Momentum-Anchored Multi-Scale Fusion Model for Long-Tailed Chest X-Ray Classification
For medical image classification, this work addresses feature drift in long-tailed distributions, improving rare disease detection.
The paper tackles class imbalance in chest X-ray classification, where gradient updates bias toward majority classes. The proposed Momentum-Anchored Multi-Scale Fusion Network achieves 0.8682 average AUC on ChestX-ray14, outperforming SOTA with notable gains on rare pathologies like Hernia (0.9470) and Pneumonia (0.8165).
Chest X-ray classification suffers from severe class imbalance where gradient updates bias toward majority classes, causing feature drift and poor performance on rare but critical pathologies. We propose a Momentum-Anchored Multi-Scale Fusion Network that uses exponential moving averages (EMA) as a temporal anchoring mechanism to stabilize feature representations under long-tailed distributions. Our approach applies selective momentum updates to the final expansion block of an EfficientNet backbone, creating a slowly-evolving reference branch that resists gradient-induced drift while preserving discriminative patterns for minority classes. Combined with multi-scale spatial fusion ($1\times 1$, $3 \times 3$, $5 \times 5$ convolutions), this anchoring strategy maintains representational stability throughout training. On ChestX-ray14, our method achieves 0.8682 average AUC, outperforming state-of-the-art approaches and showing particular improvements on rare pathologies like Hernia (0.9470) and Pneumonia (0.8165). The results demonstrate that momentum anchoring effectively counters feature instability in long-tailed medical image classification.