Key result
Referring on average 1% of the most uncertain voxels in an image to a human expert increased segmentation performance by 8%, 7%, and 5% for the right ventricle, myocardium, and left ventricle.
Why the study?
Deep learning cardiac MRI segmentation methods lack clinical adoption due to unreliability from models that fail unnoticed and produce anatomically implausible results.
Does an automatic image segmentation method based on Bayesian dilated convolutional networks improve segmentation accuracy of cardiac MRI in a human-in-the-loop setting?
Does an automatic image segmentation method based on Bayesian dilated convolutional networks improve segmentation accuracy of cardiac MRI in a human-in-the-loop setting?
Generating spatial uncertainty maps alongside deep learning segmentation of cardiac MRI allows for targeted human-in-the-loop correction, significantly improving segmentation accuracy with minimal manual effort.
May enable efficient uncertainty-guided expert review in cardiac MRI segmentation; leaves open prospective clinical validation and workflow integration.
Current state-of-the-art deep learning segmentation methods have not yet made a broad entrance into the clinical setting in spite of high demand for such automatic methods. One important reason is the lack of reliability caused by models that fail unnoticed and often locally produce anatomically implausible results that medical experts would not make. This paper presents an automatic image segmentation method based on (Bayesian) dilated convolutional networks (DCNN) that generate segmentation masks and spatial uncertainty maps for the input image at hand. The method was trained and evaluated using segmentation of the left ventricle (LV) cavity, right ventricle (RV) endocardium and myocardium (Myo) at end-diastole (ED) and end-systole (ES) in 100 cardiac 2D MR scans from the MICCAI 2017 Challenge (ACDC). Combining segmentations and uncertainty maps and employing a human-in-the-loop setting, we provide evidence that image areas indicated as highly uncertain, regarding the obtained segmentation, almost entirely cover regions of incorrect segmentations. The fused information can be harnessed to increase segmentation performance. Our results reveal that we can obtain valuable spatial uncertainty maps with low computational effort using DCNNs.
No takes yet. Share an insight, caveat, or question.
Sander et al. (2019) studied Cardiac abnormalities (normal, dilated cardiomyopathy, hypertrophic cardiomyopathy, heart failure with infarction, right ventricular abnormality) (n=100). Bayesian dilated convolutional networks (DCNN) with uncertainty maps vs. Standard DCNN without uncertainty referral was evaluated on Segmentation performance (Dice score) improvement upon referring uncertain voxels. Referring on average 1% of the most uncertain voxels in an image to a human expert increased segmentation performance by 8%, 7%, and 5% for the right ventricle, myocardium, and left ventricle.
Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context: