPF), Support Vector Machines with Radial Basisunction (SVM-RBF) and Bayesian classifier supports the contextof automatic BE and adenocarcinoma diagnosis. We validated three approaches of classification based on patches, patients and images in two datasets (MICCAI 2015 and Augsburg) using the color-and-texture descriptors and the machine learning techniques. Concerning MICCAI 2015 dataset, the best results were obtained using the blue channel for the descriptors and the supervised OPF for classification purposes in the patch-based approach, with sensitivity nearly to 73% for positive adenocarcinoma identification and specificity close to 77% for BE (non-cancerous) patch classification. Regarding the Augsburg dataset, the most accurate results were also obtained using both OPF classifier and blue channel descriptor for the feature extraction, with sensitivity close to 67% and specificity around to76%. Our work highlights new advances in the related research area and provides a promising technique that combines color and texture information, allied to three different approaches of dataset pre-processing aiming to configure robust scenarios for the classification step.
The development of adenocarcinoma in Barrett’s esophagus is difficult to detect by endoscopic surveillance of patients with signs of dysplasia. Computer assisted diagnosis of endoscopic images (CAD) could therefore be most helpful in the demarcation and classification of neoplastic lesions. In this study we tested the feasibility of a CAD method based on Speeded up Robust Feature Detection (SURF). A given database containing 100 images from 39 patients served as benchmark for feature based classification models. Half of the images had previously been diagnosed by five clinical experts as being ”cancerous”, the other half as ”non-cancerous”. Cancerous image regions had been visibly delineated (masked) by the clinicians. SURF features acquired from full images as well as from masked areas were utilized for the supervised training and testing of an SVM classifier. The predictive accuracy of the developed CAD system is illustrated by sensitivity and specificity values. The results based on full image matching where 0.78 (sensitivity) and 0.82 (specificity) were achieved, while the masked region approach generated results of 0.90 and 0.95, respectively.
Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their level of accountability and transparency must be provided in such evaluations. The reliability related to machine learning predictions must be explained and interpreted, especially if diagnosis support is addressed. For this task, the black-box nature of deep learning techniques must be lightened up to transfer its promising results into clinical practice. Hence, we aim to investigate the use of explainable artificial intelligence techniques to quantitatively highlight discriminative regions during the classification of earlycancerous tissues in Barrett’s esophagus-diagnosed patients. Four Convolutional Neural Network models (AlexNet, SqueezeNet, ResNet50, and VGG16) were analyzed using five different interpretation techniques (saliency, guided backpropagation, integrated gradients, input × gradients, and DeepLIFT) to compare their agreement with experts’ previous annotations of cancerous tissue. We could show that saliency attributes match best with the manual experts’ delineations. Moreover, there is moderate to high correlation between the sensitivity of a model and the human-and-computer agreement. The results also lightened that the higher the model’s sensitivity, the stronger the correlation of human and computational segmentation agreement. We observed a relevant relation between computational learning and experts’ insights, demonstrating how human knowledge may influence the correct computational learning.
This work presents a systematic review concerning recent studies and technologies of machine learning for Barrett's esophagus (BE) diagnosis and treatment. The use of artificial intelligence is a brand new and promising way to evaluate such disease. We compile some works published at some well-established databases, such as Science Direct, IEEEXplore, PubMed, Plos One, Multidisciplinary Digital Publishing Institute (MDPI), Association for Computing Machinery (ACM), Springer, and Hindawi Publishing Corporation. Each selected work has been analyzed to present its objective, methodology, and results. The BE progression to dysplasia or adenocarcinoma shows a complex pattern to be detected during endoscopic surveillance. Therefore, it is valuable to assist its diagnosis and automatic identification using computer analysis. The evaluation of the BE dysplasia can be performed through manual or automated segmentation through machine learning techniques. Finally, in this survey, we reviewed recent studies focused on the automatic detection of the neoplastic region for classification purposes using machine learning methods.
Barrett's esophagus denotes a disorder in the digestive system that affects the esophagus' mucosal cells, causing reflux, and showing potential convergence to esophageal adenocarcinoma if not treated in initial stages. Thus, fast and reliable computer-aided diagnosis becomes considerably welcome. Nevertheless, such approaches usually suffer from imbalanced datasets, which can be addressed through Generative Adversarial Networks (GANs). Such techniques generate realistic images based on observed samples, even though at the cost of a proper selection of its hyperparameters. Many works employed a class of nature-inspired algorithms called metaheuristics to tackle the problem considering distinct deep learning approaches. Therefore, this paper's main contribution is to introduce metaheuristic techniques to fine-tune GANs in the context of Barrett's esophagus identification, as well as to investigate the feasibility of generating high-quality synthetic images for early-cancer assisted identification.
Quantification of growth in experimental F98 and C6 rat brain tumours was performed on 51 rat brains, 17 of which have been further assessed by 3D tumour reconstruction. Brains were cryosliced and radio-labelled with a ligand of the peripheral type benzodiazepine-receptor (pBR), 3H-Pk11195 [(1-(2-chlorophenyl)-N-methyl-N-(1-methyl-propylene)-3-isoquinoline-carboxamide)] by receptor autoradiography. Manually segmented and automatically registered tumours have been 3D-reconstructed for volumetric comparison on the basis of 3H-Pk11195-based tumour recognition. Furthermore automatically computed areas of −300 μm inner (marginal) zone as well as 300 μm and 600 μm outer tumour space were quantified. These three different regions were transferred onto other adjacent slices that had been labelled by receptor autoradiography with the A1 Adenosine receptor (A1AR)-ligand 3H-CPFPX (3H-8-cyclopentyl-3-(3-fluorpropyl)-1-propylxanthine) for quantitative assessment of A1AR in the three different tumour zones. Hence, a method is described for quantifying various receptor protein systems in the tumour as well as in the marginal invasive zones around experimentally implanted rat brain tumours and their representation in the tumour microenvironment as well as in 3D space. Furthermore, a tool for automatically reading out radio-labelled rat brain slices from auto radiographic films was developed, reconstructed into a consistent 3D-tumour model and the zones around the tumour were visualized. A1AR expression was found to depend upon the tumour volume in C6 animals, but is independent on the time of tumour development. In F98 animals, a significant increase in A1AR receptor protein was found in the Peritumoural zone as a function of time of tumour development and tumour volume.
Medical image processing provides core innovation for medical imaging. This paper is focused on recent developments from science to applications analyzing the past fifteen years of history of the proceedings of the German annual meeting on medical image processing (BVM). Furthermore, some members of the program committee present their personal points of views: (i) multi-modality for imaging and diagnosis, (ii) analysis of diffusion-weighted imaging, (iii) model-based image analysis, (iv) registration of section images, (v) from images to information in digital endoscopy, and (vi) virtual reality and robotics. Medical imaging and medical image computing is seen as field of rapid development with clear trends to integrated applications in diagnostics, treatment planning and treatment.
Computer-aided diagnosis using deep learning in the evaluation of early oesophageal adenocarcinoma
(2019)
Computer-aided diagnosis using deep learning (CAD-DL) may be an instrument to improve endoscopic assessment of Barrett’s oesophagus
(BE) and early oesophageal adenocarcinoma (EAC). Based on still images from two databases, the diagnosis of EAC by CAD-DL reached sensitivities/specificities of 97%/88% (Augsburg data) and 92%/100% (Medical Image Computing and Computer-Assisted Intervention [MICCAI]
data) for white light (WL) images and 94%/80% for narrow band images (NBI) (Augsburg data), respectively. Tumour margins delineated by
experts into images were detected satisfactorily with a Dice coefficient (D) of 0.72. This could be a first step towards CAD-DL for BE assessment. If developed further, it could become a useful
adjunctive tool for patient management.
Aims:
The delineation of outer margins of early Barrett's cancer can be challenging even for experienced endoscopists. Artificial intelligence (AI) could assist endoscopists faced with this task. As of date, there is very limited experience in this domain. In this study, we demonstrate the measure of overlap (Dice coefficient = D) between highly experienced Barrett endoscopists and an AI system in the delineation of cancer margins (segmentation task).
Methods:
An AI system with a deep convolutional neural network (CNN) was trained and tested on high-definition endoscopic images of early Barrett's cancer (n = 33) and normal Barrett's mucosa (n = 41). The reference standard for the segmentation task were the manual delineations of tumor margins by three highly experienced Barrett endoscopists. Training of the AI system included patch generation, patch augmentation and adjustment of the CNN weights. Then, the segmentation results from patch classification and thresholding of the class probabilities. Segmentation results were evaluated using the Dice coefficient (D).
Results:
The Dice coefficient (D) which can range between 0 (no overlap) and 1 (complete overlap) was computed only for images correctly classified by the AI-system as cancerous. At a threshold of t = 0.5, a mean value of D = 0.72 was computed.
Conclusions:
AI with CNN performed reasonably well in the segmentation of the tumor region in Barrett's cancer, at least when compared with expert Barrett's endoscopists. AI holds a lot of promise as a tool for better visualization of tumor margins but may need further improvement and enhancement especially in real-time settings.
Based on previous work by our group with manual annotation of visible Barrett oesophagus (BE) cancer images, a real-time deep learning artificial intelligence (AI) system was developed. While an expert endoscopist conducts the endoscopic assessment of BE, our AI system captures random images from the real-time camera livestream and provides a global prediction (classification), as well as a dense prediction (segmentation) differentiating accurately between normal BE and early oesophageal adenocarcinoma (EAC). The AI system showed an accuracy of 89.9% on 14 cases with neoplastic BE.
The early diagnosis of cancer in Barrett’s esophagus is crucial for improving the prognosis. However, identifying Barrett’s esophagus-related neoplasia (BERN) is challenging, even for experts [1]. Four-quadrant biopsies may improve the detection of neoplasia, but they can be associated with sampling errors. The application of artificial intelligence (AI) to the assessment of Barrett’s esophagus could improve the diagnosis of BERN, and this has been demonstrated in both preclinical and clinical studies [2] [3].
In this video demonstration, we show the accurate detection and delineation of BERN in two patients ([Video 1]). In part 1, the AI system detects a mucosal cancer about 20 mm in size and accurately delineates the lesion in both white-light and narrow-band imaging. In part 2, a small island of BERN with high-grade dysplasia is detected and delineated in white-light, narrow-band, and texture and color enhancement imaging. The video shows the results using a transparent overlay of the mucosal cancer in real time as well as a full segmentation preview. Additionally, the optical flow allows for the assessment of endoscope movement, something which is inversely related to the reliability of the AI prediction. We demonstrate that multimodal imaging can be applied to the AI-assisted detection and segmentation of even small focal lesions in real time.
