Refine
Year of publication
Document Type
- Article (94)
- In Proceedings (83)
- ZIB-Report (29)
- Book chapter (8)
- Poster (6)
- Other (5)
- Research data (4)
- Doctoral Thesis (2)
- In Collection (2)
- Master's Thesis (1)
Is part of the Bibliography
- no (234)
Keywords
- statistical shape and intensity models (5)
- Statistical shape analysis (3)
- 3d-reconstruction from 2d X-rays (2)
- 3d-reconstruction from 2d Xrays (2)
- Articulated Models, Statistical Shape And Intensity Models, 2D/3D Anatomy Reconstruction, Pelvic Parameters Measurement, Total Hip Arthroplasty (2)
- Classification (2)
- Kniegelenk (2)
- Lie groups (2)
- Manifold valued statistics (2)
- Medizinische Planung (2)
Institute
- Visual and Data-centric Computing (219)
- Visual Data Analysis (210)
- Therapy Planning (181)
- Numerical Mathematics (19)
- Computational Medicine (15)
- Visual Data Analysis in Science and Engineering (11)
- Geometric Data Analysis and Processing (6)
- Mathematics for Life and Materials Science (4)
- ZIB Allgemein (3)
- Computational Systems Biology (1)
Purpose
Segmentation of surgical instruments in endoscopic video streams is essential for automated surgical scene understanding and process modeling. However, relying on fully supervised deep learning for this task is challenging because manual annotation occupies valuable time of the clinical experts.
Methods
We introduce a teacher–student learning approach that learns jointly from annotated simulation data and unlabeled real data to tackle the challenges in simulation-to-real unsupervised domain adaptation for endoscopic image segmentation.
Results
Empirical results on three datasets highlight the effectiveness of the proposed framework over current approaches for the endoscopic instrument segmentation task. Additionally, we provide analysis of major factors affecting the performance on all datasets to highlight the strengths and failure modes of our approach.
Conclusions
We show that our proposed approach can successfully exploit the unlabeled real endoscopic video frames and improve generalization performance over pure simulation-based training and the previous state-of-the-art. This takes us one step closer to effective segmentation of surgical instrument in the annotation scarce setting.
The reconstruction of an object's shape or surface from a set of 3D points is a common topic in materials and life sciences, computationally handled in computer graphics. Such points usually stem from optical or tactile 3D coordinate measuring equipment. Surface reconstruction also appears in medical image analysis, e.g. in anatomy reconstruction from tomographic measurements or the alignment of intra-operative navigation and preoperative planning data. In contrast to mere 3D point clouds, medical imaging yields contextual information on the 3D point data that can be used to adopt prior information on the shape that is to be reconstructed from the measurements. In this work we propose to use a statistical shape model (SSM) as a prior for surface reconstruction. The prior knowledge is represented by a point distribution model (PDM) that is associated with a surface mesh. Using the shape distribution that is modelled by the PDM, we reformulate the problem of surface reconstruction from a probabilistic perspective based on a Gaussian Mixture Model (GMM). In order to do so, the given measurements are interpreted as samples of the GMM. By using mixture components with anisotropic covariances that are oriented according to the surface normals at the PDM points, a surface-based tting is accomplished. By estimating the parameters of the GMM in a maximum a posteriori manner, the reconstruction of the surface from the given measurements is achieved. Extensive experiments suggest that our proposed approach leads to superior surface reconstructions compared to Iterative Closest Point (ICP) methods.
