AB0864 A nomogram model combining inflammatory factors and MRI radiomics to assess the disease activity of the patients with axSpA in a prospective study

Abstract

BackgroundClinical and magnetic resonance imaging (MRI) disease activity score (DAS) are measuring different aspects of axial spondyloarthritis (axSpA), they are essential in disease activity assessment. The radiomics was on facilitating readings by clinical specialists via enhancing the medical images in which subtle data differences could be distinguished.ObjectivesIf the additional information of MRI imaging can be considered as a predictor for axSpA disease activity? In this study, we sought to construct a nomogram integrating the sacroiliac joint (SIJ)- MRI radiomics features and the inflammatory biomarkers to assess disease activity and compare it with clinical disease acitivity index in axSpA patients.Methods203 patients data were collected prospectively and confirmed as axSpA were randomly divided into training (n = 143) and validation cohorts (n = 60). 1316 radiomics features were extracted from the 3.0T SIJ-MRI. A Nomogram model was constructed using multivariate logistic regression analysis Incorporating independent clinical factors and radiomics features score (Rad-score). The performance of clinics, Rad-score and nomogram models were evaluated by ROC analysis, calibration curve and decision curve analysis (DCA), and compared with the disease activity index(Ankylosing Spondylitis DAS (ASDAS)-C reactive protein (CRP), ASDAS-erythrocyte sedimentation rate (ESR), Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Bath Ankylosing Spondylitis Functional Index (BASFI)) and Spondyloarthritis Research Consortium of Canada (SPARCC) MRI scoring system.ResultsThe Rad-score allowed a good discrimination in the training (AUC, 0.91; 95% CI, 0.85-0.96) and the validation cohort (AUC, 0.84; 95% CI, 0.73-0.96). The CRP-radiomics nomogram model also showed favorable discrimination in the training (AUC, 0.96; 95% CI, 0.93-0.99) and the validation cohort (AUC, 0.89; 95% CI, 0.80-0.98), better than BASDAI(AUC, 0.58), ASDAS-CRP(AUC, 0.72), ASDAS-ESR(AUC, 0.77), ESR(AUC, 0.72), CRP(AUC, 0.77) and BASFI(AUC, 0.73), had no statistical difference with SPARCC(AUC, 0.87). Calibration curves and DCA demonstrated the nomogram fit well (p > 0.05) and was useful for activity evaluation.ConclusionRad-score showed good discriminative ability to assess disease activity in axSpA. The nomogram can increase the efficacy for assessment axSpA disease activity, which might simplify clinical evaluation.Figure 1.Comparison of ROC curve analyses in prediction models. ROC curves of the clinical features (green curve), radiomics signature model (blue curve), and hybrid model (gold curve) of axSpA in the training cohort (A) and validation cohort (B), respectively. In addition, there are AUC of ASDAS-CRP(pink curve), ASDAS-ESR(brown curve), BASDAI(purple curve), BASFI(azure curve) and SPARCC scoring system(yellow curve) in the validation cohort (B), respectively. AUC: area under the curve; ROC: receiver operating characteristic; SPARCC: Spondyloarthritis Research Consortium of Canada; BASDAI: Bath Ankylosing Spondylitis Disease Activity Index; ASDAS: Ankylosing Spondylitis Disease Activity Score; CRP: C reactive protein; ESR: erythrocyte sedimentation rate; BASFI: Bath Ankylosing Spondylitis Disease Activity Index.