Prediction of Major Adverse Cardiovascular Events in Patients with Hypertrophic Cardiomyopathy by Deep Learning and...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Hypertrophic cardiomyopathy (HCM) patients face risks of major adverse cardiovascular events (MACE).
• Accurate risk stratification is crucial for timely intervention in high-risk HCM patients.
• Existing methods lack comprehensive predictive capabilities for MACE in HCM.

Purpose of the Study:

• To develop and validate a predictive model for MACE in HCM patients.
• To combine deep transfer learning (DTL) and radiomics for enhanced risk prediction.
• To assess the clinical utility of the novel DTL-RAD model.

Main Methods:

• Retrospective analysis of 210 HCM patients with a mean follow-up of 29.44 months.
• Extraction of radiomic features from echocardiographic images and DTL features using Resnet50.
• Development of the DTL-RAD model using selected features and machine learning algorithms.

Main Results:

• The DTL-RAD model achieved excellent diagnostic performance (AUC 0.936 training, 0.918 validation).
• The model significantly outperformed standalone radiomics or DTL models.
• Decision curve analysis confirmed superior clinical applicability and effectiveness of the DTL-RAD model.

Conclusions:

• The DTL-RAD model accurately identifies HCM patients at high risk for MACE.
• Early detection of high-risk individuals facilitates precise clinical interventions.
• This approach can effectively reduce MACE incidence in HCM patients.