A Weakly supervised approach for thoracic diseases detection

作者：

Highlights：

• Saliency guided sampling is applied for online feature selection in Chest X-rays.

• A weakly supervised approach is developed to detect the salient regions of input.

• Ensemble prediction is implemented based on disease-specific saliency maps.

• Tailored downsampling of chest X-ray images enhances disease detection performance.

• Superior disease prediction performance offers improved healthcare decision making.

摘要

•Saliency guided sampling is applied for online feature selection in Chest X-rays.•A weakly supervised approach is developed to detect the salient regions of input.•Ensemble prediction is implemented based on disease-specific saliency maps.•Tailored downsampling of chest X-ray images enhances disease detection performance.•Superior disease prediction performance offers improved healthcare decision making.

论文关键词：Chest X-ray images,Convolutional neural networks,Medical image data,Multi-label image classification

论文评审过程：Received 30 January 2022, Revised 16 June 2022, Accepted 28 September 2022, Available online 4 October 2022, Version of Record 17 October 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118942