Electrocardiographic parameter profiles for differentiating hypertrophic cardiomyopathy stages

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Artificial intelligence (AI)-enhanced electrocardiography (ECG) shows promise for detecting hypertrophic cardiomyopathy (HCM) and its dilated phase (dHCM).
• Specific ECG characteristics linked to HCM and dHCM require further characterization for improved diagnostic accuracy.

Purpose of the Study:

• To identify specific ECG parameters with high importance for detecting different subtypes of HCM and dHCM.
• To evaluate the diagnostic performance of AI-ECG models using comprehensive and reduced feature sets.

Main Methods:

• Retrospective analysis of 19,170 patients, including 140 with HCM or dHCM, from the Shinken Database (2010-2017).
• Analysis of 438 ECG parameters across P-wave, QRS complex, and ST-T segments.
• Utilized univariate and multivariate logistic regression to determine parameter importance (HPI) and diagnostic accuracy (AUROC).

Main Results:

• For HCM subtypes (basal and apical), high parameter importance (HPI) was concentrated in the ST-T segment and QRS complex.
• In dilated HCM (dHCM), HPI shifted towards the QRS complex, with lower importance in the ST-T segment.
• Area under the receiver operating characteristic curves (AUROC) for models using all ECG parameters were high across HCM subtypes (0.925–0.981).
• Models using the top 10 HPI parameters demonstrated comparable predictive performance to models using all parameters for HCM subtypes.

Conclusions:

• A shift in ECG parameter importance from the ST-T segment to the QRS complex correlates with HCM progression to dHCM.
• Efficient AI-based diagnostic models for HCM can be developed using a reduced set of top ECG parameters, achieving performance similar to comprehensive models.