Researchers built a small wearable electrocardiogram (ECG) device and a simple analysis method that together aim to tell hypertrophic cardiomyopathy (HCM) apart from acquired left ventricular hypertrophy (LVH) using ECG signals alone. HCM is a genetic cause of abnormal heart muscle thickening and a leading cause of sudden death in young athletes. Existing tests such as cardiovascular magnetic resonance (CMR), echocardiography, or genetic tests are costly, operator-dependent, or slow, so a low-cost screening tool could be useful in many settings.

The team assembled a portable system with three surface electrodes placed on the torso to record Lead I of the ECG. The electronics include an AD8232 signal conditioning module, an Arduino Nano 33 BLE microcontroller (Bluetooth Low Energy), and a lithium polymer (LiPo) battery in a small enclosure. The authors chose the three-lead, torso-mounted setup for ease of use and to preserve ECG features that relate to heart muscle thickness.

On the software side, the algorithm reduces each heartbeat to two numbers called HCM Index 1 and HCM Index 2. The system classifies a patient by comparing those indices to two statistical thresholds. To test the method, the researchers used ECG records from 483 patients with acquired LVH from PhysioNet and 29 HCM patients taken from digitized clinical records. Using this data, the method achieved 75.86% sensitivity (it found about three quarters of true HCM cases) and 99.17% specificity (it flagged very few non-HCM cases). The reported overall F1-score, a combined measure of precision and recall, was 80.00%.

The authors also ran leave-one-out cross-validation to check how well the result would generalize to new patients. That test gave similar numbers: 72.41% sensitivity, 98.96% specificity, and an F1-score of 76.36%, with 95% confidence intervals reported in the paper. They checked for possible confounds from using different data sources by performing a digitization confound analysis and concluded the classifier responds to physiological ECG features rather than artifacts from how the records were captured. A simulated device acquisition chain also suggested the wearable hardware can provide signals compatible with the algorithm.