Skeletal Muscle Ultrasound Radiomics and Machine Learning for the Earlier Detection of Type 2 Diabetes Mellitus

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Endocrinology

Background:

• Ultrasound (US) assessment of deltoid muscle echogenicity can aid in early type 2 diabetes (T2D) detection.
• Traditional hemoglobin A1c (HbA1c) testing has limitations in early T2D and prediabetes (PreD) screening.

Purpose of the Study:

• To evaluate the utility of automated skeletal muscle US radiomics and machine learning for earlier detection of T2D and PreD.
• To assess the potential of US radiomics as a supplemental screening tool alongside HbA1c testing.

Main Methods:

• A cohort of 1191 patients undergoing shoulder US was analyzed, categorized into normal, screening, risk, PreD, and diabetes groups.
• Deltoid muscle US images were processed using automatic region of interest detection.
• Radiomics features, race, age, and BMI were used as inputs for a gradient-boosted decision tree model to predict T2D risk.

Main Results:

• The combined radiomics and clinical features model achieved a mean area under the receiver operating characteristic (AUROC) of 0.86 (71% sensitivity, 96% specificity).
• In patients with obesity, the model achieved an AUROC of 0.92 (82% sensitivity, 95% specificity).

Conclusions:

• Skeletal muscle US radiomics and machine learning demonstrate significant potential for earlier T2D detection.
• This noninvasive approach can serve as a valuable supplemental screening tool for opportunistic T2D and PreD identification.