FunOTTA: On-the-Fly Adaptation on Cross-Domain Fundus Image via Stable Test-time Training

TL;DR

FunOTTA is a novel test-time adaptation framework that enables fundus image diagnosis models to adapt on-the-fly to unseen domains with strong variations, improving robustness and accuracy in real-world applications.

Contribution

We introduce FunOTTA, a stable and effective on-the-fly test-time adaptation method for cross-domain fundus image diagnosis, addressing domain shifts and reducing prior knowledge bias.

Findings

Outperforms state-of-the-art TTA methods on cross-domain fundus benchmarks.

Effective across different backbone networks and diseases.

Demonstrates robustness under strong domain shifts.

Abstract

Fundus images are essential for the early screening and detection of eye diseases. While deep learning models using fundus images have significantly advanced the diagnosis of multiple eye diseases, variations in images from different imaging devices and locations (known as domain shifts) pose challenges for deploying pre-trained models in real-world applications. To address this, we propose a novel Fundus On-the-fly Test-Time Adaptation (FunOTTA) framework that effectively generalizes a fundus image diagnosis model to unseen environments, even under strong domain shifts. FunOTTA stands out for its stable adaptation process by performing dynamic disambiguation in the memory bank while minimizing harmful prior knowledge bias. We also introduce a new training objective during adaptation that enables the classifier to incrementally adapt to target patterns with reliable class conditional estimation and consistency regularization. We compare our method with several state-of-the-art test-time adaptation (TTA) pipelines. Experiments on cross-domain fundus image benchmarks across two diseases demonstrate the superiority of the overall framework and individual components under different backbone networks. Code is available at https://github.com/Casperqian/FunOTTA.