RetiGen: Framework leveraging domain generalization and test-time adaptation for multi-view retinal diagnostics

Gongyu Zhang, Ze Chen, Jiayu Huo, Joan M. Nunez Do Rio, Charalampos Komninos, Yang Liu, Rachel Sparks, Sébastien Ourselin, Christos Bergeles, Timothy Jackson

Abstract: Highlights•Introduces a novel test-time adaptation framework (RetiGen) for domain generalization in retinal imaging analysis.•Leverages unlabeled multi-view fundus photographs to enhance diagnostic accuracy and robustness.•Combines pseudo-label distribution calibration, self-distillation, and multi-view local clustering.•Provides a comprehensive comparison with state-of-the-art domain generalization and test-time adaptation methods.•Achieves state-of-the-art performance and improves model generalizability in cross-domain retinal diagnostics.