Fakultät Informatik und Mathematik
Refine
Year of publication
Document Type
- conference proceeding (presentation, abstract) (33) (remove)
Is part of the Bibliography
- no (33)
Keywords
- Künstliche Intelligenz (6)
- Artificial Intelligence (4)
- Deep Learning (4)
- Diagnose (3)
- Speiseröhrenkrankheit (3)
- Barrett-Ösophagus (2)
- Bildgebendes Verfahren (2)
- Endoscopy (2)
- Speiseröhrenkrebs (2)
- Adenokarzinom (1)
Institute
Begutachtungsstatus
- peer-reviewed (33) (remove)
Second cancer risk after radiation of localized prostate cancer with and without flattening filter
(2017)
Purpose
The complexity of the treatment techniques IMRT (Intensity Modulated Radiation Therapy) and VMAT (Volumetric Modulated Arc Therapy) introduced the requirement of an individual plan verification. The latest development is 3D verification which allows evaluation in terms of 3D gamma values but also a judgment based on differences in dose distributions on individual patient CT data and in dose volume histograms. For a verification tool the error detection sensitivity is important. The purpose of this work is to evaluate the accuracy and the sensitivity of the gamma evaluation of the 3D verification software MobiusFx when a systematic error in one parameter is applied. The results are compared to an established 2D-measurement based method.
Methods
11 IMRT and 11 VMAT plans were selected as reference plans. For every reference plan a systematic offset of 1 mm, 2 mm and 3 mm in the same direction was applied to the Multi-Leaf-Collimator (MLC) positions. All plans were irradiated and verified simultaneously with MobiusFx and MatriXX Evolution 2D array measurement. For both verification systems the applied dose distributions were evaluated using the gamma method based on the reference plans. In MobiusFx additionally the MLC position errors were evaluated.
Results
Regarding the gamma evaluation results, MobiusFx and MatriXX Evolution 2D measurements have almost the same sensitivity: The gamma evaluation of both systems detected shifts of 1 mm in VMAT plans and shifts of 2 mm and larger in IMRT plans. The information of the MLC position error provided by MobiusFx allows further to detect errors down to a shift of 1 mm in IMRT plans, giving an advantage over the MatriXX measurement.
Conclusions
MobiusFx shows a high error detection sensitivity, comparable to MatriXX Evolution. Through the gamma calculation and the MLC position error it is possible to detect errors down to a shift in MLC positions of 1 mm. It is recommended to inspect all the above parameters during plan verification.
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.