New Study Demonstrates AI Model Can Detect Heart Attack From a Single Pulse Waveform

H
hmriadmin

PASADENA, Calif. — October 3, 2023 — A new study led by Dr. Robert A. Kloner, Dr. Wangde Dai, and Dr. Rashid Alavi of Huntington Medical Research Institutes (HMRI) presents a breakthrough in using machine learning and cardiovascular physics to detect acute myocardial infarction (heart attack) from a single carotid pressure waveform — without relying on chest pain symptoms or an electrocardiogram (ECG).

Published in European Heart Journal Open, the paper — “Impact of Symptom-to-Reperfusion-Time on Transmural Infarct Extent and Left Ventricular Strain in Patients With ST-Segment Elevation Myocardial Infarction: A 3-Dimensional View on the Wavefront Phenomenon” — introduces a hybrid physics-based machine learning (ML) method that accurately identifies ischemia and heart attack in preclinical models.

The research team — Rashid Alavi, Wangde Dai, Ray V. Matthews, Robert A. Kloner, and Niema M. Pahlevan — developed ML classifiers trained on intrinsic frequency (IF) parameters extracted from carotid artery pressure waveforms in a rat model of ischemia and reperfusion.

Key findings include:

  • The best ML model achieved 92% sensitivity and 92% specificity for detecting acute myocardial infarction in blind testing.

  • For ischemia detection, the same approach achieved 85% specificity and 92% sensitivity.

  • The method relies solely on a single, non-invasive waveform measurement, demonstrating the potential for portable diagnostic tools.

“This proof-of-concept model points toward a future where heart attacks might be detected in real time using only a pulse signal,” said Dr. Alavi, Research Scientist at HMRI. “It bridges cardiovascular physics and artificial intelligence to create new frontiers in rapid, accessible diagnosis.”

The study highlights HMRI’s growing focus on AI-driven cardiovascular innovation, led by interdisciplinary teams uniting bioengineering, clinical cardiology, and computational modeling.

Full article: European Heart Journal Open, October 2023