Uncovering Latent Pathological Signatures in Pulmonary CT via Cross-Window Knowledge Distillation

TL;DR

This paper introduces a cross-window knowledge distillation method for pulmonary CT analysis, enabling models to learn latent pathological signatures across different density windows, significantly improving diagnostic accuracy.

Contribution

It proposes a novel cross-window knowledge distillation framework that captures cross-density interactions in multi-window CT imaging, outperforming existing methods.

Findings

Distillation improved per-window AUC by 10.1-16.5 percentage points.

Ensemble AUC reached 0.9960 on COPD-CT-DF dataset.

Significant gains observed across multiple cohorts.

Abstract

Multi-window CT imaging captures complementary pathological information across anatomical structures of differing densities, yet existing deep learning methods fuse representations only at later stages, missing cross-density interactions. We propose a cross-window knowledge distillation framework in which student encoders learn latent clinical priors from a teacher trained on the most informative window. Evaluated retrospectively on three cohorts - COPD-CT-DF (n=719), RSNA PE (n=1,433), and an in-house CTEPD dataset (n=161) - distillation improved per-window AUC by 10.1-16.5 percentage points on COPD-CT-DF (0.75-0.81 to 0.90-0.94; all P<0.001), with ensemble AUC reaching 0.9960. Similar gains were observed on RSNA PE (0.80-0.83 to 0.90-0.92) and CTEPD (AUC 0.7481 vs. 0.6264). Cross-window distillation internalises pathological signatures invisible to supervised approaches, offering a generalisable solution for multi-window pulmonary CT analysis.