Construction of a Clinico-Imaging Collaborative Diagnostic Model for Dermatomyositis Combined With Interstitial Lung Disease Based on PET/CT Imaging Features and Clinical Parameters

For the features included in the final optimal model, between-group comparisons of continuous variables (interstitial lung disease group vs. non-interstitial lung disease group) were performed using the Wilcoxon rank-sum test. For categorical variables, the Chi-square test or Fisher's exact test was adopted as appropriate.In the comparison of model efficacy, the DeLong test was used to assess the statistical differences in AUC values between each machine learning classifier and the reference model.All statistical analyses were conducted using R software (version 4.4.1). The corresponding R packages applied included pROC for ROC analysis, caret for model training, and SHAP for the interpretability analysis of the XGBoost model. A two-tailed p-value < 0.05 was defined as the threshold of statistical significance for all analyses.

Inclusion criteria

• The diagnosis of dermatomyositis (DM) was made in accordance with the Bohan and Peter criteria
• The diagnosis of clinically amyopathic dermatomyositis (CADM) was established based on the Sontheimer criteria
• The diagnosis of interstitial lung disease (ILD) was confirmed in line with the criteria of the American Thoracic Society (ATS)
• ⁶⁸Ga-FAPI and ¹⁸F-FDG PET/CT scans were performed in the Department of Nuclear Medicine.

Exclusion criteria

Patients with other connective tissue diseases.