The reconstruction of an object’s shape or surface from a set of 3D points plays an important role in medical image analysis, e.g. in anatomy reconstruction from tomographic measurements or in the process of aligning intra-operative navigation and preoperative planning data. In such scenarios, one usually has to deal with sparse data, which significantly aggravates the problem of reconstruction. However, medical applications often provide contextual information about the 3D point data that allow to incorporate prior knowledge about the shape that is to be reconstructed. To this end, we propose the use of a statistical shape model (SSM) as a prior for surface reconstruction. The SSM is represented by a point distribution model (PDM), which is associated with a surface mesh. Using the shape distribution that is modelled by the PDM, we formulate the problem of surface reconstruction from a probabilistic perspective based on a Gaussian Mixture Model (GMM). In order to do so, the given points are interpreted as samples of the GMM. By using mixture components with anisotropic covariances that are “oriented” according to the surface normals at the PDM points, a surface-based fitting is accomplished. Estimating the parameters of the GMM in a maximum a posteriori manner yields the reconstruction of the surface from the given data points. We compare our method to the extensively used Iterative Closest Points method on several different anatomical datasets/SSMs (brain, femur, tibia, hip, liver) and demonstrate superior accuracy and robustness on sparse data.
Deformable model-based approaches to 3D image segmentation have been shown to be highly successful. Such methodology requires an appearance model that drives the deformation of a geometric model to the image data. Appearance models are usually either created heuristically or through supervised learning. Heuristic methods have been shown to work effectively in many applications but are hard to transfer from one application (imaging modality/anatomical structure) to another. On the contrary, supervised learning approaches can learn patterns from a collection of annotated training data. In this work, we show that the supervised joint dictionary learning technique is capable of overcoming the traditional drawbacks of the heuristic approaches. Our evaluation based on two different applications (liver/CT and knee/MR) reveals that our approach generates appearance models, which can be used effectively and efficiently in a deformable model-based segmentation framework.
We present a novel approach for nonlinear statistical shape modeling that is invariant under Euclidean motion and thus alignment-free. By analyzing metric distortion and curvature of shapes as elements of Lie groups in a consistent Riemannian setting, we construct a framework that reliably handles large deformations. Due to the explicit character of Lie group operations, our non-Euclidean method is very efficient allowing for fast and numerically robust processing. This facilitates Riemannian analysis of large shape populations accessible through longitudinal and multi-site imaging studies providing increased statistical power. Additionally, as planar configurations form a submanifold in shape space, our representation allows for effective estimation of quasi-isometric surfaces flattenings. We evaluate the performance of our model w.r.t. shape-based classification of hippocampus and femur malformations due to Alzheimer's disease and osteoarthritis, respectively. In particular, we achieve state-of-the-art accuracies outperforming the standard Euclidean as well as a recent nonlinear approach especially in presence of sparse training data. To provide insight into the model's ability of capturing biological shape variability, we carry out an analysis of specificity and generalization ability.
THIS MODEL IS FOR NON-COMMERCIAL RESEARCH PURPOSES. ONLY MEMBERS OF UNIVERSITIES OR NON-COMMERCIAL RESEARCH INSTITUTES ARE ELIGIBLE TO APPLY.
1. Download, fill, and sign the form available from:
https://media.githubusercontent.com/media/mgrewe/ovmf/main/data/fexmm_license_agreement.pdf
2. Send the signed form to: fexmm@zib.de
NOTE: Use an official email address of your institution for the request.
It has been recently suggested that mechanical loads applied at frequencies close to the natural frequencies of bone could enhance bone apposition due to the resonance phenomenon. Other applications of bone modal analysis are also suggested. For the above-mentioned applications, it is important to understand how patient-specific bone shape and density distribution influence the natural frequencies of bones. We used finite element models to study the effects of bone shape and density distribution on the natural frequencies of the femur in free boundary conditions. A statistical shape and appearance model that describes shape and density distribution independently was created, based on a training set of 27 femora. The natural frequencies were then calculated for different shape modes varied around the mean shape while keeping the mean density distribution, for different appearance modes around the mean density distribution while keeping the mean bone shape, and for the 27 training femora. Single shape or appearance modes could cause up to 15% variations in the natural frequencies with certain modes having the greatest impact. For the actual femora, shape and
density distribution changed the natural frequencies by up to 38%. First appearance mode that describes the general cortical bone thickness and trabecular bone density had one of the strongest impacts. The first appearance mode could therefore provide a sensitive measure of general bone health and disease progression. Since shape and density could cause large variations in the calculated natural frequencies, patient-specific FE models are needed for accurate estimation of bone natural frequencies.