Background and aims: The accurate differentiation between T1a and T1b Barrett’s cancer has both therapeutic and prognostic implications but is challenging even for experienced physicians. We trained an Artificial Intelligence (AI) system on the basis of deep artificial neural networks (deep learning) to differentiate between T1a and T1b Barrett’s cancer white-light images.
Methods: Endoscopic images from three tertiary care centres in Germany were collected retrospectively. A deep learning system was trained and tested using the principles of cross-validation. A total of 230 white-light endoscopic images (108 T1a and 122 T1b) was evaluated with the AI-system. For comparison, the images were also classified by experts specialized in endoscopic diagnosis and treatment of Barrett’s cancer.
Results: The sensitivity, specificity, F1 and accuracy of the AI-system in the differentiation between T1a and T1b cancer lesions was 0.77, 0.64, 0.73 and 0.71, respectively. There was no statistically significant difference between the performance of the AI-system and that of human experts with sensitivity, specificity, F1 and accuracy of 0.63, 0.78, 0.67 and 0.70 respectively.
Conclusion: This pilot study demonstrates the first multicenter application of an AI-based system in the prediction of submucosal invasion in endoscopic images of Barrett’s cancer. AI scored equal to international experts in the field, but more work is necessary to improve the system and apply it to video sequences and in a real-life setting. Nevertheless, the correct prediction of submucosal invasion in Barret´s cancer remains challenging for both experts and AI.
In this study, we aimed to develop an artificial intelligence clinical decision support solution to mitigate operator-dependent limitations during complex endoscopic procedures such as endoscopic submucosal dissection and peroral endoscopic myotomy, for example, bleeding and perforation. A DeepLabv3-based model was trained to delineate vessels, tissue structures and instruments on endoscopic still images from such procedures. The mean cross-validated Intersection over Union and Dice Score were 63% and 76%, respectively. Applied to standardised video clips from third-space endoscopic procedures, the algorithm showed a mean vessel detection rate of 85% with a false-positive rate of 0.75/min. These performance statistics suggest a potential clinical benefit for procedure safety, time and also training.
Introduction
We present a clinical case showing the real-time detection, characterization and delineation of an early Barrett’s cancer using AI.
Patients and methods
A 70-year old patient with a long-segment Barrett’s esophagus (C5M7) was assessed with an AI algorithm.
Results
The AI system detected a 10 mm focal lesion and AI characterization predicted cancer with a probability of >90%. After ESD resection, histopathology showed mucosal adenocarcinoma (T1a (m), R0) confirming AI diagnosis.
Conclusion
We demonstrate the real-time AI detection, characterization and delineation of a small and early mucosal Barrett’s cancer.
The evaluation and assessment of Barrett’s esophagus is challenging for both expert and nonexpert endoscopists. However, the early diagnosis of cancer in Barrett’s esophagus is crucial for its prognosis, and could save costs. Pre-clinical and clinical studies on the application of Artificial Intelligence (AI) in Barrett’s esophagus have shown promising results. In this review, we focus on the current challenges and future perspectives of implementing AI systems in the management of patients with Barrett’s esophagus.
The growing number of publications on the application of artificial intelligence (AI) in medicine underlines the enormous importance and potential of this emerging field of research.
In gastrointestinal endoscopy, AI has been applied to all segments of the gastrointestinal tract most importantly in the detection and characterization of colorectal polyps. However, AI research has been published also in the stomach and esophagus for both neoplastic and non-neoplastic disorders.
The various technical as well as medical aspects of AI, however, remain confusing especially for non-expert physicians.
This physician-engineer co-authored review explains the basic technical aspects of AI and provides a comprehensive overview of recent publications on AI in gastrointestinal endoscopy. Finally, a basic insight is offered into understanding publications on AI in gastrointestinal endoscopy.
Aims
Artificial Intelligence (AI) systems in gastrointestinal endoscopy are narrow because they are trained to solve only one specific task. Unlike Narrow-AI, general AI systems may be able to solve multiple and unrelated tasks. We aimed to understand whether an AI system trained to detect, characterize, and segment early Barrett’s neoplasia (Barrett’s AI) is only capable of detecting this pathology or can also detect and segment other diseases like early squamous cell cancer (SCC).
Methods
120 white light (WL) and narrow-band endoscopic images (NBI) from 60 patients (1 WL and 1 NBI image per patient) were extracted from the endoscopic database of the University Hospital Augsburg. Images were annotated by three expert endoscopists with extensive experience in the diagnosis and endoscopic resection of early esophageal neoplasias. An AI system based on DeepLabV3+architecture dedicated to early Barrett’s neoplasia was tested on these images. The AI system was neither trained with SCC images nor had it seen the test images prior to evaluation. The overlap between the three expert annotations („expert-agreement“) was the ground truth for evaluating AI performance.
Results
Barrett’s AI detected early SCC with a mean intersection over reference (IoR) of 92% when at least 1 pixel of the AI prediction overlapped with the expert-agreement. When the threshold was increased to 5%, 10%, and 20% overlap with the expert-agreement, the IoR was 88%, 85% and 82%, respectively. The mean Intersection Over Union (IoU) – a metric according to segmentation quality between the AI prediction and the expert-agreement – was 0.45. The mean expert IoU as a measure of agreement between the three experts was 0.60.
Conclusions
In the context of this pilot study, the predictions of SCC by a Barrett’s dedicated AI showed some overlap to the expert-agreement. Therefore, features learned from Barrett’s cancer-related training might be helpful also for SCC prediction. Our results allow different possible explanations. On the one hand, some Barrett’s cancer features generalize toward the related task of assessing early SCC. On the other hand, the Barrett’s AI is less specific to Barrett’s cancer than a general predictor of pathological tissue. However, we expect to enhance the detection quality significantly by extending the training to SCC-specific data. The insight of this study opens the way towards a transfer learning approach for more efficient training of AI to solve tasks in other domains.
For accurate registration of histological sections blockface images are frequently used as three dimensional reference. However, due to the use of endocentric lenses the images suffer from perspective errors such as scaling and seemingly relative movement of planes which are located in different distances parallel to the imaging sensor. The suggested correction of those errors is based on the estimation of scaling factors derived from image registration of regions characterized by differing distances to the point of view in neighboring sections. The correction allows the generation of a consistent three dimensional blockface volume.
Virtual training of bone milling requires realtime and realistic haptics of the interaction between the ”virtual mill” and a ”virtual bone”. We propose an exponential abrasion model between a virtual one and the mill bit and combine it with a coarse representation of the virtual bone and the mill shaft for collision detection using the Bullet Physics Engine. We compare our exponential abrasion model to a widely used linear abrasion model and evaluate it quantitatively and qualitatively. The evaluation results show, that we can provide virtual milling in real-time, with an abrasion behavior similar to that proposed in the literature and with a realistic feeling of five different surgeons.
Training data for Neural Networks is often scarce in the medical domain, which often results in models that struggle to generalize and consequently showpoor performance on unseen datasets. Generally, adding augmentation methods to the training pipeline considerably enhances a model’s performance. Using the dataset of the Foot Ulcer Segmentation Challenge, we analyze two additional augmentation methods in the domain of chronic foot wounds - local warping of wound edges along with projection and blurring of shapes inside wounds. Our experiments show that improvements in the Dice similarity coefficient and Normalized Surface Distance metrics depend on a sensible selection of those augmentation methods.
BACKGROUND
Differentiation of high-flow from low-flow vascular malformations (VMs) is crucial for therapeutic management of this orphan disease.
OBJECTIVE
A convolutional neural network (CNN) was evaluated for differentiation of peripheral vascular malformations (VMs) on T2-weighted short tau inversion recovery (STIR) MRI.
METHODS
527 MRIs (386 low-flow and 141 high-flow VMs) were randomly divided into training, validation and test set for this single-center study. 1) Results of the CNN's diagnostic performance were compared with that of two expert and four junior radiologists. 2) The influence of CNN's prediction on the radiologists' performance and diagnostic certainty was evaluated. 3) Junior radiologists' performance after self-training was compared with that of the CNN.
RESULTS
Compared with the expert radiologists the CNN achieved similar accuracy (92% vs. 97%, p = 0.11), sensitivity (80% vs. 93%, p = 0.16) and specificity (97% vs. 100%, p = 0.50). In comparison to the junior radiologists, the CNN had a higher specificity and accuracy (97% vs. 80%, p < 0.001; 92% vs. 77%, p < 0.001). CNN assistance had no significant influence on their diagnostic performance and certainty. After self-training, the junior radiologists' specificity and accuracy improved and were comparable to that of the CNN.
CONCLUSIONS
Diagnostic performance of the CNN for differentiating high-flow from low-flow VM was comparable to that of expert radiologists. CNN did not significantly improve the simulated daily practice of junior radiologists, self-training was more effective.
Utility of Smartphone-based Three-dimensional Surface Imaging for Digital Facial Anthropometry
(2024)
Background
The utilization of three-dimensional (3D) surface imaging for facial anthropometry is a significant asset for patients undergoing maxillofacial surgery. Notably, there have been recent advancements in smartphone technology that enable 3D surface imaging.
In this study, anthropometric assessments of the face were performed using a smartphone and a sophisticated 3D surface imaging system.
Methods
30 healthy volunteers (15 females and 15 males) were included in the study. An iPhone 14 Pro (Apple Inc., USA) using the application 3D Scanner App (Laan Consulting Corp., USA) and the Vectra M5 (Canfield Scientific, USA) were employed to create 3D surface models. For each participant, 19 anthropometric measurements were conducted on the 3D surface models. Subsequently, the anthropometric measurements generated by the two approaches were compared. The statistical techniques employed included the paired t-test, paired Wilcoxon signed-rank test, Bland–Altman analysis, and calculation of the intraclass correlation coefficient (ICC).
Results
All measurements showed excellent agreement between smartphone-based and Vectra M5-based measurements (ICC between 0.85 and 0.97). Statistical analysis revealed no statistically significant differences in the central tendencies for 17 of the 19 linear measurements. Despite the excellent agreement found, Bland–Altman analysis revealed that the 95% limits of agreement between the two methods exceeded ±3 mm for the majority of measurements.
Conclusion
Digital facial anthropometry using smartphones can serve as a valuable supplementary tool for surgeons, enhancing their communication with patients. However, the proposed data suggest that digital facial anthropometry using smartphones may not yet be suitable for certain diagnostic purposes that require high accuracy.
