Towards Unbiased COVID-19 Lesion Localisation and Segmentation via Weakly Supervised Learning

• DOI: 10.48550/arxiv.2103.00780
• Publication Date: 2021-01-01
• Authors: Yang Yang; Jiancong Chen; Ruixuan Wang; Ting Ma; Lingling Wang; Jie Chen; Wei‐Shi Zheng; Tong Zhang
• Keywords: FOS: Computer and information sciences; 03 medical and health sciences; 0302 clinical medicine; Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV); Computer Science - Computer Vision and Pattern Recognition; FOS: Electrical engineering, electronic engineering, information engineering; Electrical Engineering and Systems Science - Image and Video Processing; 3. Good health

Despite tremendous efforts, it is very challenging to generate a robust model assist in the accurate quantification assessment of COVID-19 on chest CT images. Due nature blurred boundaries, supervised segmentation methods usually suffer from annotation biases. To support unbiased lesion localisation and minimise labeling costs, we propose data-driven framework by only image-level labels. The can explicitly separate potential lesions original images, with help generative adversarial network lesion-specific decoder. Experiments two datasets demonstrate effectiveness proposed its superior performance several existing methods.