Realistic prediction of individual facial emotion expressions for craniofacial surgery simulations
(2003)
Purpose: Morphological data pertaining to the pelvis and lower extremity muscles are increasingly being used in biomechanical modeling to compare healthy and pathological conditions. Very few data sets exist that encompass all of the muscles of the lower limb, allowing for comparisons between regions. The aims of this study were to (a) provide physiological cross-sectional area (PCSA) data for the pelvic, thigh, and leg muscles in young, healthy participants, using magnetic resonance imaging (MRI), and (b) to compare these data with summarized PCSAs obtained from the literature.
Materials and Methods: Six young and healthy volunteers participated and were scanned using 3 T MRI. PCSAs were calculated from volumetric segmentations obtained bilaterally of 28 muscles/muscle groups of the pelvis, thigh, and leg. These data were compared to published, summarized PCSA data derived from cadaveric, computed tomography, MRI and ultrasound studies.
Results: The PCSA of the pelvis, thigh, and leg muscles tended to be 20–130% larger in males than in females, except for the gemelli which were 34% smaller in males, and semitendinosus and triceps surae which did not differ (<20% different). The dominant and the non-dominant sides showed similar and minutely different PCSA with less than 18% difference between sides. Comparison to other studies revealed wide ranges within, and large differences between, the cadaveric and imaging PCSA data. Comparison of the PCSA of this study and published literature revealed major differences in the iliopsoas, gluteus minimus, tensor fasciae latae, gemelli, obturator internus, biceps femoris, quadriceps femoris, and the deep leg flexor muscles.
Conclusions: These volume-derived PCSAs of the pelvic and lower limb muscles alongside the data synthesised from the literature may serve as a basis for comparative and biomechanical studies of the living and healthy young, and enable calculation of muscle forces. Comparison of the literature revealed large variations in PCSA from each of the different investigative modalities, hampering omparability between studies. Sample size, age, post-mortem changes of muscle tone, chemical fixation of cadaveric tissues, and the underlying physics of the imaging techniques may potentially influence PCSA calculations.
Patientenspezifische Simulationsmodelle für die funktionelle Analyse von künstlichem Gelenkersatz
(2012)
Background: Currently, there is no fully sufficient way to differentiate between symptomatic and normal nasal breathing. Using the nose’s total resistance is disputed as a valid means to objectify nasal airflow, and the need for a more comprehensive diagnostic method is increasing. This work’s aim was to test a novel approach considering intranasal wall shear stress as well as static pressure maps obtained by computational fluid dynamics (CFD).
Methods: X-ray computed tomography (CT) scan data of six symptom-free subjects and seven symptomatic patients were used. Patient-specific geometries of the nasal cavity were segmented from these data sets. Inspiratory and expiratory steady airflow simulations were performed using CFD. Calculated static pressures and wall shear stresses (WSS) were mapped onto a common template of the nasal septum, allowing for comparison of these parameters between the two patient groups.
Results: Significant differences in wall shear stress distributions during the inspiratory phase could be identified between the two groups, whereas no differences were found for the expiratory phase. It is assumed that one essential feature of normal nasal breathing probably consists in distinctively different intranasal flow fields for inspiration and expiration. This is in accordance with previous investigations.
Conclusion: The proposed method seems to be a promising tool for developing a new kind of patient-specific assessment of nasal breathing. However, more studies and a greater case number of data with an expanded focus, would be ideal.
Several non-invasive methods are common practice in natural sciences today. Here we present how they can be applied and contribute to current topics in cephalopod (paleo-) biology. Different methods will be compared in terms of time necessary to acquire the data, amount of data, accuracy/resolution, minimum/maximum size of objects that can be studied, the degree of post-processing needed and availability. The main application of the methods is seen in morphometry and volumetry of cephalopod shells. In particular we present a method for precise buoyancy calculation. Therefore, cephalopod shells were scanned together with different reference bodies, an approach developed in medical sciences. It is necessary to know the volume of the reference bodies, which should have similar absorption properties like the object of interest. Exact volumes can be obtained from surface scanning. Depending on the dimensions of the study object different computed tomography techniques were applied.