Purpose: In this trial, we used a previously developed prototype software to assess aesthetic results after reconstructive surgery for congenital breast asymmetry using automated anthropometry. To prove the consensus between the manual and automatic digital measurements, we evaluated the software by comparing the manual and automatic measurements of 46 breasts.
Methods: Twenty-three patients who underwent reconstructive surgery for congenital breast asymmetry at our institution were examined and underwent 3D surface imaging. Per patient, 14 manual and 14 computer-based anthropometric measurements were obtained according to a standardized protocol. Manual and automatic measurements, as well as the previously proposed Symmetry Index (SI), were compared.
Results: The Wilcoxon signed-rank test revealed no significant differences in six of the seven measurements between the automatic and manual assessments. The SI showed robust agreement between the automatic and manual methods.
Conclusion: The present trial validates our method for digital anthropometry. Despite the discrepancy in one measurement, all remaining measurements, including the SI, showed high agreement between the manual and automatic methods. The proposed data bring us one step closer to the long-term goal of establishing robust instruments to evaluate the results of breast surgery.
Background
Breast reconstruction is an important coping tool for patients undergoing a mastectomy. There are numerous surgical techniques in breast reconstruction surgery (BRS). Regardless of the technique used, creating a symmetric outcome is crucial for patients and plastic surgeons. Three-dimensional surface imaging enables surgeons and patients to assess the outcome’s symmetry in BRS. To discriminate between autologous and alloplastic techniques, we analyzed both techniques using objective optical computerized symmetry analysis. Software was developed that enables clinicians to assess optical breast symmetry using three-dimensional surface imaging.
Methods
Twenty-seven patients who had undergone autologous (n = 12) or alloplastic (n = 15) BRS received three-dimensional surface imaging. Anthropomorphic data were collected digitally using semiautomatic measurements and automatic measurements. Automatic measurements were taken using the newly developed software. To quantify symmetry, a Symmetry Index is proposed.
Results
Statistical analysis revealed that there is no dif- ference in the outcome symmetry between the two groups (t test for independent samples; p = 0.48, two-tailed).
Conclusion
This study’s findings provide a foundation for qualitative symmetry assessment in BRS using automatized digital anthropometry. In the present trial, no difference in the outcomes’ optical symmetry was detected between autologous and alloplastic approaches.
Video Documentation of endoscopic findings simplifies diagnostic counseling of the patient and aids pre-operative discussion among the medical team. Judgment of such images is still subjective and can not give a quantitative evaluation of the disease process regarding diagnosis or response to treatment. Modern treatment of early laryngeal cancer with laserablation requires intensive follow up and frequent direct laryngoscopy under general anesthesia with blind biopsies to detect any tumor residual or recurrence. Inflammatory conditions of the larynx are frequently confused with other causes of dysphonia. Mapping anddigital analysis of the documented image will suggest the tumor site and avoids undue blind biopsies under anesthesia. However, varying illumination results in different colors reflected from the same object. To achieve quantitative analysis, color constancy has to be assured. Inthis paper, the environment is presented which allow the objective judgment of larngoscopies.
The purpose of this study was to evaluate the quality of surface contouring of chondromalacic cartilage by bipolar radio frequency energy using different treatment patterns in an animal model, as well as examining the impact of the treatment onto chondrocyte viability by two different methods. Our experiments were conducted on 36 fresh osteochondral sections from the tibia plateau of slaughtered 6-month-old pigs, where the thickness of the cartilage is similar to that of human wrist cartilage. An area of 1 cm(2) was first treated with emery paper to simulate the chondromalacic cartilage. Then, the treatment with RFE followed in 6 different patterns. The osteochondral sections were assessed for cellular viability (live/dead assay, caspase (cell apoptosis marker) staining, and quantitative analysed images obtained by fluorescent microscopy). For a quantitative characterization of none or treated cartilage surfaces, various roughness parameters were measured using confocal laser scanning microscopy (Olympus LEXT OLS 4000 3D). To describe the roughness, the Root-Mean-Square parameter (Sq) was calculated. A smoothing effect of the cartilage surface was detectable upon each pattern of RFE treatment. The Sq for native cartilage was Sq=3.8 +/- 1.1 mu m. The best smoothing pattern was seen for two RFE passes and a 2-second pulsed mode (B2p2) with an Sq=27.3 +/- 4.9 mu m. However, with increased smoothing, an augmentation in chondrocyte death up to 95% was detected. Using bipolar RFE treatment in arthroscopy for small joints like the wrist or MCP joints should be used with caution. In the case of chondroplasty, there is a high chance to destroy the joint cartilage.
Despite multimodal treatment, the prognosis of high-grade gliomas is grim. As tumor growth is critically dependent on new blood vessel formation, antiangiogenic treatment approaches offer an innovative treatment strategy. Bevacizumab, a humanized monoclonal antibody, has been in the spotlight of antiangiogenic approaches for several years. Currently, MRI including contrast-enhanced T1-weighted and T2/fluid-attenuated inversion recovery (FLAIR) images is routinely used to evaluate antiangiogenic treatment response (Response Assessment in Neuro-Oncology criteria). However, by restoring the blood–brain barrier, bevacizumab may reduce T1 contrast enhancement and T2/FLAIR hyperintensity, thereby obscuring the imaging-based detection of progression. The aim of this review is to highlight the recent role of imaging biomarkers from MR and PET imaging on measurement of disease progression and treatment effectiveness in antiangiogenic therapies. Based on the reviewed studies, multimodal imaging combining standard MRI with new physiological MRI techniques and metabolic PET imaging, in particular amino acid tracers, may have the ability to detect antiangiogenic drug susceptibility or resistance prior to morphological changes. As advances occur in the development of therapies that target specific biochemical or molecular pathways and alter tumor physiology in potentially predictable ways, the validation of physiological and metabolic imaging biomarkers will become increasingly important in the near future.
Whilst considerable progress has been made in enhancing the quality of indirect laryngoscopy and image processing, the evaluation of clinical findings is still based on the clinician's judgement. The aim of this paper was to examine the feasibility of an objective computer-based method for evaluating laryngeal disease. Digitally recorded images obtained by 90 degree- and 70 degree-angled indirect rod laryngoscopy using standardized white balance values were made of 16 patients and 19 healthy subjects. The digital images were evaluated manually by the clinician based on a standardized questionnaire, and suspect lesions were marked and classified on the image. Following colour separation, normal vocal cord areas as well as suspect lesions were analyzed automatically using co-occurrence matrices, which compare colour differences between neighbouring pixels over a predefined distance. Whilst colour histograms did not provide sufficient information for distinguishing between healthy and diseased tissues, consideration of the blue content of neighbouring pixels enabled a correct classification in 81.4% of cases. If all colour channels (red, green and blue) were regarded simultaneously, the best classification correctness obtained was 77.1%. Although only a very basic classification differentiating between healthy and diseased tissue was attempted, the results showed progress compared to grey-scale histograms, which have been evaluated before. The results document a first step towards an objective, machine-based classification of laryngeal images, which could provide the basis for further development of an expert system for use in indirect laryngoscopy.
Background:
Reliable, time- and cost-effective, and clinician-friendly diagnostic tools are cornerstones in facial palsy (FP) patient management. Different automated FP grading systems have been developed but revealed persisting downsides such as insufficient accuracy and cost-intensive hardware. We aimed to overcome these barriers and programmed an automated grading system for FP patients utilizing the House and Brackmann scale (HBS).
Methods:
Image datasets of 86 patients seen at the Department of Plastic, Hand, and Reconstructive Surgery at the University Hospital Regensburg, Germany, between June 2017 and May 2021, were used to train the neural network and evaluate its accuracy. Nine facial poses per patient were analyzed by the algorithm.
Results:
The algorithm showed an accuracy of 100%. Oversampling did not result in altered outcomes, while the direct form displayed superior accuracy levels when compared to the modular classification form (n = 86; 100% vs. 99%). The Early Fusion technique was linked to improved accuracy outcomes in comparison to the Late Fusion and sequential method (n = 86; 100% vs. 96% vs. 97%).
Conclusions:
Our automated FP grading system combines high-level accuracy with cost- and time-effectiveness. Our algorithm may accelerate the grading process in FP patients and facilitate the FP surgeon’s workflow.
Biological applications like vesicle membrane analysis involve the precise segmentation of 3D structures in noisy volumetric data, obtained by techniques like magnetic resonance imaging (MRI) or laser scanning microscopy (LSM). Dealing with such data is a challenging task and requires robust and accurate segmentation methods. In this article, we propose a novel energy model for 3D segmentation fusing various cues like regional intensity subdivision, edge alignment and orientation information. The uniqueness of the approach consists in the definition of a new anisotropic regularizer, which accounts for the unbalanced slicing of the measured volume data, and the generalization of an efficient numerical scheme for solving the arising minimization problem, based on linearization and fixed-point iteration. We show how the proposed energy model can be optimized globally by making use of recent continuous convex relaxation techniques. The accuracy and robustness of the presented approach are demonstrated by evaluating it on multiple real data sets and comparing it to alternative segmentation methods based on level sets. Although the proposed model is designed with focus on the particular application at hand, it is general enough to be applied to a variety of different segmentation tasks.
The German workshop on medical image computing (BVM) has been held in different locations in Germany for more than 20 years. In terms of content, BVM focused on the computer-aided analysis of medical image data with a wide range of applications, e.g. in the area of imaging, diagnostics, operation planning, computer-aided intervention and visualization.
During this time, there have been remarkable methodological developments and upheavals, on which the BVM community has worked intensively. The area of machine learning should be emphasized, which has led to significant improvements, especially for tasks of classification and segmentation, but increasingly also in image formation and registration. As a result, work in connection with deep learning now dominates the BVM. These developments have also contributed to the establishment of medical image processing at the interface between computer science and medicine as one of the key technologies for the digitization of the health system.
In addition to the presentation of current research results, a central aspect of the BVM is primarily the promotion of young scientists from the diverse BVM community, covering not only Germany but also Austria, Switzerland, The Netherland and other European neighbors. The conference serves primarily doctoral students and postdocs, but also students with excellent bachelor and master theses as a platform to present their work, to enter into professional discourse with the community, and to establish networks with specialist colleagues. Despite the many conferences and congresses that are also relevant for medical image processing, the BVM has therefore lost none of its importance and attractiveness and has retained its permanent place in the annual conference rhythm.
