A predictive clinical-radiomics nomogram for diagnosing of axial spondyloarthritis using MRI and clinical risk factors

Objectives Construct and validate a nomogram model integrating the radiomics features and the clinical risk factors to differentiating axial spondyloarthritis (axSpA) in low back pain patients undergone sacroiliac joint (SIJ)-MRI. Methods A total of 638 patients confirmed as axSpA (n = 424) or non-axSpA (n = 214) who were randomly divided into training (n = 447) and validation cohorts (n = 191). Optimal radiomics signatures were constructed from the 3.0 T SIJ-MRI using maximum relevance-minimum redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithm in the training cohort. We also included six clinical risk predictors to build the clinical model. Incorporating the independent clinical factors and Rad-score, a nomogram model was constructed by multivariable logistic regression analysis. The performance of the clinical, Rad-score, and nomogram models were evaluated by ROC analysis, calibration curve and decision curve analysis (DCA). Results A total of 1316 features were extracted and reduced to 15 features to build the Rad-score. The Rad-score allowed a good discrimination in the training (AUC, 0.82; 95% CI: 0.77, 0.86) and the validation cohort (AUC, 0.82; 95% CI: 0.76, 0.88). The clinical-radiomics nomogram model also showed favourable discrimination in the training (AUC, 0.90; 95% CI: 0.86, 0.93) and the validation cohort (AUC, 0.90; 95% CI: 0.85, 0.94). Calibration curves (P >0.05) and DCA demonstrated the nomogram was useful for axSpA diagnosis in the clinical environment. Conclusion The study proposed a radiomics model was able to separate axSpA and non-axSpA. The clinical-radiomics nomogram can increase the efficacy for differentiating axSpA, which might facilitate clinical decision-making process.