Self-supervised multimodal reconstruction pre-training for retinal computer-aided diagnosis
作者：
Highlights：
• Self-supervised multimodal pre-training improves retinal computer-aided diagnosis.
• Multimodal reconstruction provides domain-specific knowledge using unlabeled images.
• Age-related macular degeneration and glaucoma diagnosis using deep learning.
• Deep learning approach for computer-aided diagnosis with limited annotated data.
• The proposal outperforms other self-supervised and fully-supervised alternatives.
摘要
•Self-supervised multimodal pre-training improves retinal computer-aided diagnosis.•Multimodal reconstruction provides domain-specific knowledge using unlabeled images.•Age-related macular degeneration and glaucoma diagnosis using deep learning.•Deep learning approach for computer-aided diagnosis with limited annotated data.•The proposal outperforms other self-supervised and fully-supervised alternatives.
论文关键词：Deep learning,Medical imaging,Self-supervised learning,Eye fundus,Transfer learning,Computer-aided diagnosis
论文评审过程：Received 4 December 2020, Revised 15 May 2021, Accepted 10 July 2021, Available online 24 July 2021, Version of Record 31 July 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115598
