Computed Tomography-Based Radiomics Analysis Adds Value to Glycometabolism Assessment in Adrenal Adenoma: An Exploratory Observational Study.

Abstract

Objective: This exploratory observational study examined computed tomography (CT) radiomics features associated with glucose metabolism in adrenal adenoma.

Methods: A total of 114 patients were initially enrolled, 106 who completed continuous glucose monitoring were analyzed and classified as impaired glucose metabolism (IGM; A1C ≥5.7%, n = 58) or normal glucose metabolism (NGM; n = 48). Radiomics features were extracted from noncontrast, arterial-phase, and venous-phase CT images. After reproducibility assessment, correlation filtering, and least absolute shrinkage and selection operator selection, 7 features were integrated into a Rad-score. Logistic regression evaluated the association between Rad-score and IGM after adjustment for age, body mass index, and midnight cortisol. Discrimination was assessed by receiver operating characteristic analysis, with 1,000 bootstrap resamples for internal validation.

Results: Rad-score was higher in IGM than in NGM (0.377 ± 0.510 vs -0.029 ± 0.808; P = .002). It was associated with IGM in univariable analysis (odds ratio [OR] = 2.718, 95% CI 1.360-5.433; P = .005) and remained significant after multivariable adjustment (adjusted OR = 2.681, 95% CI 1.304-5.512; P = .007). Area under the receiver operating characteristic curves were 0.623 for the clinical model, 0.673 for the radiomics model, and 0.704 for the combined model; the combined-versus-clinical difference was not significant by the DeLong test (P = .101). The bootstrap-corrected area under the receiver operating characteristic curve of the combined model was 0.606.

Conclusion: A seven-feature CT radiomics signature was independently associated with IGM in patients with adrenal adenoma. These findings are exploratory and require validation in independent cohorts.