Machine learning detection of obstructive hypertrophic cardiomyopathy using a wearable biosensor.

Area of Science:

• Cardiology
• Biomedical Engineering
• Medical Diagnostics

Background:

• Hypertrophic cardiomyopathy (HCM) is a genetic heart muscle disease with significant risks, including heart failure and sudden death.
• Current diagnosis rates are low (10-20%), highlighting the need for accessible, non-clinical screening methods.
• Photoplethysmography (PPG) is a noninvasive optical technique measuring blood volume changes, commonly available in smartwatches.

Purpose of the Study:

• To evaluate the efficacy of PPG-based analysis for detecting obstructive HCM (oHCM).
• To develop and validate a machine learning classifier for oHCM screening using PPG data.

Main Methods:

• Collected PPG recordings and echocardiograms from 19 oHCM patients and 64 healthy controls.
• Utilized automated analysis to compare 42 morphometric pulse wave features between groups.
• Developed a machine learning classifier to distinguish oHCM from controls.

Main Results:

• Significant differences (38/42 features) were observed in pulse wave morphology between oHCM patients and controls.
• The machine learning classifier achieved a high C-statistic of 0.99 for oHCM detection.
• Key differentiating features included systolic ejection time, rate of rise, and respiratory variation.

Conclusions:

• PPG analysis combined with machine learning shows promise as a noninvasive screening tool for oHCM.
• This technology could significantly improve early detection rates for hypertrophic cardiomyopathy.
• Further development may lead to a widely accessible screening method for obstructive HCM via smartwatches.