SUPER-Net: Trustworthy Image Segmentation via Uncertainty Propagation in Encoder-Decoder Networks
This addresses the brittleness of DL models in sensitive fields like medical imaging by enabling trustworthy segmentation with self-assessment capabilities, though it is an incremental improvement over existing uncertainty methods.
The paper tackles the problem of deep learning models lacking uncertainty quantification in image segmentation, proposing SUPER-Net, a Bayesian framework that propagates uncertainty to generate segmentation and uncertainty maps without Monte Carlo sampling, showing it outperforms state-of-the-art models in robustness and accuracy on MRI and CT scans under noisy and adversarial conditions.
Deep Learning (DL) holds great promise in reshaping the industry owing to its precision, efficiency, and objectivity. However, the brittleness of DL models to noisy and out-of-distribution inputs is ailing their deployment in sensitive fields. Current models often lack uncertainty quantification, providing only point estimates. We propose SUPER-Net, a Bayesian framework for trustworthy image segmentation via uncertainty propagation. Using Taylor series approximations, SUPER-Net propagates the mean and covariance of the model's posterior distribution across nonlinear layers. It generates two outputs simultaneously: the segmented image and a pixel-wise uncertainty map, eliminating the need for expensive Monte Carlo sampling. SUPER-Net's performance is extensively evaluated on MRI and CT scans under various noisy and adversarial conditions. Results show that SUPER-Net outperforms state-of-the-art models in robustness and accuracy. The uncertainty map identifies low-confidence areas affected by noise or attacks, allowing the model to self-assess segmentation reliability, particularly when errors arise from noise or adversarial examples.