A Multi-Task Cross-Task Learning Architecture for Ad-hoc Uncertainty Estimation in 3D Cardiac MRI Image Segmentation
This work addresses the need for reliable segmentation in medical imaging, specifically for cardiac MRI, by enhancing model generalizability and failure detection, though it is incremental as it builds on existing multi-task and semi-supervised learning trends.
The paper tackles the problem of generating accurate and smooth 3D cardiac MRI segmentation masks by proposing a multi-task cross-task learning approach that enforces consistency between segmentation and distance map tasks, resulting in improved performance with varied labeled data quantities and the ability to flag low-quality segmentations using uncertainty estimates.
Medical image segmentation has significantly benefitted thanks to deep learning architectures. Furthermore, semi-supervised learning (SSL) has recently been a growing trend for improving a model's overall performance by leveraging abundant unlabeled data. Moreover, learning multiple tasks within the same model further improves model generalizability. To generate smoother and accurate segmentation masks from 3D cardiac MR images, we present a Multi-task Cross-task learning consistency approach to enforce the correlation between the pixel-level (segmentation) and the geometric-level (distance map) tasks. Our extensive experimentation with varied quantities of labeled data in the training sets justifies the effectiveness of our model for the segmentation of the left atrial cavity from Gadolinium-enhanced magnetic resonance (GE-MR) images. With the incorporation of uncertainty estimates to detect failures in the segmentation masks generated by CNNs, our study further showcases the potential of our model to flag low-quality segmentation from a given model.