Development of machine learning model in prediction of the rapid progression of interstitial lung disease in patients with idiopathic inflammatory myopathy

<bold>Objectives:</bold> This study aims to establish and validate a machine learning model to predict rapidly progressive interstitial lung disease (RP-ILD) in patients with idiopathic inflammatory myopathy-related interstitial lung disease (IIM-ILD) based on quantitative high-resolution CT (HRCT) and clinical features. <bold>Material & Methods:</bold> 511 patients (367 females, median age 54 years) with IIM-ILD were studied, with 193 showing a decline in symptoms, imaging, or physiology (pulmonary function and blood gas) within 3 months (identified as RP-ILD). Deep learning was used to predict RP-ILD based on HRCT images before treatment, and a final prediction model was created using demographic factors, pulmonary function tests, and blood gas analysis. <bold>Results:</bold> Logistic regression was selected as the best model for predicting RP-ILD due to its high AUC and interpretability. The final model showed minimal prediction bias (concordance index: 0.887, 95% confidence interval from 0.800 to 0.974, P<0.001) and provided greater net benefit in calibration and clinical decision curve analysis. The nomogram, which included variables like subtype, gender, FEV1%, DLCO%, OI, and quantitative HRCT features, showed improved benefits at various thresholds. It accurately predicted RP-ILD in patients with antisynthetase syndrome or anti-MDA5+ and remained stable in these subpopulations. <bold>Conclusion:</bold> Our study developed and confirmed machine learning models to predict RP-ILD in IIM patients, showing the connection between HRCT results, clinical features, and disease progression for practical use in healthcare.