Several non-invasive methods are common practice in natural sciences today. Here we present how they can be applied and contribute to current topics in cephalopod (paleo-) biology. Different methods will be compared in terms of time necessary to acquire the data, amount of data, accuracy/resolution, minimum/maximum size of objects that can be studied, the degree of post-processing needed and availability. The main application of the methods is seen in morphometry and volumetry of cephalopod shells. In particular we present a method for precise buoyancy calculation. Therefore, cephalopod shells were scanned together with different reference bodies, an approach developed in medical sciences. It is necessary to know the volume of the reference bodies, which should have similar absorption properties like the object of interest. Exact volumes can be obtained from surface scanning. Depending on the dimensions of the study object different computed tomography techniques were applied.
Non-invasive approach to shed new light on the buoyancy business of chambered cephalopods (Mollusca)
(2011)
Background: Fossil ticks are extremely rare, whereby Ixodes succineus Weidner, 1964 from Eocene (ca. 44-49 Ma) Baltic amber is one of the oldest examples of a living hard tick genus (Ixodida: Ixodidae). Previous work suggested it was most closely related to the modern and widespread European sheep tick Ixodes ricinus (Linneaus, 1758).
Results: Restudy using phase contrast synchrotron x-ray tomography yielded images of exceptional quality. These confirm the fossil's referral to Ixodes Latreille, 1795, but the characters resolved here suggest instead affinities with the Asian subgenus Partipalpiger Hoogstraal et al., 1973 and its single living (and medically significant) species Ixodes ovatus Neumann, 1899. We redescribe the amber fossil here as Ixodes (Partipalpiger) succineus.
Conclusions: Our data suggest that Ixodes ricinus is unlikely to be directly derived from Weidner's amber species, but instead reveals that the Partipalpiger lineage was originally more widely distributed across the northern hemisphere. The closeness of Ixodes (P.) succineus to a living vector of a wide range of pathogens offers the potential to correlate its spatial and temporal position (northern Europe, nearly 50 million years ago) with the estimated origination dates of various tick-borne diseases.
Polygonale Schädelmodelle bilden ein wichtiges Hilfsmittel für computergestützte Planungen im Bereich der plastischen Chirurgie. Wir beschreiben, wie derartige Modelle automatisch aus hochaufgelösten CT-Datensätzen erzeugt werden können. Durch einen lokal steuerbaren Simplifizierungsalgorithmus werden die Modelle so weit vereinfacht, daß auch auf kleineren Graphikcomputern interaktives Arbeiten möglich wird. Die Verwendung eines speziellen Transparenzmodells ermöglicht den ungehinderten Blick auf die bei der Planung relevanten Knochenstrukturen und läßt den Benutzer zugleich die Kopfumrisse des Patienten erkennen.
Der Artikel gibt einen Einblick in das reiche Feld der Zusammenarbeit zwischen Mathematik und Medizin. Beispielhaft werden drei Erfolgsmodelle dargestellt: Medizinische Bildgebung, mathematische Modellierung und Biosignalverarbeitung im Bereich der Dynamik des Herzens sowie mathematische Modellierung und Simulation in der Krebstherapie Hyperthermie und der Mund-Kiefer-Gesichts-Chirurgie. In allen Fällen existiert ein Gleichklang der Interessen von Medizin und Mathematik: Beide Disziplinen wollen die Resultate schnell und zuverlässig. Für die Klinik heißt das, dass notwendige Rechnungen in möglichst kurzer Zeit, und zwar auf dem PC, ablaufen müssen und dass die Resultate so genau und belastbar sein müssen, dass medizinische Entscheidungen darauf aufbauen können. Für die Mathematik folgt daraus, dass höchste Anforderungen an die Effizienz der verwendeten Algorithmen und die darauf aufbauende Software in Numerik und Visualisierung zu stellen sind. Jedes Kapitel endet mit einer Darstellung der Perspektive des jeweiligen Gebietes. Abschließend werden mögliche Handlungsoptionen für Politik und Wirtschaft diskutiert.
