Search papers, labs, and topics across Lattice.
The paper introduces a task-agnostic continual learning (CL) approach, CARL-XRay, for chest radiograph classification, addressing the challenge of sequentially incorporating heterogeneous datasets without task identifiers at inference. CARL-XRay employs a fixed backbone with incrementally allocated task-specific adapters and classifier heads, guided by a latent task selector that uses prototypes and feature-level replay to maintain historical context. Experiments on large chest X-ray datasets demonstrate that CARL-XRay achieves robust performance retention and reliable task-aware inference, outperforming joint training in task-unknown scenarios and using significantly fewer trainable parameters.
Forget retraining: a new method lets chest X-ray classifiers learn from new datasets without catastrophic forgetting, even when they don't know which dataset they're looking at.
Clinical deployment of chest radiograph classifiers requires models that can be updated as new datasets become available without retraining on previously ob- served data or degrading validated performance. We study, for the first time, a task-incremental continual learning setting for chest radiograph classification, in which heterogeneous chest X-ray datasets arrive sequentially and task identifiers are unavailable at inference. We propose a continual adapter-based routing learning strategy for Chest X-rays (CARL-XRay) that maintains a fixed high-capacity backbone and incrementally allocates lightweight task-specific adapters and classifier heads. A latent task selector operates on task-adapted features and leverages both current and historical context preserved through compact prototypes and feature-level experience replay. This design supports stable task identification and adaptation across sequential updates while avoiding raw-image storage. Experiments on large-scale public chest radiograph datasets demonstrate robust performance retention and reliable task-aware inference under continual dataset ingestion. CARL-XRay outperforms joint training under task-unknown deployment, achieving higher routing accuracy (75.0\% vs.\ 62.5\%), while maintaining competitive diagnostic performance with AUROC of 0.74 in the oracle setting with ground-truth task identity and 0.75 under task-unknown inference, using significantly fewer trainable parameters. Finally, the proposed framework provides a practical alternative to joint training and repeated full retraining in continual clinical deployment.