Building on this foundation, there are some innovations and changes this year. The BVM 2021 was organized for the first time at the Ostbayerische Technische Hochschule Regensburg (OTH Regensburg, a technical university of applied sciences). After Aachen, Berlin, Erlangen, Freiburg, Hamburg, Heidelberg, Leipzig, Lübeck, and Munich, Regensburg is not just a new venue. OTH Regensburg is the first representative of the universities of applied sciences (HAW) to organize the conference, which differs to universities, university hospitals, or research centers like Fraunhofer or Helmholtz. This also considers the further development of the research landscape in Germany, where HAWs increasingly contribute to applied research in addition to their focus on teaching. This development is also reflected in the contributions submitted to the BVM in recent years.
At BVM 2021, which was held in a virtual format for the first time due to the Corona pandemic, an attractive and high-quality program was offered. Fortunately, the number of submissions increased significantly. Out of 97 submissions, 26 presentations, 51 posters and 5 software demonstrations were accepted via an anonymized reviewing process with three reviews each. The three best works have been awarded BVM prizes, selected by a separate committee.
Based on these high-quality submissions, we are able to present another special issue in the International Journal of Computer Assisted Radiology and Surgery (IJCARS). Out of the 97 submissions, the ones with the highest scores have been invited to submit an extended version of their paper to be presented in IJCARS. As a result, we are now able to present this special issue with seven excellent articles. Many submissions focus on machine learning in a medical context.
In order to fix a fracture using minimally invasive surgery approaches, surgeons are drilling complex and tiny bones with a 2 dimensional X-ray as single imaging modality in the operating room. Our novel haptic force-feedback and visual assisted training system will potentially help hand surgeons to learn the drilling procedure in a realistic visual environment. Within the simulation, the collision detection as well as the interaction between virtual drill, bone voxels and surfaces are important. In this work, the chai3d collision detection and force calculation algorithms are combined with a physics engine to simulate the bone drilling process. The chosen Bullet-Physics-Engine provides a stable simulation of rigid bodies, if the collision model of the drill and the tool holder is generated as a compound shape. Three haptic points are added to the K-wire tip for removing single voxels from the bone. For the drilling process three modes are proposed to emulate the different phases of drilling in restricting the movement of a haptic device.
One common method to fix fractures of the human hand after an accident is an osteosynthesis with Kirschner wires (K-wires) to stabilize the bone fragments. The insertion of K-wires is a delicate minimally invasive surgery, because surgeons operate almost without a sight. Since realistic training methods are time consuming, costly and insufficient, a virtual-reality (VR) based training system for the placement of K-wires was developed. As part of this, the current work deals with the real-time bone drilling simulation using a haptic force-feedback device.
To simulate the drilling, we introduce a virtual fixture based force-feedback drilling approach. By decomposition of the drilling task into individual phases, each phase can be handled individually to perfectly control the drilling procedure. We report about the related finite state machine (FSM), describe the haptic feedback of each state and explain, how to avoid jerking of the haptic force-feedback during state transition.
The usage of the virtual fixture approach results in a good haptic performance and a stable drilling behavior. This was confirmed by 26 expert surgeons, who evaluated the virtual drilling on the simulator and rated it as very realistic. To make the system even more convincing, we determined real drilling feed rates through experimental pig bone drilling and transferred them to our system. Due to a constant simulation thread we can guarantee a precise drilling motion.
Virtual fixtures based force-feedback calculation is able to simulate force-feedback assisted bone drilling with high quality and, thus, will have a great potential in developing medical applications.
Background: Currently, it is common practice to use three-dimensional (3D) printers not only for rapid prototyping in the industry, but also in the medical area to create medical applications for training inexperienced surgeons. In a clinical training simulator for minimally invasive bone drilling to fix hand fractures with Kirschner-wires (K-wires), a 3D-printed hand phantom must not only be geometrically but also haptically correct. Due to a limited view during an operation, surgeons need to perfectly localize underlying risk structures only by feeling of specific bony protrusions of the human hand.
Methods: The goal of this experiment is to imitate human soft tissue with its haptic and elasticity for a realistic hand phantom fabrication, using only a dual-material 3D printer and support-material-filled metamaterial between skin and bone. We present our workflow to generate lattice structures between hard bone and soft skin with iterative cube edge (CE) or cube face (CF) unit cells. Cuboid and finger shaped sample prints with and without inner hard bone in different lattice thickness are constructed and 3D printed.
Results: The most elastic available rubber-like material is too firm to imitate soft tissue. By reducing the amount of rubber in the inner volume through support material (SUP), objects become significantly softer. Without metamaterial, after disintegration, the SUP can be shifted through the volume and thus the body loses its original shape. Although the CE design increases the elasticity, it cannot restore the fabric form. In contrast to CE, the CF design increases not only the elasticity but also guarantees a local limitation of the SUP. Therefore, the body retains its shape and internal bones remain in its intended place. Various unit cell sizes, lattice thickening and skin thickness regulate the rubber material and SUP ratio. Test prints with higher SUP and lower rubber material percentage appear softer and vice versa. This was confirmed by an expert surgeon evaluation. Subjects adjudged pure rubber-like material as too firm and samples only filled with SUP or lattice structure in CE design as not suitable for imitating tissue. 3D-printed finger samples in CF design were rated as realistic compared to the haptic of human tissue with a good palpable bone structure.
Conclusions: We developed a new dual-material 3D print technique to imitate soft tissue of the human hand with its haptic properties. Blowy SUP is trapped within a lattice structure to soften rubber-like 3D print material, which makes it possible to reproduce a realistic replica of human hand soft tissue.
Currently, it is common practice to use three-dimensional (3D) printers not only for rapid prototyping in the industry, but also in the medical area to create medical applications for training inexperienced surgeons. In a clinical training simulator for minimally invasive bone drilling to fix hand fractures with Kirschner-wires (K-wires), a 3D printed hand phantom must not only be geometrically but also haptically correct. Due to a limited view during an operation, surgeons need to perfectly localize underlying risk structures only by feeling of specific bony protrusions of the human hand.
Background: For surgical fixation of bone fractures of the human hand, so-called Kirschner-wires (K-wires) are drilled through bone fragments. Due to the minimally invasive drilling procedures without a view of risk structures like vessels and nerves, a thorough training of young surgeons is necessary. For the development of a virtual reality (VR) based training system, a three-dimensional (3D) printed phantom hand is required. To ensure an intuitive operation, this phantom hand has to be realistic in both, its position relative to the driller as well as in its haptic features. The softest 3D printing material available on the market, however, is too hard to imitate human soft tissue. Therefore, a support-material (SUP) filled metamaterial is used to soften the raw material. Realistic haptic features are important to palpate protrusions of the bone to determine the drilling starting point and angle. An optical real-time tracking is used to transfer position and rotation to the training system.
Methods: A metamaterial already developed in previous work is further improved by use of a new unit cell. Thus, the amount of SUP within the volume can be increased and the tissue is softened further. In addition, the human anatomy is transferred to the entire hand model. A subcutaneous fat layer and penetration of air through pores into the volume simulate shiftability of skin layers. For optical tracking, a rotationally symmetrical marker attached to the phantom hand with corresponding reference marker is developed. In order to ensure trouble-free position transmission, various types of marker point applications are tested.
Results: Several cuboid and forearm sample prints lead to a final 30 centimeter long hand model. The whole haptic phantom could be printed faultless within about 17 hours. The metamaterial consisting of the new unit cell results in an increased SUP share of 4.32%. Validated by an expert surgeon study, this allows in combination with a displacement of the uppermost skin layer a good palpability of the bones. Tracking of the hand marker in dodecahedron design works trouble-free in conjunction with a reference marker attached to the worktop of the training system.
Conclusions: In this work, an optically tracked and haptically correct phantom hand was developed using dual-material 3D printing, which can be easily integrated into a surgical training system.
Object
The consistency of parametric registration in multi-temporal magnetic resonance (MR) imaging studies was evaluated.
Materials and methods
Serial MRI scans of adult patients with a brain tumor (glioma) were aligned by parametric registration. The performance of low-order spatial alignment (6/9/12 degrees of freedom) of different 3D serial MR-weighted images is evaluated. A registration protocol for the alignment of all images to one reference coordinate system at baseline is presented. Registration results were evaluated for both, multimodal intra-timepoint and mono-modal multi-temporal registration. The latter case might present a challenge to automatic intensity-based registration algorithms due to ill-defined correspondences. The performance of our algorithm was assessed by testing the inverse registration consistency. Four different similarity measures were evaluated to assess consistency.
Results
Careful visual inspection suggests that images are well aligned, but their consistency may be imperfect. Sub-voxel inconsistency within the brain was found for allsimilarity measures used for parametric multi-temporal registration. T1-weighted images were most reliable for establishing spatial correspondence between different timepoints.
Conclusions
The parametric registration algorithm is feasible for use in this application. The sub-voxel resolution mean displacement error of registration transformations demonstrates that the algorithm converges to an almost identical solution for forward and reverse registration.
Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) has been established as a powerful technique for the determination of metal and nonmetal distributions within biological systems with high sensitivity. An imaging LA-ICP-MS technique for Fe, Cu, Zn, and Mn was developed to produce large series of quantitative element maps in native brain sections of mice subchronically intoxicated with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridin (MPTP) as a model of Parkinson’s disease. Images were calibrated using matrix-matched laboratory standards. A software solution allowing a precise delineation of anatomical structures was implemented. Coronal brain sections were analyzed crossing the striatum and the substantia nigra, respectively. Animals sacrificed 2 h, 7 d, or 28 d after the last MPTP injection and controls were investigated.
We observed significant decreases of Cu concentrations in the periventricular zone and the fascia dentata at 2 h and 7d and a recovery or overcompensation at 28 d, most pronounced in the rostral periventricular zone (+40%). In the cortex Cu decreased slightly to −10%. Fe increased in the interpeduncular nucleus (+40%) but not in the substantia nigra. This pattern is in line with a differential regulation of periventricular and parenchymal Cu, and with the histochemical localization of Fe, and congruent to regions of preferential MPTP binding described in the rodent brain.
The LA-ICP-MS technique yielded valid and statistically robust results in the present study on 39 slices from 19 animals. Our findings underline the value of routine micro-local analytical techniques in the life sciences and affirm a role of Cu availability in Parkinson’s disease.
Background
This study evaluated the effect of an artificial intelligence (AI)-based clinical decision support system on the performance and diagnostic confidence of endoscopists in their assessment of Barrett’s esophagus (BE).
Methods
96 standardized endoscopy videos were assessed by 22 endoscopists with varying degrees of BE experience from 12 centers. Assessment was randomized into two video sets: group A (review first without AI and second with AI) and group B (review first with AI and second without AI). Endoscopists were required to evaluate each video for the presence of Barrett’s esophagus-related neoplasia (BERN) and then decide on a spot for a targeted biopsy. After the second assessment, they were allowed to change their clinical decision and confidence level.
