Learning to Segment using Summary Statistics and Weak Supervision
For medical experts who discard segmentation annotations, this method reduces annotation burden by leveraging retained summary statistics and minimal weak supervision.
The paper proposes a method to train segmentation models using only summary statistics (e.g., area) and a few weak pixel annotations, achieving performance improvements over using statistics alone. Experiments on medical imaging datasets show the approach is effective.
Medical experts often manually segment images to obtain diagnostic statistics and discard the resulting annotations. We aim to train segmentation models to alleviate this burden, but constrained to the retained summary statistics (e.g., the area of the annotated region). Empirical results suggest that statistics alone are insufficient for this task, but adding weak information in the form of a few pixels within the area of interest significantly improves performance. We use a novel loss function that combines terms for image reconstruction quality, matching to summary statistics, and overlap between the predicted foreground and the weak supervisory signal. Experiments on standard image, ultrasound (breast cancer), and Computed Tomography (CT) scan (kidney tumors) data demonstrate the utility and potential of the approach.