大数据与互联网学院关于“Deep Learning Based Semi-Supervised Medical Image Segmentation”讲座的通知

一、时间：2024年11月5日（周二）20:30

二、地点：#腾讯会议:260-991-122

三、报告主题

Deep Learning Based Semi-Supervised Medical Image Segmentation

四、报告摘要

  Semi-supervised medical image segmentation has shown promise in training models with limited labeled data and abundant unlabeled data. However, the performance of state-of-the-art semi-supervised models still lags significantly behind fully supervised models. These challenges highlight the need for methods that can accurately capture boundaries of interest areas by leveraging inherent image features rather than relying solely on extensive annotations.

  In this talk, I will go through our recent works, which narrow the gap between semi-supervised and fully supervised approaches significantly while using less than one-fifth labeled data. We first focus on representation learning for semi-supervised learning by developing a novel contrastive learning framework to extract robust feature representations that reflect intra- and inter-slice relationships across the whole dataset [1, 2]. Secondly, existing works ignore a potentially valuable source of unsupervised semantic information—spatial registration transforms between image volumes. We propose a contrastive cross-teaching framework incorporating registration information to further boost semi-supervised segmentation performance [3]. Extensive experiments on multi-structure medical segmentation datasets (CT and MRI) demonstrate the effectiveness of our methods.

[1] Liu, Q., Gu, X., Henderson, P., & Deligianni, F. (2023). Multi-Scale Cross Contrastive Learning for Semi-Supervised Medical Image Segmentation, BMVC.

[2] Liu, Q., Gu, X., Henderson, P., Hang, D. & Deligianni, F. (2024). Certainty-Guided Cross Contrastive Learning for Semi-Supervised Medical Image Segmentation. (submitted to JBHI)

[3] Liu, Q., Henderson, P., Gu, X., Hang, D. & Deligianni, F. (2025). Learning Semi-Supervised Medical Image Segmentation from Spatial Registration, WACV.

五、主讲人简介

Qianying Liu is a final-year Ph.D. student fully funded by CSC at the School of Computing Science, University of Glasgow, UK. She is supervised by Dr. Fani Deligianni, Dr. Paul Henderson, and Dr. Hang Dai. Her research focuses on understanding the visual information of medical images with fully and limited supervision, particularly in identifying organs and learning their boundaries, e.g., medical image segmentation/classification. This work lies at the intersection of Machine Learning, Computer Vision, and Medical Image Processing.

六、参会人员

欢迎大数据与互联网学院及其他学院感兴趣的师生参加。

大数据与互联网学院

2024年11月4日