Integrated imaging-informed and machine learning-based prognostic models for risk stratification in connective tissue...

Area of Science:

• Pulmonology and Radiology
• Medical Informatics

Background:

• Connective tissue disease-associated interstitial lung disease (CTD-ILD) has variable outcomes, and current prognostic tools like the ILD-GAP score lack precision.
• There is a need for improved prognostic models that incorporate imaging data to better stratify CTD-ILD patient risk.

Purpose of the Study:

• To develop and externally validate integrated prognostic systems for CTD-ILD risk stratification.
• To compare the prognostic discrimination of imaging-derived fibrosis measures against traditional physiology-based models.

Main Methods:

• A multicenter retrospective study enrolled patients with CTD-ILD.
• Three prognostic systems were evaluated: System A (ILD-GAP and fibrosis score), System B (composite model), and System C (machine learning models).
• Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC) with internal and external validation.

Main Results:

• The fibrosis score showed better discrimination than the ILD-GAP index in external validation (AUC 0.798 vs. 0.768).
• Machine learning models achieved the highest discrimination (Random Forest AUC 0.833, Support Vector Machine AUC 0.826).
• SHAP analysis identified fibrosis extent, DLCO, and age as key predictors of adverse outcomes.

Conclusions:

• Imaging-informed and data-driven models significantly improve prognostic discrimination in CTD-ILD compared to traditional indices.
• The fibrosis score is crucial for outcome prediction, with composite and machine learning models offering refined, individualized risk stratification.
• Combining imaging (fibrosis extent) and physiological data (DLCO, age) enhances prognostic accuracy for CTD-ILD.