Novel Hypertrophic Cardiomyopathy Diagnosis Index Using Deep Features and Local Directional Pattern Techniques.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Hypertrophic cardiomyopathy (HCM) is a genetic heart condition with varied presentations, including sudden death.
• Early diagnosis of HCM is crucial for timely intervention and preventing adverse outcomes.
• Current diagnosis relies on imaging, primarily echocardiography (ultrasound), which can be subject to human error.

Purpose of the Study:

• To develop and validate an automated computer-aided diagnostic tool for discriminating HCM from healthy subjects using ultrasound images.
• To enhance the accuracy and reduce variability in HCM diagnosis through artificial intelligence.

Main Methods:

• Utilized a local directional pattern combined with the ResNet-50 pretrained network for image classification.
• Trained and tested the model on heart ultrasound images from 62 HCM patients and 101 healthy controls.
• Identified the most significant deep feature (SigFea) using Student's t-test and derived an integrated index for diagnosis.

Main Results:

• The automated tool achieved 100% accuracy in differentiating HCM from healthy subjects in the test dataset.
• An integrated index, calculated as 100·log10(SigFea/2), effectively separated the two groups.
• A diagnostic threshold of 0.5 for the integrated index demonstrated perfect classification.

Conclusions:

• An automated AI-driven approach using local directional patterns and ResNet-50 shows promise for accurate HCM detection via ultrasound.
• This tool has the potential to improve diagnostic consistency and reduce errors in HCM identification.
• The developed method offers a reliable, high-accuracy solution for early hypertrophic cardiomyopathy diagnosis.