ROP lesion segmentation via sequence coding and block balancing

Xiping Jia, Jianying Qiu, Dong Nie, Tian Liu

Abstract: Highlights•We propose a network SeBSNet for segmentation of multiple types of ROP lesions.•Feature learning integrates domain knowledge via image category coding.•A block-weighted strategy boosts the lesion area importance during network training.•We labeled 350 ROP images with 4 lesion types; data to be released upon publication.