On-Device Continual Learning with Dual-Stage Buffer and Dynamic Loss for Point-of-Care Pneumonia Diagnosis

TL;DR

PneumoNet is a lightweight, on-device continual learning method that maintains high pneumonia diagnosis accuracy across domain shifts using a dual-stage buffer and dynamic loss, suitable for resource-limited settings.

Contribution

The paper introduces PneumoNet, a novel on-device continual learning approach combining a dual-stage buffer and dynamic loss for robust pneumonia diagnosis under domain shifts.

Findings

Achieves 86.6% accuracy on domain-shifted PneumoniaMNIST dataset.

Maintains only 1.4% forgetting during continual learning.

Smaller and faster than existing baseline methods.

Abstract

Deep learning models detect pneumonia from chest X-rays with high accuracy, but the performance declines under domain shifts caused by differences in devices, patients, or institutions. We present PneumoNet, a domain-incremental learning method for point-of-care pneumonia diagnosis in resource-limited settings. PneumoNet combines a lightweight CNN for on-device prediction, a dual-stage balanced buffer for class-balanced replay, and a dynamic class-weighted loss to correct training-batch imbalances. Evaluated on a domain-shifted PneumoniaMNIST dataset simulating five realistic domain change scenarios, PneumoNet achieves 86.6% accuracy with 1.4% forgetting while being smaller and faster than existing baselines. These results highlight PneumoNet's potential to enable adaptive, privacy-preserving diagnostic AI directly on point-of-care medical devices in real-world and pandemic-ready healthcare.