A Machine Learning-Based Multimodal Model Integrating Radiomics and Clinical Features for Predicting Interstitial
Yi Wang1, Ziyan Zhang2, Xiangyu Dai1
1Department of Pulmonary and Critical Care Medicine, The Affiliated Huai'an Hospital of Xuzhou Medical University, Huai'an, China (Y.W., X.D., R.F.).
Academic Radiology
|July 23, 2026
Summary
A new machine learning (ML) model combining CT radiomics and clinical data effectively predicts interstitial lung disease (ILD) risk. This multimodal approach shows promise for early ILD risk stratification in diverse patient groups.
Area of Science:
- Medical Imaging Analysis
- Machine Learning in Healthcare
- Pulmonary Medicine
Background:
- Interstitial lung disease (ILD) poses diagnostic challenges.
- Early risk stratification is crucial for timely intervention.
- Integrating diverse data sources can improve predictive accuracy.
Purpose of the Study:
- To develop and validate a multimodal machine learning (ML) model for interstitial lung disease (ILD) risk prediction.
- To integrate radiomics features from CT scans with clinical variables.
- To assess the model's performance across multiple centers.
Main Methods:
- A multicenter retrospective study included 2456 at-risk subjects.
- Radiomics features were extracted from chest CT images.
- A multimodal fusion model was developed using ML, integrating radiomics and clinical data.
Main Results:
- The multimodal fusion model demonstrated robust discriminative performance (AUC range: 0.715-0.738) across training, internal, and external validation cohorts.
- The model outperformed single-modality approaches.
- No significant performance gain was observed with more complex ML algorithms.
Conclusions:
- An ML-based multimodal model integrating CT radiomics and clinical features provides stable ILD risk prediction.
- This approach holds potential for early ILD risk stratification.
- The model's multicenter validation supports its generalizability.