Results
AI had a stand-alone sensitivity, specificity, and accuracy of 92.2%, 68.9%, and 81.3%, respectively. Without AI, BE experts had an overall sensitivity, specificity, and accuracy of 83.3%, 58.1%, and 71.5%, respectively. With AI, BE nonexperts showed a significant improvement in sensitivity and specificity when videos were assessed a second time with AI (sensitivity 69.8% [95%CI 65.2%–74.2%] to 78.0% [95%CI 74.0%–82.0%]; specificity 67.3% [95%CI 62.5%–72.2%] to 72.7% [95%CI 68.2%–77.3%]). In addition, the diagnostic confidence of BE nonexperts improved significantly with AI.
Conclusion
BE nonexperts benefitted significantly from additional AI. BE experts and nonexperts remained significantly below the stand-alone performance of AI, suggesting that there may be other factors influencing endoscopists’ decisions to follow or discard AI advice.
Effect of AI on performance of endoscopists to detect Barrett neoplasia: A Randomized Tandem Trial
()
Background and study aims
To evaluate the effect of an AI-based clinical decision support system (AI) on the performance and diagnostic confidence of endoscopists during the assessment of Barrett's esophagus (BE).
Patients and Methods
Ninety-six standardized endoscopy videos were assessed by 22 endoscopists from 12 different centers with varying degrees of BE experience.
The assessment was randomized into two video sets: Group A (review first without AI and second with AI) and group B (review first with AI and second without AI). Endoscopists were required to evaluate each video for the presence of Barrett's esophagus-related neoplasia (BERN) and then decide on a spot for a targeted biopsy. After the second assessment, they were allowed to change their clinical decision and confidence level.
Results
AI had a standalone sensitivity, specificity, and accuracy of 92.2%, 68.9%, and 81.6%, respectively. Without AI, BE experts had an overall sensitivity, specificity, and accuracy of 83.3%, 58.1 and 71.5%, respectively. With AI, BE nonexperts showed a significant improvement in sensitivity and specificity when videos were assessed a second time with AI (sensitivity 69.7% (95% CI, 65.2% - 74.2%) to 78.0% (95% CI, 74.0% - 82.0%); specificity 67.3% (95% CI, 62.5% - 72.2%) to 72.7% (95 CI, 68.2% - 77.3%). In addition, the diagnostic confidence of BE nonexperts improved significantly with AI.
Conclusion
BE nonexperts benefitted significantly from the additional AI. BE experts and nonexperts remained below the standalone performance of AI, suggesting that there may be other factors influencing endoscopists to follow or discard AI advice.
Aims
Evaluation of the add-on effect an artificial intelligence (AI) based clinical decision support system has on the performance of endoscopists with different degrees of expertise in the field of Barrett's esophagus (BE) and Barrett's esophagus-related neoplasia (BERN).
Methods
The support system is based on a multi-task deep learning model trained to solve a segmentation and several classification tasks. The training approach represents an extension of the ECMT semi-supervised learning algorithm. The complete system evaluates a decision tree between estimated motion, classification, segmentation, and temporal constraints, to decide when and how the prediction is highlighted to the observer. In our current study, ninety-six video cases of patients with BE and BERN were prospectively collected and assessed by Barrett's specialists and non-specialists. All video cases were evaluated twice – with and without AI assistance. The order of appearance, either with or without AI support, was assigned randomly. Participants were asked to detect and characterize regions of dysplasia or early neoplasia within the video sequences.
Results
Standalone sensitivity, specificity, and accuracy of the AI system were 92.16%, 68.89%, and 81.25%, respectively. Mean sensitivity, specificity, and accuracy of expert endoscopists without AI support were 83,33%, 58,20%, and 71,48 %, respectively. Gastroenterologists without Barrett's expertise but with AI support had a comparable performance with a mean sensitivity, specificity, and accuracy of 76,63%, 65,35%, and 71,36%, respectively.
Conclusions
Non-Barrett's experts with AI support had a similar performance as experts in a video-based study.
Aims
Barrett´s esophagus related neoplasia (BERN) is difficult to detect and characterize during endoscopy, even for expert endoscopists. We aimed to assess the add-on effect of an Artificial Intelligence (AI) algorithm (Barrett-Ampel) as a decision support system (DSS) for non-expert endoscopists in the evaluation of Barrett’s esophagus (BE) and BERN.
Methods
Twelve videos with multimodal imaging white light (WL), narrow-band imaging (NBI), texture and color enhanced imaging (TXI) of histologically confirmed BE and BERN were assessed by expert and non-expert endoscopists. For each video, endoscopists were asked to identify the area of BERN and decide on the biopsy spot. Videos were assessed by the AI algorithm and regions of BERN were highlighted in real-time by a transparent overlay. Finally, endoscopists were shown the AI videos and asked to either confirm or change their initial decision based on the AI support.
Results
Barrett-Ampel correctly identified all areas of BERN, irrespective of the imaging modality (WL, NBI, TXI), but misinterpreted two inflammatory lesions (Accuracy=75%). Expert endoscopists had a similar performance (Accuracy=70,8%), while non-experts had an accuracy of 58.3%. When AI was implemented as a DSS, non-expert endoscopists improved their diagnostic accuracy to 75%.
Conclusions
AI may have the potential to support non-expert endoscopists in the assessment of videos of BE and BERN. Limitations of this study include the low number of videos used. Randomized clinical trials in a real-life setting should be performed to confirm these results.
Pixel-level classification is an essential part of computer vision. For learning from labeled data, many powerful deep learning models have been developed recently. In this work, we augment such supervised segmentation models by allowing them to learn from unlabeled data. Our semi-supervised approach, termed Error-Correcting Supervision, leverages a collaborative strategy. Apart from the supervised training on the labeled data, the segmentation network is judged by an additional network.
We propose an automatic approach for early detection of adenocarcinoma in the esophagus. High-definition endoscopic images (50 cancer, 50 Barrett) are partitioned into a dataset containing approximately equal amounts of patches showing cancerous and non-cancerous regions. A deep convolutional neural network is adapted to the data using a transfer learning approach. The final classification of an image is determined by at least one patch, for which the probability being a cancer patch exceeds a given threshold. The model was evaluated with leave one patient out cross-validation. With sensitivity and specificity of 0.94 and 0.88, respectively, our findings improve recently published results on the same image data base considerably. Furthermore, the visualization of the class probabilities of each individual patch indicates, that our approach might be extensible to the segmentation domain.
Semantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student’s prediction. The combined outputs of the students’ prediction with the teachers’ correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.
We investigate contrastive learning in a multi-task learning setting classifying and segmenting early Barrett’s cancer. How can contrastive learning be applied in a domain with few classes and low inter-class and inter-sample variance, potentially enabling image retrieval or image attribution? We introduce a data sampling strategy that mines per-lesion data for positive samples and keeps a queue of the recent projections as negative samples. We propose a masking strategy for the NT-Xent loss that keeps the negative set pure and removes samples from the same lesion. We show cohesion and uniqueness improvements of the proposed method in feature space. The introduction of the auxiliary objective does not affect the performance but adds the ability to indicate similarity between lesions. Therefore, the approach could enable downstream auto-documentation tasks on homogeneous medical image data.
Real-time computational speed and a high degree of precision are requirements for computer-assisted interventions. Applying a segmentation network to a medical video processing task can introduce significant inter-frame prediction noise. Existing approaches can reduce inconsistencies by including temporal information but often impose requirements on the architecture or dataset. This paper proposes a method to include temporal information in any segmentation model and, thus, a technique to improve video segmentation performance without alterations during training or additional labeling. With Motion-Corrected Moving Average, we refine the exponential moving average between the current and previous predictions. Using optical flow to estimate the movement between consecutive frames, we can shift the prior term in the moving-average calculation to align with the geometry of the current frame. The optical flow calculation does not require the output of the model and can therefore be performed in parallel, leading to no significant runtime penalty for our approach. We evaluate our approach on two publicly available segmentation datasets and two proprietary endoscopic datasets and show improvements over a baseline approach.
Local gray level dependencies of natural images can be modelled by means of co-occurrence matrices containing joint probabilities of gray-level pairs. Texture, however, is a resolution-dependent phenomenon and hence, classification depends on the chosen scale. Since there is no optimal scale for all textures we employ a multiscale approach that acquires textural features at several scales. Thus linear and nonlinear scale-spaces are analyzed by multiscale co-occurrence matrices that describe the statistical behavior of a texture in scale-space. Classification is then performed on the basis of texture features taken from the individual scale with the highest discriminatory power. By considering cross-scale occurrences of gray level pairs, the impact of filters on the feature is described and used for classification of natural textures. This novel method was found to improve classification rates of the common co-occurrence matrix approach on standard textures significantly.
The success of artificial intelligence in medicine is based on the need for large amounts of high quality training data. Sharing of medical image data, however, is often restricted by laws such as doctor-patient confidentiality. Although there are publicly available medical datasets, their quality and quantity are often low. Moreover, datasets are often imbalanced and only represent a fraction of the images generated in hospitals or clinics and can thus usually only be used as training data for specific problems. The introduction of generative adversarial networks (GANs) provides a mean to generate artificial images by training two convolutional networks. This paper proposes a method which uses GANs trained on medical images in order to generate a large number of artificial images that could be used to train other artificial intelligence algorithms. This work is a first step towards alleviating data privacy concerns and being able to publicly share data that still contains a substantial amount of the information in the original private data. The method has been evaluated on several public datasets and quantitative and qualitative tests showing promising results.
Purpose
Congenital breast asymmetry is a serious gynecological malformation for affected patients. The condition hits young women in puberty and is associated with socio-esthetic handicap, depression, and psychosexual problems. Surgical treatment is usually early in the patient's lifetime, so a long-term sustainable solution is important. Although postoperative outcome has been evaluated in several studies before, this study is the first to analyze which objective parameters have the greatest influence on subjective satisfaction with long-term results.
Methods
Thirty-four patients diagnosed with congenital breast asymmetry that underwent either lipofilling or implant therapy between the years of 2008 to 2019 were examined. On average, our collective comprised patients seven years after surgery. Data were mainly gathered through manual measurements, patient-reported outcome measures (Breast Q™), and breast volumetry based on 3D scans (Vectra® H2, Canfield Scientific).
Results
Among all analyzed parameters, only areolar diameter correlated significantly negatively with the subjective outcome satisfaction of the patient. Regarding the subjective assessment of postoperative satisfaction with similarity of the breasts, again the mean areolar diameter, but also the difference in areolar diameter and breast volume between the right and left breasts correlated significantly negatively.
