Unmasking myocardial fibrosis by cardiac MRI in patients with frequent ventricular extrasystoles and normal echocardiography
R Brandao, I Miranda, F Gerardo, L Cotrim, I Madeira Santos, J Bicho AugustoAbstract
Background
Frequent ventricular extrasystoles (VEs) with normal echocardiogram findings often prompt cardiac MRI evaluation to detect myocardial fibrosis.
Objective
To characterize (1) the prevalence and location of fibrosis and (2) to assess the predictors of fibrosis in such patients.
Methods
We retrospectively analyzed patients referred for CMR for frequent VEs with normal echocardiograms between January 2022 and November 2024. Fibrosis was indicated by late gadolinium enhancement (LGE). Feature selection via SelectKBest (ANOVA F-test) retained the top five predictors, and a logistic regression model with L2 regularization was developed. Data were split (70% training, 30% testing) and validated using stratified 5-fold cross-validation. Model performance was assessed via precision, recall, F1-score, and accuracy.
Results
Among 98 patients (54.3±12.1 years, 55% male), 26.5% exhibited fibrosis. Of these, 19.2% had subendocardial fibrosis, consistent with a myocardial infarct pattern, while 80.8% had midwall and/or subepicardial fibrosis. The basal inferolateral and basal inferior regions were the most affected (23%). Patients with fibrosis were older (58.2±10.4 vs 53.0±12.8 years, p=0.03). Predictors included age (p=0.01), hypertension (p=0.02), obesity (p=0.04) and left ventricular ejection fraction (p=0.03). The model achieved 98.8% cross-validation accuracy and 70.4% accuracy on the holdout set. Non-fibrosis cases were reliably predicted (precision 77.3%, recall 85%, F1-score 80.9%), whereas performance for fibrosis cases was reduced (precision 40%, recall 28.6%, F1-score 33.3%).
Conclusion
Myocardial fibrosis was detected in over a quarter of patients with frequent VEs and normal echocardiograms, predominantly in the basal inferolateral region. Older age, hypertension, obesity, and reduced ejection fraction were key predictors. Cardiac MRI and clinical integration remain crucial for risk assessment.