Maxillofacial surgery planning with 3D soft tissue prediction - modeling, planning, simulation
(2005)
In dentistry, software-based medical image analysis and visualization provide efficient and accurate diagnostic and therapy planning capabilities. We present an approach for the automatic recognition of tooth types and positions in digital volume tomography (DVT). By using deep learning techniques in combination with dimensionality reduction through non-planar reformatting of the jaw anatomy, DVT data can be efficiently processed and
teeth reliably recognized and classified, even in the presence of imaging artefacts, missing or dislocated teeth.
We evaluated our approach, which is based on 2D Convolutional Neural Networks (CNNs), on 118 manually
annotated cases of clinical DVT datasets. Our proposed method correctly classifies teeth with an accuracy of
94% within a limit of 2mm distance to ground truth labels.
Reconstructing anatomical shapes from sparse or partial measurements relies on prior knowledge of shape variations that occur within a given population. Such shape priors are learned from example shapes, obtained by segmenting volumetric medical images. For existing models, the resolution of a learned shape prior is limited to the resolution of the training data. However, in clinical practice, volumetric images are often acquired with highly anisotropic voxel sizes, e.g. to reduce image acquisition time in MRI or radiation exposure in CT imaging. The missing shape information between the slices prohibits existing methods to learn a high-resolution shape prior. We introduce a method for high-resolution shape reconstruction from sparse measurements without relying on high-resolution ground truth for training. Our method is based on neural implicit shape representations and learns a continuous shape prior only from highly anisotropic segmentations. Furthermore, it is able to learn from shapes with a varying field of view and can reconstruct from various sparse input configurations. We demonstrate its effectiveness on two anatomical structures: vertebra and femur, and successfully reconstruct high-resolution shapes from sparse segmentations, using as few as three orthogonal slices.
Statistical shape models are an essential tool for various tasks in medical image analysis, including shape generation, reconstruction and classification. Shape models are learned from a population of example shapes, which are typically obtained through segmentation of volumetric medical images. In clinical practice, highly anisotropic volumetric scans with large slice distances are prevalent, e.g., to reduce radiation exposure in CT or image acquisition time in MR imaging. For existing shape modeling approaches, the resolution of the emerging model is limited to the resolution of the training shapes. Therefore, any missing information between slices prohibits existing methods from learning a high-resolution shape prior. We propose a novel shape modeling approach that can be trained on sparse, binary segmentation masks with large slice distances. This is achieved through employing continuous shape representations based on neural implicit functions. After training, our model can reconstruct shapes from various sparse inputs at high target resolutions beyond the resolution of individual training examples. We successfully reconstruct high-resolution shapes from as few as three orthogonal slices. Furthermore, our shape model allows us to embed various sparse segmentation masks into a common, low-dimensional latent space — independent of the acquisition direction, resolution, spacing, and field of view. We show that the emerging latent representation discriminates between healthy and pathological shapes, even when provided with sparse segmentation masks. Lastly, we qualitatively demonstrate that the emerging latent space is smooth and captures characteristic modes of shape variation. We evaluate our shape model on two anatomical structures: the lumbar vertebra and the distal femur, both from publicly available datasets.
Statistical shape modeling aims at capturing shape variations of an anatomical structure that occur within a given population. Shape models are employed in many tasks, such as shape reconstruction and image segmentation, but also shape generation and classification. Existing shape priors either require dense correspondence between training examples or lack robustness and topological guarantees. We present FlowSSM, a novel shape modeling approach that learns shape variability without requiring dense correspondence between training instances. It relies on a hierarchy of continuous deformation flows, which are parametrized by a neural network. Our model outperforms state-of-the-art methods in providing an expressive and robust shape prior for distal femur and liver. We show that the emerging latent representation is discriminative by separating healthy from pathological shapes. Ultimately, we demonstrate its effectiveness on two shape reconstruction tasks from partial data. Our source code is publicly available (https://github.com/davecasp/flowssm).