Uncertainty-Aware Leave-One-Out Collaborative Learning for Scribble-Supervised Medical Image Segmentation

DOI: https://doi.org/10.62517/jbdc.202601322

Author(s)

Zhaoyang Chen*, Tuchao Li

Affiliation(s)

College of Artificial Intelligence and Big Data, Guangzhou Vocational University of Science and Technology, Guangzhou, Guangdong, China *Corresponding Author

Abstract

Dense pixel-level annotation is expensive in medical image segmentation, while scribble annotation provides a practical weak-supervision alternative. This study addresses pseudo-label self-confirmation in tri-branch scribble-supervised medical image segmentation. We propose an uncertainty-aware leave-one-out (LOO) collaborative learning framework built on heterogeneous CNN, Swin-UNet, and Mamba-UNet branches. Each branch is anchored by partial cross entropy on scribble pixels and receives dense pseudo supervision generated only from the other two branches; entropy-based confidence further weights pseudo labels during training and branch fusion during inference. Five-fold experiments on the ACDC cardiac MRI dataset show that the method preserves the strong baseline overlap performance, with Dice of 0.8794 compared with 0.8803 and IoU of 0.7921 compared with 0.7934. Precision changes slightly from 0.8662 to 0.8666, although HD95 and ASD do not consistently improve. These results demonstrate that the proposed strategy reduces direct self-confirmation and offers a lightweight pseudo-label governance mechanism without redesigning the backbone. The main contribution is a simple collaborative supervision formulation that separates sparse reliable scribble anchors from dense noisy pseudo-label expansion, providing a reproducible basis for future uncertainty- and boundary-aware weakly supervised medical segmentation.

Keywords

Medical Image Segmentation; Weak Supervision; Pseudo Label; Collaborative Learning; Vision Mamba

References

[1] O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation," in Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015, LNCS 9351, pp. 234-241, 2015. [2] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, "nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation," Nature Methods, vol. 18, no. 2, pp. 203-211, 2021. [3] H. Cao, Y. Wang, J. Chen, D. Jiang, X. Zhang, Q. Tian, and M. Wang, "Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation," arXiv:2105.05537, 2021. [4] Y. Liu, Y. Tian, Y. Zhao, H. Yu, L. Xie, Y. Wang, Q. Ye, J. Jiao, and Y. Liu, "VMamba: Visual State Space Model," arXiv:2401.10166, 2024. [5] Z. Wang and C. Ma, "Weak-Mamba-UNet: Visual Mamba Makes CNN and ViT Work Better for Scribble-based Medical Image Segmentation," arXiv:2402.10887, 2024. [6] M. Han, X. Luo, and X. Xie, "DMSPS: Dynamically mixed soft pseudo-label supervision for scribble-supervised medical image segmentation," Medical Image Analysis, vol. 97, Art. no. 103274, 2024. [7] T. Wang, X. Zhang, Y. Chen, Y. Zhou, L. Zhao, T. Tan, and T. Tong, "ScribbleVS: Scribble-Supervised Medical Image Segmentation via Dynamic Competitive Pseudo Label Selection," arXiv:2411.10237, 2024. [8] O. Bernard, A. Lalande, C. Zotti et al., "Deep Learning Techniques for Automatic MRI Cardiac Multi-Structures Segmentation and Diagnosis: Is the Problem Solved?" IEEE Transactions on Medical Imaging, vol. 37, no. 11, pp. 2514-2525, 2018. [9] D. Lin, J. Dai, J. Jia, K. He, and J. Sun, "ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, 2016, pp. 3159-3167. [10] K. Zhang and X. Zhuang, "CycleMix: A Holistic Strategy for Medical Image Segmentation from Scribble Supervision," in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), New Orleans, LA, USA, 2022, pp. 11656-11665