DenseMP: Unsupervised Dense Pre-training for Few-shot Medical Image Segmentation

TL;DR

DenseMP introduces an unsupervised dense pre-training pipeline for few-shot medical image segmentation, significantly improving performance and achieving state-of-the-art results on Abd-CT and Abd-MRI datasets.

Contribution

It proposes a novel two-stage unsupervised dense pre-training method tailored for few-shot medical image segmentation, addressing data scarcity and over-fitting issues.

Findings

Achieves state-of-the-art results on Abd-CT and Abd-MRI datasets.

Enhances the performance of the PA-Net segmentation model.

Effectively mitigates over-fitting in few-shot medical image segmentation.

Abstract

Few-shot medical image semantic segmentation is of paramount importance in the domain of medical image analysis. However, existing methodologies grapple with the challenge of data scarcity during the training phase, leading to over-fitting. To mitigate this issue, we introduce a novel Unsupervised Dense Few-shot Medical Image Segmentation Model Training Pipeline (DenseMP) that capitalizes on unsupervised dense pre-training. DenseMP is composed of two distinct stages: (1) segmentation-aware dense contrastive pre-training, and (2) few-shot-aware superpixel guided dense pre-training. These stages collaboratively yield a pre-trained initial model specifically designed for few-shot medical image segmentation, which can subsequently be fine-tuned on the target dataset. Our proposed pipeline significantly enhances the performance of the widely recognized few-shot segmentation model, PA-Net, achieving state-of-the-art results on the Abd-CT and Abd-MRI datasets. Code will be released after acceptance.