A Structure-Aware Relation Network for Thoracic Diseases Detection and Segmentation

TL;DR

This paper introduces SAR-Net, a structure-aware relation network that leverages anatomical and disease relations for improved detection and segmentation of thoracic diseases in chest X-ray images, demonstrating significant performance gains.

Contribution

The paper proposes SAR-Net, a novel extension of Mask R-CNN that incorporates domain knowledge through relation modules for better disease detection and segmentation in chest X-rays.

Findings

SAR-Net outperforms baseline Mask R-CNN on chest X-ray datasets.

ChestX-Det dataset provides detailed annotations for thoracic diseases.

Experimental results show significant accuracy improvements.

Abstract

Instance level detection and segmentation of thoracic diseases or abnormalities are crucial for automatic diagnosis in chest X-ray images. Leveraging on constant structure and disease relations extracted from domain knowledge, we propose a structure-aware relation network (SAR-Net) extending Mask R-CNN. The SAR-Net consists of three relation modules: 1. the anatomical structure relation module encoding spatial relations between diseases and anatomical parts. 2. the contextual relation module aggregating clues based on query-key pair of disease RoI and lung fields. 3. the disease relation module propagating co-occurrence and causal relations into disease proposals. Towards making a practical system, we also provide ChestX-Det, a chest X-Ray dataset with instance-level annotations (boxes and masks). ChestX-Det is a subset of the public dataset NIH ChestX-ray14. It contains ~3500 images of 13 common disease categories labeled by three board-certified radiologists. We evaluate our SAR-Net on it and another dataset DR-Private. Experimental results show that it can enhance the strong baseline of Mask R-CNN with significant improvements. The ChestX-Det is released at https://github.com/Deepwise-AILab/ChestX-Det-Dataset.