Shape-intensity knowledge distillation for robust medical image segmentation

TL;DR

This paper introduces a shape-intensity knowledge distillation method that enhances medical image segmentation accuracy and generalization by transferring shape-intensity priors from a teacher to a student network.

Contribution

It presents a novel approach to incorporate shape-intensity priors into segmentation networks via knowledge distillation, improving robustness across datasets.

Findings

Consistent improvement over baseline models on five segmentation tasks.

Enhanced cross-dataset generalization ability.

Effective integration of shape-intensity priors into segmentation networks.

Abstract

Many medical image segmentation methods have achieved impressive results. Yet, most existing methods do not take into account the shape-intensity prior information. This may lead to implausible segmentation results, in particular for images of unseen datasets. In this paper, we propose a novel approach to incorporate joint shape-intensity prior information into the segmentation network. Specifically, we first train a segmentation network (regarded as the teacher network) on class-wise averaged training images to extract valuable shape-intensity information, which is then transferred to a student segmentation network with the same network architecture as the teacher via knowledge distillation. In this way, the student network regarded as the final segmentation model can effectively integrate the shape-intensity prior information, yielding more accurate segmentation results. Despite its simplicity, experiments on five medical image segmentation tasks of different modalities demonstrate that the proposed Shape-Intensity Knowledge Distillation (SIKD) consistently improves several baseline models (including recent MaxStyle and SAMed) under intra-dataset evaluation, and significantly improves the cross-dataset generalization ability. The code is available at https://github.com/whdong-whu/SIKD.