Online Reflective Learning for Robust Medical Image Segmentation

被引:4
|
作者
Huang, Yuhao [1 ,2 ,3 ]
Yang, Xin [1 ,2 ,3 ]
Huang, Xiaoqiong [1 ,2 ,3 ]
Liang, Jiamin [1 ,2 ,3 ]
Zhou, Xinrui [1 ,2 ,3 ]
Chen, Cheng [4 ]
Dou, Haoran [5 ]
Hu, Xindi [6 ]
Cao, Yan [1 ,2 ,3 ]
Ni, Dong [1 ,2 ,3 ]
机构
[1] Shenzhen Univ, Sch Biomed Engn, Hlth Sci Ctr, Natl Reg Key Technol Engn Lab Med Ultrasound, Shenzhen, Peoples R China
[2] Shenzhen Univ, Med Ultrasound Image Comp MUSIC Lab, Shenzhen, Peoples R China
[3] Shenzhen Univ, Marshall Lab Biomed Engn, Shenzhen, Peoples R China
[4] Chinese Univ Hong Kong, Dept Comp Sci & Engn, Hong Kong, Peoples R China
[5] Univ Leeds, Ctr Computat Imaging & Simulat Technol Biomed CIS, Leeds, England
[6] Shenzhen RayShape Med Technol Co Ltd, Shenzhen, Peoples R China
基金
中国国家自然科学基金;
关键词
Segmentation; Robustness; Online learning; DOMAIN ADAPTATION;
D O I
10.1007/978-3-031-16452-1_62
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep segmentation models often face the failure risks when the testing image presents unseen distributions. Improving model robustness against these risks is crucial for the large-scale clinical application of deep models. In this study, inspired by human learning cycle, we propose a novel online reflective learning framework (RefSeg) to improve segmentation robustness. Based on the reflection-on-action conception, our RefSeg firstly drives the deep model to take action to obtain semantic segmentation. Then, RefSeg triggers the model to reflect itself. Because making deep models realize their segmentation failures during testing is challenging, RefSeg synthesizes a realistic proxy image from the semantic mask to help deep models build intuitive and effective reflections. This proxy translates and emphasizes the segmentation flaws. By maximizing the structural similarity between the raw input and the proxy, the reflection-on-action loop is closed with segmentation robustness improved. RefSeg runs in the testing phase and is general for segmentation models. Extensive validation on three medical image segmentation tasks with a public cardiac MR dataset and two in-house large ultrasound datasets show that our RefSeg remarkably improves model robustness and reports state-of-the-art performance over strong competitors.
引用
收藏
页码:652 / 662
页数:11
相关论文
共 50 条
  • [1] Style Curriculum Learning for Robust Medical Image Segmentation
    Liu, Zhendong
    Manh, Van
    Yang, Xin
    Huang, Xiaoqiong
    Lekadir, Karim
    Campello, Victor
    Ravikumar, Nishant
    Frangi, Alejandro F.
    Ni, Dong
    MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION - MICCAI 2021, PT I, 2021, 12901 : 451 - 460
  • [2] Learning Robust Shape Regularization for Generalizable Medical Image Segmentation
    Chen, Kecheng
    Qin, Tiexin
    Lee, Victor Ho-Fun
    Yan, Hong
    Li, Haoliang
    IEEE TRANSACTIONS ON MEDICAL IMAGING, 2024, 43 (07) : 2693 - 2706
  • [3] Learning With Context Feedback Loop for Robust Medical Image Segmentation
    Girum, Kibrom Berihu
    Crehange, Gilles
    Lalande, Alain
    IEEE TRANSACTIONS ON MEDICAL IMAGING, 2021, 40 (06) : 1542 - 1554
  • [4] Learning a Single Network for Robust Medical Image Segmentation With Noisy Labels
    Ye, Shuquan
    Xu, Yan
    Chen, Dongdong
    Han, Songfang
    Liao, Jing
    IEEE TRANSACTIONS ON MEDICAL IMAGING, 2024, 43 (09) : 3188 - 3199
  • [5] DISTRIBUTION-AWARE CONTRASTIVE LEARNING FOR ROBUST MEDICAL IMAGE SEGMENTATION
    Qin, Zheyun
    Xi, Xiaoming
    Yin, Yilong
    2024 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, ICASSP 2024, 2024, : 1991 - 1995
  • [6] GSAL: Geometric structure adversarial learning for robust medical image segmentation
    Wang, Kun
    Zhang, Xiaohong
    Lu, Yuting
    Zhang, Wei
    Huang, Sheng
    Yang, Dan
    PATTERN RECOGNITION, 2023, 140
  • [7] A robust medical image segmentation method
    Hao, J. T.
    Li, M. L.
    DYNAMICS OF CONTINUOUS DISCRETE AND IMPULSIVE SYSTEMS-SERIES B-APPLICATIONS & ALGORITHMS, 2006, 13E : 2285 - 2289
  • [8] Pixel level deep reinforcement learning for accurate and robust medical image segmentation
    Liu, Yunxin
    Yuan, Di
    Xu, Zhenghua
    Zhan, Yuefu
    Zhang, Hongwei
    Lu, Jun
    Lukasiewicz, Thomas
    SCIENTIFIC REPORTS, 2025, 15 (01):
  • [9] A Hybrid Robust-Learning Architecture for Medical Image Segmentation with Noisy Labels
    Shi, Jialin
    Guo, Chenyi
    Wu, Ji
    FUTURE INTERNET, 2022, 14 (02)
  • [10] Learning robust medical image segmentation from multi-source annotations
    Wang, Yifeng
    Luo, Luyang
    Wu, Mingxiang
    Wang, Qiong
    Chen, Hao
    MEDICAL IMAGE ANALYSIS, 2025, 101