Domain Incremental Learning for Pandemic-Resilient Chest X-Ray Analysis

TL;DR

This paper presents a replay-based continual learning method that enables pneumonia detection models to adapt across different clinical domains without forgetting previous knowledge.

Contribution

It introduces a class-aware balanced replay and loss mechanism to improve cross-domain generalization in chest X-ray analysis.

Findings

Achieved an average accuracy of 88.66% on a domain-shifted PneumoniaMNIST dataset.

Outperformed Experience Replay, Fine-Tuning, and Joint Training baselines.

Demonstrated robustness and consistency in pneumonia detection across varied clinical environments.

Abstract

Deep learning models achieved high accuracy in pneumonia detection from chest X-rays. However, their generalization across clinical domains remains limited due to variations in imaging devices, acquisition protocols, and institutional conditions. This study introduces a replay-based domain-incremental continual learning designed to enable continual adaptation to cross-domain variations without catastrophic forgetting. The proposed method incorporates a class-aware balanced replay to maintain balanced class representation within a constrained memory and a class-aware loss to dynamically reweight class imbalance during training. Experiments conducted on a domain-shifted PneumoniaMNIST dataset consisting of five simulated domains demonstrate that the proposed method achieves an average accuracy of 88.66%, outperforming Experience Replay, Fine-Tuning, and Joint Training baselines. These findings highlight the efficacy of the proposed approach in achieving robust and consistent pneumonia detection across clinical environment variations.