Few-shot image segmentation for cross-institution male pelvic organs using registration-assisted prototypical learning

TL;DR

This paper introduces a novel 3D few-shot segmentation method for cross-institution male pelvic organs, leveraging registration-assisted prototypical learning to improve accuracy with limited data from diverse sources.

Contribution

It presents the first 3D few-shot interclass segmentation network using registration to enhance generalization across institutions in medical imaging.

Findings

Significantly improved segmentation accuracy on unseen institution data

Achieved 75% fewer parameters compared to 2D approaches

Effective utilization of prior anatomical knowledge through registration

Abstract

The ability to adapt medical image segmentation networks for a novel class such as an unseen anatomical or pathological structure, when only a few labelled examples of this class are available from local healthcare providers, is sought-after. This potentially addresses two widely recognised limitations in deploying modern deep learning models to clinical practice, expertise-and-labour-intensive labelling and cross-institution generalisation. This work presents the first 3D few-shot interclass segmentation network for medical images, using a labelled multi-institution dataset from prostate cancer patients with eight regions of interest. We propose an image alignment module registering the predicted segmentation of both query and support data, in a standard prototypical learning algorithm, to a reference atlas space. The built-in registration mechanism can effectively utilise the prior knowledge of consistent anatomy between subjects, regardless whether they are from the same institution or not. Experimental results demonstrated that the proposed registration-assisted prototypical learning significantly improved segmentation accuracy (p-values<0.01) on query data from a holdout institution, with varying availability of support data from multiple institutions. We also report the additional benefits of the proposed 3D networks with 75% fewer parameters and an arguably simpler implementation, compared with existing 2D few-shot approaches that segment 2D slices of volumetric medical images.