Joint few-shot registration and segmentation self-training of 3D medical images

• Published in: Biomedical signal processing and control Vol. 80; p. 104294
• Main Authors: Shi, Huabang, Lu, Liyun, Yin, Mengxiao, Zhong, Cheng, Yang, Feng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2023
• Subjects: Few-shot; Image registration; Image segmentation; Pseudo-labels; Self-training; Image registration; Image segmentation; Self-training; Few-shot; Pseudo-labels
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2022.104294

•Joint registration and segmentation self-training framework for few-shot scenario.•Weakly supervised registration and semi-supervised segmentation complement each other.•Quality-assessed and screened pseudo-labels ensure a benign dual-task self-training.•Outperforms fully supervised single-task models on 3D medical images. Medical image segmentation and registration are very important related steps in clinical medical diagnosis. In the past few years, deep learning techniques for joint segmentation and registration have achieved good results in both segmentation and registration tasks through one-way assisted learning or mutual utilization. However, they often rely on large labeled datasets for supervised training or directly use pseudo-labels without quality estimation. We propose a joint registration and segmentation self-training framework (JRSS), which aims to use segmentation pseudo-labels to promote shared learning between segmentation and registration in scenarios with few manually labeled samples while improving the performance of dual tasks. JRSS combines weakly supervised registration and semi-supervised segmentation learning in a self-training framework. Segmentation self-training generates high-quality pseudo-labels for unlabeled data by injecting noise, pseudo-labels screening, and uncertainty correction. Registration utilizes pseudo-labels to facilitate weakly supervised learning, and as input noise as well as data augmentation to facilitate segmentation self-training. Experiments on two public 3D medical image datasets, abdominal CT and brain MRI, demonstrate that our proposed method achieves simultaneous improvements in segmentation and registration accuracy under few-shot scenarios. Outperforms the single-task fully-supervised training state-of-the-art model in the metrics of Dice similarity coefficient and standard deviation of the Jacobian determinant.