Abstract
Pulmonary function, commonly assessed using FEV1 is a key indicator of COPD severity and progression. Although recent deep learning methods can estimate pulmonary function from chest CT, most rely solely on structural imaging and do not explicitly model how structural and functional information jointly contribute to pulmonary function. We propose a region-wise gated multimodal framework that integrates structural CT and CT-derived ventilation imaging for longitudinal prediction of post-bronchodilator FEV1% predicted. Modality-specific features are extracted using independent 3D convolutional encoders and adaptively fused across anatomically corresponding lung regions through a learned gating mechanism. Experiments on 5,263 subjects from the COPDGene cohort demonstrate that the proposed model outperforms both CT-only and ventilation-only baselines, achieving an RMSE of 14.58 (95% CI: 13.82–15.25) and a Pearson correlation of 0.813 (95% CI: 0.789–0.836) for predicting 5-year postbronchodilator FEV1% predicted from baseline imaging. Analysis of the learned fusion gates revealed decreasing regional fusion heterogeneity with increasing GOLD severity, suggesting evolving structure–function relationships during disease progression. These findings demonstrate that multimodal fusion of structural and functional imaging improves pulmonary function prediction, while regionwise gating provides interpretable insight into regional structure–function interactions in COPD.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/TIA_019.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{MelAla_Interpretable_MICCAISAT2026,
author = { Melek, Alaa AND Sheshadri, Ajay AND Zhao, Lili AND Nair, Girish AND Castillo, Edward},
title = { { Interpretable Region-wise Gated CT–Ventilation Fusion for FEV1 Prediction in COPD } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17269},
month = {pending},
page = {pending}
}