Electrocardiographic Detection of Non-ST Elevation Myocardial Events for Accelerated Classification of Chest Pain Encounters (ECG-SMART 2)

Electrocardiographic Detection of Non-ST Elevation Myocardial Events for Accelerated
Classification of Chest Pain Encounters (ECG-SMART-2)
ABSTRACT
There is a clear need to develop improved tools to stratify risk in patients who seek emergency care for chest
pain, one of the most common and potentially deadly conditions encountered in acute care settings. The 12-
lead ECG has been the mainstay of initial evaluation of chest pain yet is currently only diagnostic for a small
subset of patients with ST-elevation myocardial infarction. Over the past funding period, we have built the
largest database of multi-hospital, outcome-linked, prehospital 12-lead ECG repository known to us (n=4,132).
Using this multi-expert, multi-tier ground truth annotated database, we have developed and validated novel,
machine learning-based, ECG interpretation algorithms that could identify non-ST elevation acute coronary
events. Using state-of-the-art interpretability toolkits, we identified ECG signatures that are mechanistically
linked to ischemia and can serve as plausible markers of acute coronary syndrome. We now aim to move
these extensive efforts to clinical use by expanding and building these models at the bedside for prospective
validation and real-time clinical deployment. The specific aims of this renewal application are: 1) to build and
externally validate a multi-task, ECG-based intelligent decision support system; 2) to build and deploy a real-
time architecture for this intelligent system along with a clinician-facing graphical user interface platform; and 3)
to perform a prospective clinical validation of this intelligent ECG system, including silent deployment and
evaluation at two clinical sites. The final deliverable is an intelligent ECG interpretation system for detecting
and stratifying patients with suspected acute coronary syndrome of sufficient readiness to be deployed in
clinical trials aimed at improving outcomes in non-ST elevation coronary syndromes. Such intelligent system,
when combined with the judgment of trained emergency personnel (physicians, nurses, and paramedics),
would more accurately identify patients with acute coronary occlusions for ultra-early intervention. This system
will streamline the care provided to non-specific chest pain beyond the costly and time-consuming overnight
observations for serial cardiac enzymes and provocative testing.

Grant Number: 3R01HL137761-08S1
NIH Institute/Center: NIH
Principal Investigator: Salah Al-Zaiti