Conclusion
Areolar diameter was revealed as being a significant factor influencing subjective long-term satisfaction in breast asymmetry patients. Moreover, 3D volumetry proves to be an effective tool to substantiate subjective patient assessments. Our findings may lead to further improvements to surgical planning and will be expanded in further studies.
Metals are involved in many processes of life. They are needed for enzymatic reactions, are involved in healthy processes but also yield diseases if the metal homeostasis is disordered. Therefore, the interest to assess the spatial distribution of metals is rising in biomedical science. Imaging metal (and non-metal) isotopes by laser ablation mass spectrometry with inductively coupled plasma (LA-ICP-MS) requires a special software solution to process raw data obtained by scanning a sample line-by-line. As no software ready to use was available we developed an interactive software tool for Image Generation and Analysis (IMAGENA). Unless optimised for LA-ICP-MS, IMAGENA can handle other raw data as well. The general purpose was to reconstruct images from a continuous list of raw data points, to visualise these images, and to convert them into a commonly readable image file format that can be further analysed by standard image analysis software. The generation of the image starts with loading a text file that holds a data column of every measured isotope. Specifying general spatial domain settings like the data offset and the image dimensions is done by the user getting a direct feedback by means of a preview image. IMAGENA provides tools for calibration and to correct for a signal drift in the y-direction. Images are visualised in greyscale as well a pseudo-colours with possibilities for contrast enhancement. Image analysis is performed in terms of smoothed line plots in row and column direction.
GinJinn: An object-detection pipeline for automated feature extraction from herbarium specimens
(2020)
PREMISE:
The generation of morphological data in evolutionary, taxonomic, and ecological studies of plants using herbarium material has traditionally been a labor-intensive task. Recent progress in machine learning using deep artificial neural networks (deep learning) for image classification and object detection has facilitated the establishment of a pipeline for the automatic recognition and extraction of relevant structures in images of herbarium specimens.
METHODS AND RESULTS:
We implemented an extendable pipeline based on state-of-the-art deep-learning object-detection methods to collect leaf images from herbarium specimens of two species of the genus Leucanthemum. Using 183 specimens as the training data set, our pipeline extracted one or more intact leaves in 95% of the 61 test images.
CONCLUSIONS:
We establish GinJinn as a deep-learning object-detection tool for the automatic recognition and extraction of individual leaves or other structures from herbarium specimens. Our pipeline offers greater flexibility and a lower entrance barrier than previous image-processing approaches based on hand-crafted features.
The field of AI is characterized by robust promises, astonishing successes, and remarkable breakthroughs. AI will play a major role in all domains of clinical medicine, but the role of AI in relation to the physician is not yet completely determined. The term artificial intelligence or AI is broad, and several different terms are used in this context that must be organized and demystified. This chapter will review the key concepts and methods of AI, and will introduce some of the different roles for AI in relation to the physician.
Polarised light imaging (PLI) utilises the birefringence of the myelin sheaths in order to visualise the orientation of nerve fibres in microtome sections of adult human post-mortem brains at ultra-high spatial resolution. The preparation of post-mortem brains for PLI involves fixation, freezing and cutting into 100-μm-thick sections. Hence, geometrical distortions of histological sections are inevitable and have to be removed for 3D reconstruction and subsequent fibre tracking. We here present a processing pipeline for 3D reconstruction of these sections using PLI derived multimodal images of post-mortem brains. Blockface images of the brains were obtained during cutting; they serve as reference data for alignment and elimination of distortion artefacts. In addition to the spatial image transformation, fibre orientation vectors were reoriented using the transformation fields, which consider both affine and subsequent non-linear registration. The application of this registration and reorientation approach results in a smooth fibre vector field, which reflects brain morphology. PLI combined with 3D reconstruction and fibre tracking is a powerful tool for human brain mapping. It can also serve as an independent method for evaluating in vivo fibre tractography.
3D rat brain tumors
(2005)
Fusion of histology and MRI is frequently demanded in biomedical research to study in vitro tissue properties in an in vivo reference space. Distortions and artifacts caused by cutting and staining of histological slices as well as differences in spatial resolution make even the rigid fusion a difficult task. State-of- the-art methods start with a mono-modal restacking yielding a histological pseudo-3D volume. The 3D information of the MRI reference is considered subsequently. However, consistency of the histology volume and consistency due to the corresponding MRI seem to be diametral goals. Therefore, we propose a novel fusion framework optimizing histology/histology and histology/MRI consistency at the same time finding a balance between both goals. Method - Direct slice-to-slice correspondence even in irregularly-spaced cutting sequences is achieved by registration-based interpolation of the MRI. Introducing a weighted multi-image mutual information metric (WI), adjacent histology and corresponding MRI are taken into account at the same time. Therefore, the reconstruction of the histological volume as well as the fusion with the MRI is done in a single step. Results - Based on two data sets with more than 110 single registrations in all, the results are evaluated quantitatively based on Tanimoto overlap measures and qualitatively showing the fused volumes. In comparison to other multi-image metrics, the reconstruction based on WI is significantly improved. We evaluated different parameter settings with emphasis on the weighting term steering the balance between intra- and inter-modality consistency.
Objective:
Complex hue/saturation images as a new approach for color texture classification using Gabor filters are introduced and compared with common techniques.
Method:
The interpretation of hue and saturationas polar coordinates allows direct use of the HSV-colorspace for Fourier transform. This technique is applied for Gabor feature extraction of color textures. In contrast to other color features based on the RGB-colorspace [1] the combination of color bands is done previous to the filtering.
Results:
The performance of the new HS-featuresis compared with that of RGB based as well as grayscale Gabor features by evaluating the classifi-cation of 30 natural textures. The new HS-featuresshow same results like the best RGB features but allow a more compact representation. On the averagethe color features improve the results of grayscale features.
Conclusion:
The consideration of the color information enhances the classification of color texture. The choice of colorspace cannot be adjudged finally, but the introduced features suggest the use of the HSV-colorspace with less features than RGB.
Laryngoscopic images of the vocal tract are used for diagnostic purposes. Quantitative mea-surements like changes of the glottis size or the surface of the vocal cords during an image sequence can be helpful to describe the healing process or to compare the findings of diffe-rent patients. Typically the endoscopic images are circulary symmetric distorted (barrel di-stortion). Therefore measurements of geometric dimensions depend on the object´s position in the image. In this paper an algorithm is presented which allows the computation of the translational invariant "real" object size by correcting the image distortion without using additional calibration of the optical environment.
Time-Dependent Joint Probability Speed Function for Level-Set Segmentation of Rat-Brain Slices
(2008)
The segmentation of rat brain slices suffers from illumination inhomogeneities and staining effects. State-of-the-art level-set methods model slice and background with intensity mixture densities defining the speed function as difference between the respective probabilites. Nevertheless, the overlap of these distributions causes an inaccurate stopping at the slice border. In this work, we propose the characterisation of the border area with intensity pairs for inside and outside estimating joint intensity probabilities. Method - In contrast to global object and background models, we focus on the object border characterised by a joint mixture density. This specifies the probability of the occurance of an inside and an outside value in direct adjacency. These values are not known beforehand, because inside and outside depend on the level-set evolution and change during time. Therefore, the speed function is computed time-dependently at the position of the current zero level-set. Along this zero level-set curve, the inside and outside values are derived as mean along the curvature normal directing inside and outside the object. Advantage of the joint probability distribution is to resolve the distribution overlaps, because these are assumed to be not located at the same border position. Results - The novel time-dependent joint probability based speed function is compared expermimentally with single probability based speed functions. Two rat brains with about 40 slices are segmented and the results analysed using manual segmentations and the Tanimoto overlap measure. Improved results are recognised for both data sets.
Introduction – Diagnosis of abnormal cilia function is based on ultrastructural analysis of axoneme defects, especialy the features of inner and outer dynein arms which are the motors of ciliar motility. Sub-optimal biopsy material, methodical, and intrinsic electron microscopy factors pose difficulty in ciliary defects evaluation. We present a computer-assisted approach based on state-of-the-art image analysis and object recognition methods yielding a time-saving and efficient diagnosis of cilia dysfunction. Method – The presented approach is based on a pipeline of basal image processing methods like smoothing, thresholding and ellipse fitting. However, integration of application specific knowledge results in robust segmentations even in cases of image artifacts. The method is build hierarchically starting with the detection of cilia within the image, followed by the detection of nine doublets within each analyzable cilium, and ending with the detection of dynein arms of each doublet. The process is concluded by a rough classification of the dynein arms as basis for a computer-assisted diagnosis. Additionally, the interaction possibilities are designed in a way, that the results are still reproducible given the completion report. Results – A qualitative evaluation showed reasonable detection results for cilia, doublets and dynein arms. However, since a ground truth is missing, the variation of the computer-assisted diagnosis should be within the subjective bias of human diagnosticians. The results of a first quantitative evaluation with five human experts and six images with 12 analyzable cilia showed, that with default parameterization 91.6% of the cilia and 98% of the doublets were found. The computer-assisted approach rated 66% of those inner and outer dynein arms correct, where all human experts agree. However, especially the quality of the dynein arm classification may be improved in future work.
In neuroscience, small-animal studies frequently involve dealing with series of images from multiple modalities such as histology and autoradiography. The consistent and bias-free restacking of multi-modality image series is obligatory as a starting point for subsequent non-rigid registration procedures and for quantitative comparisons with positron emission tomography (PET) and other in vivo data. Up to now, consistency between 2D slices without cross validation using an inherent 3D modality is frequently presumed to be close to the true morphology due to the smooth appearance of the contours of anatomical structures. However, in multi-modality stacks consistency is difficult to assess. In this work, consistency is defined in terms of smoothness of neighboring slices within a single modality and between different modalities. Registration bias denotes the distortion of the registered stack in comparison to the true 3D morphology and shape. Based on these metrics, different restacking strategies of multi-modality rat brain slices are experimentally evaluated. Experiments based on MRI-simulated and real dual-tracer autoradiograms reveal a clear bias of the restacked volume despite quantitatively high consistency and qualitatively smooth brain structures. However, different registration strategies yield different inter-consistency metrics. If no genuine 3D modality is available, the use of the so-called SOP (slice-order preferred) or MOSOP (modality-and-slice-order preferred) strategy is recommended.
