Scribble-supervised multi-organ segmentation via epistemic-driven hardness-adaptive focusing

Tools

Han, Xiaoxiang and Liu, Yiman and Shang, Jiang and Chen, Haobo and Liu, Xiaohong and Qiu, Zhen and Wang, Yan and Zhang, Qi (2026) Scribble-supervised multi-organ segmentation via epistemic-driven hardness-adaptive focusing. IEEE Transactions on Medical Imaging. ISSN 1558-254X (https://doi.org/10.1109/TMI.2026.3665556)

[thumbnail of Han-etal-Scribble-supervised-multi-organ-segmentation-via-epistemic-driven-hardness-adaptive-focusing]
Preview
 Text. Filename: Han-etal-Scribble-supervised-multi-organ-segmentation-via-epistemic-driven-hardness-adaptive-focusing.pdf
Accepted Author Manuscript
License: Creative Commons Attribution 4.0 logo
Download (5MB)| Preview

Abstract

Scribble supervision reduces annotation costs in multi-organ segmentation. However, its sparsity results in insufficient supervision for most regions and inadequate feature learning in hard areas (e.g., organ boundaries). These hard areas cause model confirmation bias and high epistemic uncertainty, which existing methods fail to address. To overcome these core challenges, we propose an epistemic-driven hardness-adaptive focusing framework. This framework establishes a self-improving loop: quantified epistemic uncertainty guides hard sample generation, while hard sample learning and feature alignment jointly reduce epistemic uncertainty. Specifically, we first propose a phase-adaptive hardness-aware loss function to quantify epistemic uncertainty and generate dynamic hardness maps during training. Based on these maps, we employ a distribution-divergence-aware copy-paste operation to create hard samples, which are progressively incorporated into learning to reduce epistemic uncertainty. Furthermore, we introduce feature distribution alignment to mitigate bias and epistemic uncertainty by aligning organ-specific hard regions with global features. Extensive experiments on multi-organ CT and ultrasound datasets demonstrate the competitiveness and effectiveness of our method. The framework’s generalizability and robustness are further validated under cross-dataset and noise-corrupted scenarios. This work offers a practical solution for clinical applications where annotation efficiency is critical.

ORCID iDs

Han, Xiaoxiang, Liu, Yiman, Shang, Jiang, Chen, Haobo, Liu, Xiaohong, Qiu, Zhen ORCID logoORCID: https://orcid.org/0000-0002-0226-7855, Wang, Yan and Zhang, Qi;
CORE (COnnecting REpositories)