Single-Phase CT with Clinical Data Matches Dual-Phase CT for COPD Staging: Study

Predictions from intraclass correlation coefficient (ICC) on the assessment of spirometry, and accuracy measures for the classification into the GOLD stage

Key findings

Agreement Between CNN Predictions and Reference Spirometry:-

• CNN-predicted spirometry values were of moderate to good agreement compared to the reference spirometry values, with an ICC ranging between 0.66–0.79.

• The inclusion of clinical data increased the agreement (ICC, 0.70–0.85; p ≤0.04), except for FEV1/FVC in the inspiratory-phase CNN model (p=0.35) and FEV1 in the expiratory-phase CNN model (p =0.33).

Accuracy by GOLD Stage

• Single-phase CT CNNs had moderate to good agreement (ICC, 0.68–0.70) with accuracies of 59.8% to 84.1% (682–959 of 1140).

• Dual-phase CT CNNs had higher accuracies of 60.0% to 86.3% (684–984 of 1140) with an ICC of 0.72.

Impact of Clinical Data:

Inclusion of clinical data markedly improved both single-phase and dual-phase CNN models:

• The single-phase CNN achieved an accuracy of 65.2%–85.8% (743–978 of 1140) with an ICC of 0.72.

• The dual-phase CNN achieved an accuracy of 67.6%–88.0% (771–1003 of 1140) with an ICC of 0.77–0.78.

• But, for the expiratory CNN model, no improvement in GOLD stage ICC was seen (p=0.08).

Single-phase CT combined with clinical data yields accuracy comparable to dual-phase CT for CNN-based COPD staging, thus providing a streamlined and effective approach for disease diagnosis and management. This approach represents a significant step forward in integrating imaging and clinical data for advanced diagnostic applications.

Reference:

Lee, A. N., Hsiao, A., & Hasenstab, K. A. (2024). Evaluating the cumulative benefit of inspiratory CT, expiratory CT, and clinical data for COPD diagnosis and staging through deep learning. Radiology. Cardiothoracic Imaging, 6(6). https://doi.org/10.1148/ryct.240005