Artificial Intelligence Based ECG Scores Can Predict Response to Treatment in Hypertrophic Obstructive, Is This the Future?

Historically, early detection of hypertrophic cardiomyopathy (HCM) has been challenging, and prior artificial intelligence (AI) based electrocardiography (ECG) detection algorithms have proven helpful in the early diagnosis of HCM.¹ As per a new research letter published in the Journal of American College of Cardiology, researchers from the PIONEER-OLE (Extension Study of Mavacamten [MYK-461] in Adults With Symptomatic Obstructive Hypertrophic Cardiomyopathy Previously Enrolled in PIONEER) report AI-based ECG can also help in monitoring disease-related physiological and hemodynamic parameters.²

The researchers developed two AI-based-ECG algorithms independently at the University of California San Francisco (UCSF) and Mayo Clinic, and it was validated for HCM diagnosis during the pre-treatment and on-treatment ECGs of phase-2 PIONEER-OLE trial. The researchers developed AI-ECG-predicted HCM scores with the combination of echocardiographic and laboratory parameters at day 0, at weeks 4, 8, and every 12 weeks after that. A total of 216 patients (mean age 57.8 years, 69.2% men) were enrolled with a median follow-up of 79 weeks.

Longitudinal follow-up of the patients revealed a consistent decrease in the mean AI-ECG-predicted HCM scores during the treatment across each data point, with a mean HCM score reduction of 43% (0.67 pre-treatment to 0.38 at 72 weeks) for the UCSF algorithm. Similarly, the Mayo algorithm showed a 56% reduction (0.85 pre-treatment to 0.37 at 72 weeks). There was a longitudinal trend in the reduction of HCM scores to that of reduction in the echocardiographic parameters of left ventricular tract obstruction and laboratory parameters of NT-pro-BNP.²

The results of this study are intriguing, as this is the first study validating the use of a simple and readily available ECG paired with AI to be helpful in the diagnosis of the HCM and revealed to monitor the longitudinal trend in the response of the treatment. Furthermore, the results suggested that HCM scores could identify the changes in the ECG not seen by naked eyes interpretation of the ECG.

The novel findings of this research may lay a framework for further studies to diagnose and assess the treatment response in patients with other cardiovascular diseases. While the ECG continues to remain a powerful tool for cardiovascular assessment, ECG coupled with artificial intelligence may further help in earlier and easier diagnosis of complex cardiac conditions along with longitudinal follow-up regarding the effect of medications. Furthermore, it will be even more interesting to see similar AI-based predictive algorithms in the screening of life-threatening cardiac condition in future.