This project evaluates the diagnostic accuracy of standard and novel ECG criteria for detecting cardiac structural abnormalities, validated against cardiac MRI reference standards available in UK Biobank. Primary research questions include: (1) What are the sensitivity and specificity of conventional ECG voltage criteria (e.g., Sokolow-Lyon, Cornell) for left ventricular hypertrophy (LVH) when benchmarked against cardiac MRI-derived left ventricular mass index (LVMI) and geometry? (2) How do advanced ECG measures (e.g., QRS fragmentation, T-wave morphology, global electrical heterogeneity) perform for identifying structural remodeling, left ventricular systolic/diastolic dysfunction, and atrial enlargement when compared with cardiac MRI parameters? (3) Can machine-learning ECG models improve diagnostic accuracy over traditional thresholds across demographic subgroups?
Objectives: derive optimal ECG thresholds against MRI phenotypes; evaluate diagnostic accuracy using STARD principles; perform subgroup analyses by age, sex, BMI, ethnicity, and comorbidities; and develop validated ECG-based prediction tools for structural heart disease.
Scientific rationale: ECG is globally available, inexpensive, and scalable, but its accuracy for structural phenotyping remains uncertain because prior validation cohorts lacked large-scale MRI reference imaging. UK Biobank’s extensive ECG data paired with the world’s largest population-based cardiac MRI resource enables robust evaluation of ECG diagnostic performance.
Expected outputs include peer-reviewed publications defining evidence-based ECG criteria and externally validated machine-learning models that may support earlier detection and improved cardiovascular risk stratification.