The number of patients with Barret’s esophagus (BE) has increased in the last decades. Considering the dangerousness of the disease and its evolution to adenocarcinoma, an early diagnosis of BE may provide a high probability of cancer remission. However, limitations regarding traditional methods of detection and management of BE demand alternative solutions. As such, computer-aided tools have been recently used to assist in this problem, but the challenge still persists. To manage the problem, we introduce the infinity Restricted Boltzmann Machines (iRBMs) to the task of automatic identification of Barrett’s esophagus from endoscopic images of the lower esophagus. Moreover, since iRBM requires a proper selection of its meta-parameters, we also present a discriminative iRBM fine-tuning using six meta-heuristic optimization techniques. We showed that iRBMs are suitable for the context since it provides competitive results, as well as the meta-heuristic techniques showed to be appropriate for such task.
Volumetric representations of autoradiographic and histological images gain ever more interest as a base to interpret data obtained with /spl mu/-imaging devices like microPET. Beyond supporting spatial orientation within rat brains especially autoradiographic images may serve as a base to quantitatively evaluate the complex uptake patterns of microPET studies with receptor ligands or tumor tracers. They may also serve for the development of rat brain atlases or data models, which can be explored during further image analysis or simulation studies. In all cases a consistent spatial representation of the rat brain, i.e. its anatomy and the corresponding quantitative uptake pattern, is required. This includes both, a restacking of the individual two-dimensional images and the exact registration of the respective volumes. We propose strategies how these volumes can be created in a consistent way and trying to limit the requirements on the circumstances during data acquisition, i.e. being independent from other sources like video imaging of the block face prior to cutting or high resolution micro-X-ray CT or micro MRI.
Anatomical differences between individuals are often neglected in musculoskeletal models, but they are necessary in case of subject-specific questions regarding the lumbar spine. A modification of models to each subject is complex and the effects on lumbar loading are difficult to assess. The objective of this thesis is to create a validated musculoskeletal model of the human body, which facilitates a subject-specific modification of the geometry of the lumbar spine. Furthermore, important parameters are identified in sensitivity studies and a case study regarding multifidus muscle atrophy after a disc herniation is conducted. Therefore, a generic model is heavily modified and a semi-automatic process is implemented. This procedure remodels the geometry of the lumbar spine to a subject-specific one on basis of segmented medical images. The resulting five models are validated with regard to the lumbar loading at the L4/L5 level. The influence of lumbar ligament stiffness is determined by changing the stiffness values of all lumbar ligaments in eleven steps during a flexion motion. Sensitivities of lumbar loading to an altered geometry of the lumbar spine are identified by varying ten lumbar parameters in simulations with each model in four postures. The case study includes an analysis of the loading of the multifidus muscle and of the lumbar discs throughout various stages of disc herniation. This time each model performs four motions with two different motion rhythms. The results indicate that lumbar motion and loading is dependent on lumbar ligament stiffness. Furthermore, subject-specific modelling of the lumbar spine should include at least the vertebral height, disc height and lumbar lordosis. The results of the case study suggest that an overloading of the multifidus muscle could follow disc herniation. Additionally, a subsequent atrophy of the muscles could expose adjacent levels to an increased loading, but these findings are highly dependent on the individual.
Celiac disease is an autoimmune disorder caused by gluten that results in an inflammatory response of the small intestine.We investigated whether celiac disease can be detected using endoscopic images through a deep learning approach. The results show that additional clinical parameters can improve the classification accuracy. In this work, we distinguished between healthy tissue and Marsh III, according to the Marsh score system. We first trained a baseline network to classify endoscopic images of the small bowel into these two classes and then augmented the approach with a multimodality component that took the antibody status into account.
Heavy smoke development represents an important challenge for operating physicians during laparoscopic procedures and can potentially affect the success of an intervention due to reduced visibility and orientation. Reliable and accurate recognition of smoke is therefore a prerequisite for the use of downstream systems such as automated smoke evacuation systems. Current approaches distinguish between non-smoked and smoked frames but often ignore the temporal context inherent in endoscopic video data. In this work, we therefore present a method that utilizes the pixel-wise displacement from randomly sampled images to the preceding frames determined using the optical flow algorithm by providing the transformed magnitude of the displacement as an additional input to the network. Further, we incorporate the temporal context at evaluation time by applying an exponential moving average on the estimated class probabilities of the model output to obtain more stable and robust results over time. We evaluate our method on two convolutional-based and one state-of-the-art transformer architecture and show improvements in the classification results over a baseline approach, regardless of the network used.
In the field of computer- and robot-assisted minimally invasive surgery, enormous progress has been made in recent years based on the recognition of surgical instruments in endoscopic images. Especially the determination of the position and type of the instruments is of great interest here. Current work involves both spatial and temporal information with the idea, that the prediction of movement of surgical tools over time may improve the quality of final segmentations. The provision of publicly available datasets has recently encouraged the development of new methods, mainly based on deep learning. In this review, we identify datasets used for method development and evaluation, as well as quantify their frequency of use in the literature. We further present an overview of the current state of research regarding the segmentation and tracking of minimally invasive surgical instruments in endoscopic images. The paper focuses on methods that work purely visually without attached markers of any kind on the instruments, taking into account both single-frame segmentation approaches as well as those involving temporal information. A discussion of the reviewed literature is provided, highlighting existing shortcomings and emphasizing available potential for future developments. The publications considered were identified through the platforms Google Scholar, Web of Science, and PubMed. The search terms used were "instrument segmentation", "instrument tracking", "surgical tool segmentation", and "surgical tool tracking" and result in 408 articles published between 2015 and 2022 from which 109 were included using systematic selection criteria.
Digital nerve lesions result in a loss of tactile sensation reflected by an anesthetic area (AA) at the radial or ulnar aspect of the respective digit. Yet, available tools to monitor the recovery of tactile sense have been criticized for their lack of validity. However, the precise quantification of AA dynamics by three-dimensional (3-D) imaging could serve as an accurate surrogate to monitor recovery following digital nerve repair.
For validation, AAs were marked on digits of healthy volunteers to simulate the AA of an impaired cutaneous innervation. Three dimensional models were composed from raw images that had been acquired with a 3-D camera (Vectra H2) to precisely quantify relative AA for each digit (3-D models, n= 80). Operator properties varied regarding individual experience in 3-D imaging and image processing. Additionally, the concept was applied in a clinical case study.
Images taken by experienced photographers were rated better quality (p< 0.001) and needed less processing time (p= 0.020). Quantification of the relative AA was neither altered significantly by experience levels of the photographer (p= 0.425) nor the image assembler (p= 0.749).
The proposed concept allows precise and reliable surface quantification of digits and can be performed consistently without relevant distortion by lack of examiner experience. Routine 3-D imaging of the AA has the great potential to provide visual evidence of various returning states of sensation and to convert sensory nerve recovery into a metric variable with high responsiveness to temporal progress.
The endoscopic features associated with eosinophilic esophagitis (EoE) may be missed during routine endoscopy. We aimed to develop and evaluate an Artificial Intelligence (AI) algorithm for detecting and quantifying the endoscopic features of EoE in white light images, supplemented by the EoE Endoscopic Reference Score (EREFS). An AI algorithm (AI-EoE) was constructed and trained to differentiate between EoE and normal esophagus using endoscopic white light images extracted from the database of the University Hospital Augsburg. In addition to binary classification, a second algorithm was trained with specific auxiliary branches for each EREFS feature (AI-EoE-EREFS). The AI algorithms were evaluated on an external data set from the University of North Carolina, Chapel Hill (UNC), and compared with the performance of human endoscopists with varying levels of experience. The overall sensitivity, specificity, and accuracy of AI-EoE were 0.93 for all measures, while the AUC was 0.986. With additional auxiliary branches for the EREFS categories, the AI algorithm (AI-EoEEREFS) performance improved to 0.96, 0.94, 0.95, and 0.992 for sensitivity, specificity, accuracy, and AUC, respectively. AI-EoE and AI-EoE-EREFS performed significantly better than endoscopy beginners and senior fellows on the same set of images. An AI algorithm can be trained to detect and quantify endoscopic features of EoE with excellent performance scores. The addition of the EREFS criteria improved the performance of the AI algorithm, which performed significantly better than endoscopists with a lower or medium experience level.
Aims
Eosinophilic esophagitis (EoE) is easily missed during endoscopy, either because physicians are not familiar with its endoscopic features or the morphologic changes are too subtle. In this preliminary paper, we present the first attempt to detect EoE in endoscopic white light (WL) images using a deep learning network (EoE-AI).
Methods
401 WL images of eosinophilic esophagitis and 871 WL images of normal esophageal mucosa were evaluated. All images were assessed for the Endoscopic Reference score (EREFS) (edema, rings, exudates, furrows, strictures). Images with strictures were excluded. EoE was defined as the presence of at least 15 eosinophils per high power field on biopsy. A convolutional neural network based on the ResNet architecture with several five-fold cross-validation runs was used. Adding auxiliary EREFS-classification branches to the neural network allowed the inclusion of the scores as optimization criteria during training. EoE-AI was evaluated for sensitivity, specificity, and F1-score. In addition, two human endoscopists evaluated the images.
Results
EoE-AI showed a mean sensitivity, specificity, and F1 of 0.759, 0.976, and 0.834 respectively, averaged over the five distinct cross-validation runs. With the EREFS-augmented architecture, a mean sensitivity, specificity, and F1-score of 0.848, 0.945, and 0.861 could be demonstrated respectively. In comparison, the two human endoscopists had an average sensitivity, specificity, and F1-score of 0.718, 0.958, and 0.793.
Conclusions
To the best of our knowledge, this is the first application of deep learning to endoscopic images of EoE which were also assessed after augmentation with the EREFS-score. The next step is the evaluation of EoE-AI using an external dataset. We then plan to assess the EoE-AI tool on endoscopic videos, and also in real-time. This preliminary work is encouraging regarding the ability for AI to enhance physician detection of EoE, and potentially to do a true “optical biopsy” but more work is needed.
In the field of computer- and robot-assisted minimally invasive surgery, enormous progress has been made in recent years based on the recognition of surgical instruments in endoscopic images and videos. In particular, the determination of the position and type of instruments is of great interest. Current work involves both spatial and temporal information, with the idea that predicting the movement of surgical tools over time may improve the quality of the final segmentations. The provision of publicly available datasets has recently encouraged the development of new methods, mainly based on deep learning. In this review, we identify and characterize datasets used for method development and evaluation and quantify their frequency of use in the literature. We further present an overview of the current state of research regarding the segmentation and tracking of minimally invasive surgical instruments in endoscopic images and videos. The paper focuses on methods that work purely visually, without markers of any kind attached to the instruments, considering both single-frame semantic and instance segmentation approaches, as well as those that incorporate temporal information. The publications analyzed were identified through the platforms Google Scholar, Web of Science, and PubMed. The search terms used were “instrument segmentation”, “instrument tracking”, “surgical tool segmentation”, and “surgical tool tracking”, resulting in a total of 741 articles published between 01/2015 and 07/2023, of which 123 were included using systematic selection criteria. A discussion of the reviewed literature is provided, highlighting existing shortcomings and emphasizing the available potential for future developments.
