Development and Implementation of an Artificial Intelligence-Driven Multimodal Skeletal Muscle Feature Fusion Model for Risk Prediction of Sudden Cardiac Death in Patients With Implantable Cardioverter-Defibrillators: The SMART-SCD Study.

This study is designed as a prospective, single-center, observational cohort study (the SMART-SCD Study, full name: Skeletal Muscle Multi-omics Analysis and Risk Tailoring in Sudden Cardiac Death), which enrolls high-risk populations meeting the criteria for implantable cardioverter defibrillator (ICD) implantation.

The research focuses on the mechanistic association between skeletal muscle metabolic disorders and ventricular arrhythmia (VA) as well as sudden cardiac death (SCD), and aims to construct a "muscle-heart crosstalk" risk early warning system through integration of multimodal skeletal muscle data. We will systematically collect the following data:

Baseline handgrip strength measurement (Biomi-h500+X5); Functional diagnosis and phenotyping of sarcopenia conducted via the InBody 270 body composition analyzer; Non-contrast chest and abdominal computed tomography (CT) images (to extract novel imaging phenotypes including skeletal muscle density at the T12 vertebra level, intermuscular adipose tissue, subcutaneous adipose tissue, etc.); Serum biomarkers (GDF-8, Irisin, IL-6); Metabolomics data of skeletal muscle tissue from the ICD pocket (lipid/energy metabolism profiles detected via the UPLC-QTOF/MS platform); Ambulatory electrocardiographic data. All treatment and intervention regimens for patients will be independently formulated by clinicians in accordance with clinical guidelines, and the study itself does not involve any intervention measures. Prospective follow-up will be conducted at 3/6/12 months after ICD implantation. The primary endpoint is composite ventricular arrhythmia events (including SCD, appropriate ICD therapy documented by the device, and hemodynamically unstable ventricular tachycardia/ventricular fibrillation), and the secondary endpoint is all-cause mortality.

Through the above prospective cohort study, we will integrate multimodal data including novel CT imaging phenotypes of skeletal muscle, metabolomics profiles and functional phenotyping of sarcopenia using artificial intelligence techniques, so as to construct a precision prediction model for SCD, screen novel CT imaging phenotypes of sarcopenia and myogenic metabolites, and finally establish a generalizable SCD risk assessment tool and individualized intervention strategies.