Digital Poster 8: Deep Learning-Based Cardiac MRI Analysis for Hypertrophic Cardiomyopathy Detection

Background: Hypertrophic Cardiomyopathy (HCM) is a prevalent genetic cardiovascular disorder, distinguished by LV hypertrophy and distinct fibrosis patterns within the myocardium.[1] Although Late Gadolinium Enhancement (LGE) in cardiovascular magnetic resonance (CMR) remains the gold standard for fibrosis detection in HCM, it poses certain limitations, including retention in the brain.[2] Consequently, there is a growing interest in needle-free CMR sequences.[3, 4] This study leverages AI to improve CMR efficiency in HCM by identifying pathological patterns in the myocardium in native cine images. Our goal was to use deep learning techniques to distinguish between fibrotic and healthy tissue, while also visualizing fibrosis patterns through heat maps.

Methods: We employed a transfer learning approach using a 3D ResNet architecture, tailored to analyze both systolic and diastolic phases of HCM in 10-slice SAX cine stacks from 23 HCM patients and 28 healthy individuals. Leveraging the deep residual learning framework of the ResNet, pre-trained on the ImageNet database, we achieved robust feature extraction essential for distinguishing between pathological and normal myocardial tissues. The model was optimized with the Adam optimizer at a learning rate of 0.000295, incorporating a 50% dropout rate to enhance generalization and prevent overfitting. Training was conducted over 50 epochs.

Results: After training, the model demonstrated exceptional discriminative capability, achieving a validation accuracy, precision, and recall of 92.7% each, complemented by an outstanding ROC-AUC score of 0.992 . Significantly, heat maps generated from AI feature maps were used to visualize and identify fibrosis patterns in the cine images, closely matching the abnormalities observed in Late Gadolinium Enhancement (LGE), the current gold standard (Figure 1).

Conclusion:

Our study demonstrated that a 3D ResNet architecture can effectively differentiate between HCM and normal cases by analyzing cardiac cine MRI images during the systolic and diastolic phases. The AI model successfully identified myocardial fibrosis. This finding underscores the potential for AI-driven models for using native CMR images for tissue characterization and eventually allow for needle-free and highly efficient CMR scans.