SS-3DCapsNet: Self-supervised 3D Capsule Networks for Medical Segmentation on Less Labeled Data

TL;DR

This paper introduces SS-3DCapsNet, a self-supervised 3D capsule network architecture that enhances medical image segmentation performance on limited labeled data by pre-training with self-reconstruction tasks.

Contribution

It extends capsule networks for volumetric segmentation using self-supervised pre-training, improving initialization and performance over previous methods.

Findings

Outperforms previous capsule networks and 3D-UNets on multiple datasets

Effective self-supervised pre-training improves segmentation accuracy

Demonstrates robustness with less labeled data

Abstract

Capsule network is a recent new deep network architecture that has been applied successfully for medical image segmentation tasks. This work extends capsule networks for volumetric medical image segmentation with self-supervised learning. To improve on the problem of weight initialization compared to previous capsule networks, we leverage self-supervised learning for capsule networks pre-training, where our pretext-task is optimized by self-reconstruction. Our capsule network, SS-3DCapsNet, has a UNet-based architecture with a 3D Capsule encoder and 3D CNNs decoder. Our experiments on multiple datasets including iSeg-2017, Hippocampus, and Cardiac demonstrate that our 3D capsule network with self-supervised pre-training considerably outperforms previous capsule networks and 3D-UNets.