Machine learning phenotyping of scarred myocardium from cine in hypertrophic cardiomyopathy

Aims Cardiovascular magnetic resonance (CMR) with late-gadolinium enhancement (LGE) is increasingly being used in hypertrophic cardiomyopathy (HCM) for diagnosis, risk stratification, and monitoring. However, recent data demonstrating brain gadolinium deposits have raised safety concerns. We developed and validated a machine-learning (ML) method that incorporates features extracted from cine to identify HCM patients without fibrosis in whom gadolinium can be avoided. Methods and results An XGBoost ML model was developed using regional wall thickness and thickening, and radiomic features of myocardial signal intensity, texture, size, and shape from cine. A CMR dataset containing 1099 HCM patients collected using 1.5T CMR scanners from different vendors and centres was used for model development (n=882) and validation (n=217). Among the 2613 radiomic features, we identified 7 features that provided best discrimination between +LGE and −LGE using 10-fold stratified cross-validation in the development cohort. Subsequently, an XGBoost model was developed using these radiomic features, regional wall thickness and thickening. In the independent validation cohort, the ML model yielded an area under the curve of 0.83 (95% CI: 0.77–0.89), sensitivity of 91%, specificity of 62%, F1-score of 77%, true negatives rate (TNR) of 34%, and negative predictive value (NPV) of 89%. Optimization for sensitivity provided sensitivity of 96%, F2-score of 83%, TNR of 19% and NPV of 91%; false negatives halved from 4% to 2%. Conclusion An ML model incorporating novel radiomic markers of myocardium from cine can rule-out myocardial fibrosis in one-third of HCM patients referred for CMR reducing unnecessary gadolinium administration.

Influence of longitudinal position on the evolution of steady-state signal in cardiac cine balanced steady-state free precession imaging

Background Emerging quantitative cardiac magnetic resonance imaging (CMRI) techniques use cine balanced steady-state free precession (bSSFP) to measure myocardial signal intensity and probe underlying physiological parameters. This correlation assumes that steady-state is maintained uniformly throughout the heart in space and time. Purpose To determine the effects of longitudinal cardiac motion and initial slice position on signal deviation in cine bSSFP imaging by comparing two-dimensional (2D) and three-dimensional (3D) acquisitions. Material and Methods Nine healthy volunteers completed cardiac MRI on a 1.5-T scanner. Short axis images were taken at six slice locations using both 2D and 3D cine bSSFP. 3D acquisitions spanned two slices above and below selected slice locations. Changes in myocardial signal intensity were measured across the cardiac cycle and compared to longitudinal shortening. Results For 2D cine bSSFP, 46% ± 9% of all frames and 84% ± 13% of end-diastolic frames remained within 10% of initial signal intensity. For 3D cine bSSFP the proportions increased to 87% ± 8% and 97% ± 5%. There was no correlation between longitudinal shortening and peak changes in myocardial signal. The initial slice position significantly impacted peak changes in signal intensity for 2D sequences ( P < 0.001). Conclusion The initial longitudinal slice location significantly impacts the magnitude of deviation from steady-state in 2D cine bSSFP that is only restored at the center of a 3D excitation volume. During diastole, a transient steady-state is established similar to that achieved with 3D cine bSSFP regardless of slice location.