Aims
VA is an endoscopic finding of celiac disease (CD), which can easily be missed if pretest probability is low. In this study, we aimed to develop an artificial intelligence (AI) algorithm for the detection of villous atrophy on endoscopic images.
Methods
858 images from 182 patients with VA and 846 images from 323 patients with normal duodenal mucosa were used for training and internal validation of an AI algorithm (ResNet18). A separate dataset was used for external validation, as well as determination of detection performance of experts, trainees and trainees with AI support. According to the AI consultation distribution, images were stratified into “easy” and “difficult”.
Results
Internal validation showed 82%, 85% and 84% for sensitivity, specificity and accuracy. External validation showed 90%, 76% and 84%. The algorithm was significantly more sensitive and accurate than trainees, trainees with AI support and experts in endoscopy. AI support in trainees was associated with significantly improved performance. While all endoscopists showed significantly lower detection for “difficult” images, AI performance remained stable.
Conclusions
The algorithm outperformed trainees and experts in sensitivity and accuracy for VA detection. The significant improvement with AI support suggests a potential clinical benefit. Stable performance of the algorithm in “easy” and “difficult” test images may indicate an advantage in macroscopically challenging cases.
Clinical setting
Third space procedures such as endoscopic submucosal dissection (ESD) and peroral endoscopic myotomy (POEM) are complex minimally invasive techniques with an elevated risk for operator-dependent adverse events such as bleeding and perforation. This risk arises from accidental dissection into the muscle layer or through submucosal blood vessels as the submucosal cutting plane within the expanding resection site is not always apparent. Deep learning algorithms have shown considerable potential for the detection and characterization of gastrointestinal lesions. So-called AI – clinical decision support solutions (AI-CDSS) are commercially available for polyp detection during colonoscopy. Until now, these computer programs have concentrated on diagnostics whereas an AI-CDSS for interventional endoscopy has not yet been introduced. We aimed to develop an AI-CDSS („Smart ESD“) for real-time intra-procedural detection and delineation of blood vessels, tissue structures and endoscopic instruments during third-space endoscopic procedures.
Characteristics of Smart ESD
An AI-CDSS was invented that delineates blood vessels, tissue structures and endoscopic instruments during third-space endoscopy in real-time. The output can be displayed by an overlay over the endoscopic image with different modes of visualization, such as a color-coded semitransparent area overlay, or border tracing (demonstration video). Hereby the optimal layer for dissection can be visualized, which is close above or directly at the muscle layer, depending on the applied technique (ESD or POEM). Furthermore, relevant blood vessels (thickness> 1mm) are delineated. Spatial proximity between the electrosurgical knife and a blood vessel triggers a warning signal. By this guidance system, inadvertent dissection through blood vessels could be averted.
Technical specifications
A DeepLabv3+ neural network architecture with KSAC and a 101-layer ResNeSt backbone was used for the development of Smart ESD. It was trained and validated with 2565 annotated still images from 27 full length third-space endoscopic videos. The annotation classes were blood vessel, submucosal layer, muscle layer, electrosurgical knife and endoscopic instrument shaft. A test on a separate data set yielded an intersection over union (IoU) of 68%, a Dice Score of 80% and a pixel accuracy of 87%, demonstrating a high overlap between expert and AI segmentation. Further experiments on standardized video clips showed a mean vessel detection rate (VDR) of 85% with values of 92%, 70% and 95% for POEM, rectal ESD and esophageal ESD respectively. False positive measurements occurred 0.75 times per minute. 7 out of 9 vessels which caused intraprocedural bleeding were caught by the algorithm, as well as both vessels which required hemostasis via hemostatic forceps.
Future perspectives
Smart ESD performed well for vessel and tissue detection and delineation on still images, as well as on video clips. During a live demonstration in the endoscopy suite, clinical applicability of the innovation was examined. The lag time for processing of the live endoscopic image was too short to be visually detectable for the interventionist. Even though the algorithm could not be applied during actual dissection by the interventionist, Smart ESD appeared readily deployable during visual assessment by ESD experts. Therefore, we plan to conduct a clinical trial in order to obtain CE-certification of the algorithm. This new technology may improve procedural safety and speed, as well as training of modern minimally invasive endoscopic resection techniques.
ARTIFICIAL INTELLIGENCE (AI) – ASSISTED VESSEL AND TISSUE RECOGNITION IN THIRD-SPACE ENDOSCOPY
(2022)
Aims
Third-space endoscopy procedures such as endoscopic submucosal dissection (ESD) and peroral endoscopic myotomy (POEM) are complex interventions with elevated risk of operator-dependent adverse events, such as intra-procedural bleeding and perforation. We aimed to design an artificial intelligence clinical decision support solution (AI-CDSS, “Smart ESD”) for the detection and delineation of vessels, tissue structures, and instruments during third-space endoscopy procedures.
Methods
Twelve full-length third-space endoscopy videos were extracted from the Augsburg University Hospital database. 1686 frames were annotated for the following categories: Submucosal layer, blood vessels, electrosurgical knife and endoscopic instrument. A DeepLabv3+neural network with a 101-layer ResNet backbone was trained and validated internally. Finally, the ability of the AI system to detect visible vessels during ESD and POEM was determined on 24 separate video clips of 7 to 46 seconds duration and showing 33 predefined vessels. These video clips were also assessed by an expert in third-space endoscopy.
Results
Smart ESD showed a vessel detection rate (VDR) of 93.94%, while an average of 1.87 false positive signals were recorded per minute. VDR of the expert endoscopist was 90.1% with no false positive findings. On the internal validation data set using still images, the AI system demonstrated an Intersection over Union (IoU), mean Dice score and pixel accuracy of 63.47%, 76.18% and 86.61%, respectively.
Conclusions
This is the first AI-CDSS aiming to mitigate operator-dependent limitations during third-space endoscopy. Further clinical trials are underway to better understand the role of AI in such procedures.
Aims
AI has proven great potential in assisting endoscopists in diagnostics, however its role in therapeutic endoscopy remains unclear. Endoscopic submucosal dissection (ESD) is a technically demanding intervention with a slow learning curve and relevant risks like bleeding and perforation. Therefore, we aimed to develop an algorithm for the real-time detection and delineation of relevant structures during third-space endoscopy.
Methods
5470 still images from 59 full length videos (47 ESD, 12 POEM) were annotated. 179681 additional unlabeled images were added to the training dataset. Consequently, a DeepLabv3+ neural network architecture was trained with the ECMT semi-supervised algorithm (under review elsewhere). Evaluation of vessel detection was performed on a dataset of 101 standardized video clips from 15 separate third-space endoscopy videos with 200 predefined blood vessels.
Results
Internal validation yielded an overall mean Dice score of 85% (68% for blood vessels, 86% for submucosal layer, 88% for muscle layer). On the video test data, the overall vessel detection rate (VDR) was 94% (96% for ESD, 74% for POEM). The median overall vessel detection time (VDT) was 0.32 sec (0.3 sec for ESD, 0.62 sec for POEM).
Conclusions
Evaluation of the developed algorithm on a video test dataset showed high VDR and quick VDT, especially for ESD. Further research will focus on a possible clinical benefit of the AI application for VDR and VDT during third-space endoscopy.
Aims
While AI has been successfully implemented in detecting and characterizing colonic polyps, its role in therapeutic endoscopy remains to be elucidated. Especially third space endoscopy procedures like ESD and peroral endoscopic myotomy (POEM) pose a technical challenge and the risk of operator-dependent complications like intraprocedural bleeding and perforation. Therefore, we aimed at developing an AI-algorithm for intraprocedural real time vessel detection during ESD and POEM.
Methods
A training dataset consisting of 5470 annotated still images from 59 full-length videos (47 ESD, 12 POEM) and 179681 unlabeled images was used to train a DeepLabV3+neural network with the ECMT semi-supervised learning method. Evaluation for vessel detection rate (VDR) and time (VDT) of 19 endoscopists with and without AI-support was performed using a testing dataset of 101 standardized video clips with 200 predefined blood vessels. Endoscopists were stratified into trainees and experts in third space endoscopy.
Results
The AI algorithm had a mean VDR of 93.5% and a median VDT of 0.32 seconds. AI support was associated with a statistically significant increase in VDR from 54.9% to 73.0% and from 59.0% to 74.1% for trainees and experts, respectively. VDT significantly decreased from 7.21 sec to 5.09 sec for trainees and from 6.10 sec to 5.38 sec for experts in the AI-support group. False positive (FP) readings occurred in 4.5% of frames. FP structures were detected significantly shorter than true positives (0.71 sec vs. 5.99 sec).
Conclusions
AI improved VDR and VDT of trainees and experts in third space endoscopy and may reduce performance variability during training. Further research is needed to evaluate the clinical impact of this new technology.
Aims
Celiac disease (CD) is a complex condition caused by an autoimmune reaction to ingested gluten. Due to its polymorphic manifestation and subtle endoscopic presentation, the diagnosis is difficult and thus the disorder is underreported. We aimed to use deep learning to identify celiac disease on endoscopic images of the small bowel.
Methods
Patients with small intestinal histology compatible with CD (MARSH classification I-III) were extracted retrospectively from the database of Augsburg University hospital. They were compared to patients with no clinical signs of CD and histologically normal small intestinal mucosa. In a first step MARSH III and normal small intestinal mucosa were differentiated with the help of a deep learning algorithm. For this, the endoscopic white light images were divided into five equal-sized subsets. We avoided splitting the images of one patient into several subsets. A ResNet-50 model was trained with the images from four subsets and then validated with the remaining subset. This process was repeated for each subset, such that each subset was validated once. Sensitivity, specificity, and harmonic mean (F1) of the algorithm were determined.
Results
The algorithm showed values of 0.83, 0.88, and 0.84 for sensitivity, specificity, and F1, respectively. Further data showing a comparison between the detection rate of the AI model and that of experienced endoscopists will be available at the time of the upcoming conference.
Conclusions
We present the first clinical report on the use of a deep learning algorithm for the detection of celiac disease using endoscopic images. Further evaluation on an external data set, as well as in the detection of CD in real-time, will follow. However, this work at least suggests that AI can assist endoscopists in the endoscopic diagnosis of CD, and ultimately may be able to do a true optical biopsy in live-time.