Comprehensive multi-organ and body composition phenotyping identifies metabolic signatures associated with incident type 2 diabetes.
Authors
Affiliations (11)
Affiliations (11)
- Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea.
- Department of Biomedical Systems Informatics, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Department of Radiology, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea.
- Department of Convergence Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea.
- Department of Endocrinology, Internal Medicine, Yongin Severance Hospital, Yonsei University College of Medicine, Yongin, Republic of Korea.
- Department of Radiation Oncology, Heavy Ion Therapy Research Institute, Yonsei Cancer Center, Yonsei University Health System, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Department of Biomedical Informatics, College of Medicine, Gachon University, Incheon, Republic of Korea.
- Department of Biomedical Systems Informatics, Yonsei University College of Medicine, Seoul, Republic of Korea. [email protected].
- Institute for Innovation in Digital Healthcare, Severance Hospital, Seoul, Republic of Korea. [email protected].
- Center for Digital Health, Yongin Severance Hospital, Yonsei University Health System, Yongin, Republic of Korea. [email protected].
Abstract
To derive a multi-organ imaging biomarker panel from whole-body MRI and evaluate its predictive value for incident type 2 diabetes mellitus (T2DM), including the role of intermuscular adipose tissue (IMAT) in reconciling discrepancies between muscle volume and metabolic risk. This retrospective cohort study analyzed 46,129 UK Biobank participants without diabetes at the time of imaging. Organ volume indices and fat fractions were extracted from Dixon MRI through automated segmentation. Associations between muscle volume indices, fat fractions, and handgrip strength were evaluated, and IMAT was graded in strength-matched participants. Independent imaging biomarkers were identified using adjusted Cox models, and a deep learning model was trained. Added value relative to the Leicester risk assessment score was evaluated using decision curve analysis. During follow-up, 600 participants developed T2DM. Eight biomarkers independently predicted incident T2DM, including liver volume index (hazard ratio (HR), 1.28; 95% confidence interval (CI), 1.19-1.38), pancreas fat fraction (HR, 1.19; 95% CI, 1.10-1.28), and gluteus fat fraction (HR, 1.17; 95% CI, 1.08-1.27). Gluteal IMAT grades were higher in weak-for-volume participants than in strength-matched controls (median, 2.0 vs 1.0; p = 0.02). The deep learning model achieved a concordance index of 0.814 (95% CI, 0.797-0.832) and, when applied sequentially with clinical criteria, reduced false-positive classifications by 43.0% while preserving 97.3% of true positives. Whole-body MRI-derived multi-organ biomarkers independently predicted incident T2DM and enhanced risk stratification beyond an established clinical risk score. Muscle fat fraction may better reflect metabolic risk than muscle volume in IMAT-rich regions. Question Clinical risk scores miss the multi-organ pathophysiology of type 2 diabetes; does whole-body MRI organ size and fat quantification improve prediction of incident diabetes? Findings Eight MRI-derived biomarkers independently predicted incident type 2 diabetes; a deep learning model using whole-body MRI biomarkers reached a concordance index of 0.814. Clinical relevance Automated whole-body MRI biomarkers capture multi-organ metabolic signatures beyond conventional risk scoring. Applied after clinical screening, they remove 43.0% of false-positive classifications while preserving 97.3% of true positives.