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

Abstract

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)- magnetic resonance imaging (MRI). Methods 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 3.0T 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 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 model were evaluated by ROC analysis, calibration curve and decision curve analysis (DCA). Results 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 favorable 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.

Diagnostic clinical prediction rules for categorising low back pain: A systematic review

“…(2021), and Ye et al. (2022) have all shown strong positive evidence and low risk of bias during the early stages of development.…”

“…Table 7 presents the RoB of identified studies ( n = 16) using the QUADAS‐2 tool. Five studies had a low overall RoB (Aghaei et al., 2015; Benditz et al., 2019, 2021; Kato et al., 2015; Roux et al., 2007; Ye et al., 2022) and nine studies demonstrated acceptable applicability (Aghaei et al., 2015; Benditz et al., 2019, 2021; Chatprem et al., 2021; Kato et al., 2009, 2015; Roux et al., 2007; Tominaga et al., 2022; Ye et al., 2022).…”

Development of clinical and magnetic resonance imaging-based radiomics nomograms for the differentiation of nodular fasciitis from soft tissue sarcoma

“…Radiomic analysis involving clinical factors and imaging demonstrates good clinical value in a variety of diseases (31)(32)(33)(34)(35). By extracting the high-throughput quantitative features from MRI scans, radiomics allows the non-invasive interpretation of tumor heterogeneity and biological characteristics (22,36).…”

Sacroiliitis diagnosis based on interpretable features and multi-task learning

“…Sacroiliitis is visually assessed on CT scans using gray scale descriptors, texture descriptors, which are effective prior features for mitigating the risk of over-fitting. As a result, radiomics has been applied extensively in the diagnosis of sacroiliitis (Faleiros 2018, Castro-Zunti et al 2020, Tenório et al 2020, Kepp et al 2021, Ye et al 2022. Ye et al (2022) improved the diagnostic accuracy from 74% to 78% using a model that combined the features of radiomics and clinical practice.…”