ZScribbleSeg: A comprehensive segmentation framework with modeling of efficient annotation and maximization of scribble supervision

2026-05-07

AI Summary

This work addresses the high cost of full annotation in medical image segmentation and the limited accuracy of existing scribble-based weakly supervised methods, which often suffer from insufficient supervision and lack of shape priors. The authors propose a novel weakly supervised segmentation framework that systematically investigates principles for generating high-quality scribble annotations. By employing a supervision maximization strategy combined with stochastic simulation, the method produces efficient scribble configurations. It further enhances inference and error correction in unannotated regions through spatial relationship modeling, a shape-constrained regularization term, and an EM algorithm to estimate class mixture proportions. Using only scribble-level annotations, the approach achieves competitive performance across six diverse benchmarks: ACDC, MSCMRseg, BTCV, MyoPS, and the brain tumor and prostate tasks from the Medical Decathlon challenge.

Abstract

Curating fully annotated datasets for medical image segmentation is labour-intensive and expertise-demanding. To alleviate this problem, prior studies have explored scribble annotations for weakly supervised segmentation. Existing solutions mainly compute losses on annotated areas and generate pseudo labels by propagating annotations to adjacent regions. However, these methods often suffer from inaccurate and unrealistic segmentations due to insufficient supervision and incomplete shape information. In contrast, we first investigate the principle of good scribble annotations, which leads to efficient scribble forms via supervision maximization and randomness simulation. We further introduce regularization terms to encode the spatial relationship and the shape constraints, where the EM algorithm is utilized to estimate the mixture ratios of label classes. These ratios are critical in identifying the unlabeled pixels for each class and correcting erroneous predictions, thus the accurate estimation lays the foundation for the incorporation of spatial prior. Finally, we integrate the efficient scribble supervision with the prior into a framework, referred to as ZScribbleSeg, and apply it to multiple scenarios. Leveraging only scribble annotations, ZScribbleSeg achieves competitive performance on six segmentation tasks including ACDC, MSCMRseg, BTCV, MyoPS, Decathlon-BrainTumor and Decathlon-Prostate. Our code will be released via https://github.com/DLwbm123/ZScribbleSeg.

Problem

• medical image segmentation
• scribble annotation
• weakly supervised learning
• annotation efficiency
• shape constraints

Innovation

• scribble supervision
• weakly supervised segmentation
• EM algorithm
• shape constraints
• spatial prior