CheXGAT: A disease correlation-aware network for thorax disease diagnosis from chest X-ray images

作者：

Highlights：

• We propose a novel end-to-end trainable multilabel classification framework that employs a CNN model and a GNN module to diagnose thorax diseases in CXR images.

• We combined the image representation features with the semantic features to explore the implicit correlations between thorax diseases.

• We evaluated our method on the benchmark multilabel radiography dataset, and the method achieved impressive performance.

摘要

•We propose a novel end-to-end trainable multilabel classification framework that employs a CNN model and a GNN module to diagnose thorax diseases in CXR images.•We combined the image representation features with the semantic features to explore the implicit correlations between thorax diseases.•We evaluated our method on the benchmark multilabel radiography dataset, and the method achieved impressive performance.

论文关键词：Chest X-ray,Thorax diseases,Disease correlation,Computer-aided diagnosis,Graph neural network,Graph attention mechanism,Convolutional neural network

论文评审过程：Received 28 June 2021, Revised 7 August 2022, Accepted 19 August 2022, Available online 27 August 2022, Version of Record 30 August 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2022.102382