Towards Generalizable Medical Image Segmentation with Pixel-wise Uncertainty Estimation

TL;DR

This paper introduces a novel uncertainty estimation framework for medical image segmentation that highlights hard-to-classify pixels, improving model generalization across different datasets.

Contribution

It proposes a new pixel-wise uncertainty estimation method to identify challenging pixels, enhancing segmentation accuracy and robustness in medical imaging.

Findings

Outperforms state-of-the-art methods on prostate and fundus datasets.

Highlights hard-to-classify pixels to improve segmentation robustness.

Demonstrates better generalization in real-world medical applications.

Abstract

Deep neural networks (DNNs) achieve promising performance in visual recognition under the independent and identically distributed (IID) hypothesis. In contrast, the IID hypothesis is not universally guaranteed in numerous real-world applications, especially in medical image analysis. Medical image segmentation is typically formulated as a pixel-wise classification task in which each pixel is classified into a category. However, this formulation ignores the hard-to-classified pixels, e.g., some pixels near the boundary area, as they usually confuse DNNs. In this paper, we first explore that hard-to-classified pixels are associated with high uncertainty. Based on this, we propose a novel framework that utilizes uncertainty estimation to highlight hard-to-classified pixels for DNNs, thereby improving its generalization. We evaluate our method on two popular benchmarks: prostate and fundus datasets. The results of the experiment demonstrate that our method outperforms state-of-the-art methods.