Uncertainty-Guided Lung Nodule Segmentation with Feature-Aware Attention

TL;DR

This paper introduces UGS-Net, a segmentation network that leverages uncertainty information from multiple annotations to improve lung nodule segmentation accuracy, especially in uncertain regions, using novel modules for uncertainty estimation and feature attention.

Contribution

It proposes a new uncertainty-guided segmentation framework with modules for uncertainty estimation and feature-aware attention, enhancing segmentation performance on ambiguous regions.

Findings

Achieves superior segmentation accuracy on LIDC-IDRI dataset.

Effectively predicts regions with varying uncertainty levels.

Outperforms existing methods in handling annotation disagreements.

Abstract

Since radiologists have different training and clinical experiences, they may provide various segmentation annotations for a lung nodule. Conventional studies choose a single annotation as the learning target by default, but they waste valuable information of consensus or disagreements ingrained in the multiple annotations. This paper proposes an Uncertainty-Guided Segmentation Network (UGS-Net), which learns the rich visual features from the regions that may cause segmentation uncertainty and contributes to a better segmentation result. With an Uncertainty-Aware Module, this network can provide a Multi-Confidence Mask (MCM), pointing out regions with different segmentation uncertainty levels. Moreover, this paper introduces a Feature-Aware Attention Module to enhance the learning of the nodule boundary and density differences. Experimental results show that our method can predict the nodule regions with different uncertainty levels and achieve superior performance in LIDC-IDRI dataset.