DMDF-Net: Dual multiscale dilated fusion network for accurate segmentation of lesions related to COVID-19 in lung radiographic scans

作者：

Highlights：

• Dual multiscale dilated fusion network (DMDF-Net) is proposed for segmentation.

• Dual multiscale feature fusion is included in the encoder and decoder of DMDF-Net.

• Concept of post-region of interest (ROI) fusion is used for lesion quantification.

• Achieve superior results as an average DICE score of 75.7% and IoU score of 67.22%.

• Our trained DMDF-Net is publicly available.

摘要

•Dual multiscale dilated fusion network (DMDF-Net) is proposed for segmentation.•Dual multiscale feature fusion is included in the encoder and decoder of DMDF-Net.•Concept of post-region of interest (ROI) fusion is used for lesion quantification.•Achieve superior results as an average DICE score of 75.7% and IoU score of 67.22%.•Our trained DMDF-Net is publicly available.

论文关键词：DMDF-Net,Computer-aided diagnosis,Lung segmentation,COVID-19 lesions segmentation,Infection quantification

论文评审过程：Received 13 July 2021, Revised 24 January 2022, Accepted 25 April 2022, Available online 2 May 2022, Version of Record 6 May 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117360