Over the past decade, deep learning has become the gold standard for automatic medical image segmentation. Every segmentation task has an underlying uncertainty due to image resolution, annotation protocol, etc. Therefore, a number of methods and metrics have been proposed to quantify the uncertainty of neural networks mostly based on Bayesian deep learning, ensemble learning methods or output probability calibration. The aim of our research is to assess how reliable the different uncertainty metrics found in the literature are. We propose a quantitative and statistical comparison of uncertainty measures based on the relevance of the uncertainty map to predict misclassification. Four uncertainty metrics were compared over a set of 144 models. The application studied is the segmentation of the lumen and vessel wall of carotid arteries based on multiple sequences of magnetic resonance (MR) images in multi-center data.
|Title of host publication||Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, and Graphs in Biomedical Image Analysis - 2nd International Workshop, UNSURE 2020, and 3rd International Workshop, GRAIL 2020, Held in Conjunction with MICCAI 2020, Proceedings|
|Editors||Carole H. Sudre, Hamid Fehri, Tal Arbel, Christian F. Baumgartner, Adrian Dalca, Ryutaro Tanno, Koen Van Leemput, William M. Wells, Aristeidis Sotiras, Bartlomiej Papiez, Enzo Ferrante, Sarah Parisot|
|Publisher||Springer Science+Business Media|
|Number of pages||10|
|Publication status||Published - 2020|
|Event||2nd International Workshop on Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, UNSURE 2020, and the 3rd International Workshop on Graphs in Biomedical Image Analysis, GRAIL 2020, held in conjunction with the 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020 - Lima, Peru|
Duration: 8 Oct 2020 → 8 Oct 2020
|Series||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||2nd International Workshop on Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, UNSURE 2020, and the 3rd International Workshop on Graphs in Biomedical Image Analysis, GRAIL 2020, held in conjunction with the 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020|
|Period||8/10/20 → 8/10/20|
Bibliographical noteFunding Information:
Acknowledgments. This work was funded by Netherlands Organisation for Scientific Research (NWO) VICI project VI.C.182.042. The PARISK study was funded within the framework of CTMM, the Center for Translational Molecular Medicine, project PARISK (grant 01C-202), and supported by the Dutch Heart Foundation.
© 2020, Springer Nature Switzerland AG.