Vector Quantisation for Robust Segmentation

TL;DR

This paper introduces a vector quantisation approach within a neural network architecture to enhance the robustness of medical image segmentation models against noise, domain shifts, and perturbations.

Contribution

It proposes integrating vector quantisation into segmentation models to improve robustness, a novel application in medical imaging segmentation tasks.

Findings

Improved segmentation accuracy across three tasks.

Enhanced robustness to noise and domain shifts.

Effective in both latent and output spaces.

Abstract

The reliability of segmentation models in the medical domain depends on the model's robustness to perturbations in the input space. Robustness is a particular challenge in medical imaging exhibiting various sources of image noise, corruptions, and domain shifts. Obtaining robustness is often attempted via simulating heterogeneous environments, either heuristically in the form of data augmentation or by learning to generate specific perturbations in an adversarial manner. We propose and justify that learning a discrete representation in a low dimensional embedding space improves robustness of a segmentation model. This is achieved with a dictionary learning method called vector quantisation. We use a set of experiments designed to analyse robustness in both the latent and output space under domain shift and noise perturbations in the input space. We adapt the popular UNet architecture, inserting a quantisation block in the bottleneck. We demonstrate improved segmentation accuracy and better robustness on three segmentation tasks. Code is available at \url{https://github.com/AinkaranSanthi/Vector-Quantisation-for-Robust-Segmentation}