CT-based renal and body-composition radiomics model to improve the detection ability of diabetic kidney disease in patients with type 2 diabetes mellitus.

OBJECTIVES: Conventional imaging modalities exhibit limited capability in distinguishing diabetic kidney disease (DKD), but radiomics could expand image-derived information. Kidney, perirenal tissue and body composition, whether considered in macroscopic or microscopic profile, are associated with kidney function. The present study seeks to evaluate the efficacy of computed tomography (CT)-derived radiomic features of the kidney and body composition in detecting DKD among individuals with type 2 diabetes mellitus (T2DM). METHODS: Patients with T2DM who underwent abdominal CT and renal function test at two institutions were enrolled. Participants from one institution (n = 256) were randomly allocated into training and internal validation cohorts, whereas individuals from another institution constituted the external validation cohort (n = 64). Two- and three-dimensional radiomics features were retrieved from kidneys, perirenal and renal sinus fat, visceral and subcutaneous fat, and skeletal muscles. Radiomics, clinical, and integrated models were developed and evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves and decision curve analysis. RESULTS: The radiomics model achieved impressive AUCs of 0.869 (95% CI, 0.809-0.929), 0.831 (95% CI, 0.721-0.940), and 0.833 (95% CI, 0.687-0.979) in the training cohort, internal and external validation cohorts, respectively. The combined model outperformed the other two models, achieving AUCs (95% CI) of 0.913 (0.868-0.957), 0.919 (0.849-0.989), and 0.867 (0.762-0.973) in the training, internal validation, and external validation cohorts, respectively. CONCLUSION: CT-based renal and perirenal radiomics features demonstrate strong efficacy in identifying DKD in patients with T2DM.