Understanding and Visualizing Generalization in UNets

Abhejit Rajagopal, Vamshi Chowdary Madala, Thomas A Hope, Peder Larson
Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, PMLR 143:665-681, 2021.

Abstract

Fully-convolutional neural networks, such as the 2D or 3D UNet, are now pervasive in medical imaging for semantic segmentation, classification, image denoising, domain translation, and reconstruction. However, evaluation of UNet performance, as with most CNNs, has mostly been relegated to evaluation of a few performance metrics (e.g. accuracy, IoU, SSIM, etc.) using the network’s final predictions, which provides little insight into important issues such as dataset shift that occur in clinical application. In this paper, we propose techniques for understanding and visualizing the generalization performance of UNets in image classification and regression tasks, giving rise to metrics that are indicative of performance on a withheld test-set without the need for groundtruth annotations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v143-rajagopal21a, title = {Understanding and Visualizing Generalization in {UN}ets}, author = {Rajagopal, Abhejit and Madala, Vamshi Chowdary and Hope, Thomas A and Larson, Peder}, booktitle = {Proceedings of the Fourth Conference on Medical Imaging with Deep Learning}, pages = {665--681}, year = {2021}, editor = {Heinrich, Mattias and Dou, Qi and de Bruijne, Marleen and Lellmann, Jan and Schläfer, Alexander and Ernst, Floris}, volume = {143}, series = {Proceedings of Machine Learning Research}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v143/rajagopal21a/rajagopal21a.pdf}, url = {https://proceedings.mlr.press/v143/rajagopal21a.html}, abstract = {Fully-convolutional neural networks, such as the 2D or 3D UNet, are now pervasive in medical imaging for semantic segmentation, classification, image denoising, domain translation, and reconstruction. However, evaluation of UNet performance, as with most CNNs, has mostly been relegated to evaluation of a few performance metrics (e.g. accuracy, IoU, SSIM, etc.) using the network’s final predictions, which provides little insight into important issues such as dataset shift that occur in clinical application. In this paper, we propose techniques for understanding and visualizing the generalization performance of UNets in image classification and regression tasks, giving rise to metrics that are indicative of performance on a withheld test-set without the need for groundtruth annotations.} }
Endnote
%0 Conference Paper %T Understanding and Visualizing Generalization in UNets %A Abhejit Rajagopal %A Vamshi Chowdary Madala %A Thomas A Hope %A Peder Larson %B Proceedings of the Fourth Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2021 %E Mattias Heinrich %E Qi Dou %E Marleen de Bruijne %E Jan Lellmann %E Alexander Schläfer %E Floris Ernst %F pmlr-v143-rajagopal21a %I PMLR %P 665--681 %U https://proceedings.mlr.press/v143/rajagopal21a.html %V 143 %X Fully-convolutional neural networks, such as the 2D or 3D UNet, are now pervasive in medical imaging for semantic segmentation, classification, image denoising, domain translation, and reconstruction. However, evaluation of UNet performance, as with most CNNs, has mostly been relegated to evaluation of a few performance metrics (e.g. accuracy, IoU, SSIM, etc.) using the network’s final predictions, which provides little insight into important issues such as dataset shift that occur in clinical application. In this paper, we propose techniques for understanding and visualizing the generalization performance of UNets in image classification and regression tasks, giving rise to metrics that are indicative of performance on a withheld test-set without the need for groundtruth annotations.
APA
Rajagopal, A., Madala, V.C., Hope, T.A. & Larson, P.. (2021). Understanding and Visualizing Generalization in UNets. Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 143:665-681 Available from https://proceedings.mlr.press/v143/rajagopal21a.html.

Related Material