Machine Learning for Diagnosis of Occlusive MI in LBBB Patients

N/ARecruitingNCT07620119

Konya City Hospital50 enrolled

Overview

This study investigates a new way to diagnose severe heart attacks in patients who have a specific electrical heart pattern called a Left Bundle Branch Block (LBBB). When patients present to the emergency department with chest pain, doctors routinely perform an electrocardiogram (ECG) to check for a heart attack. However, the presence of an LBBB can alter the heart's electrical signals on the ECG, effectively masking or hiding the typical signs of an ongoing acute coronary occlusion (a completely blocked artery). This making it highly challenging for emergency physicians to make an accurate and rapid diagnosis. The primary purpose of this prospective and observational research is to develop and evaluate an artificial intelligence/machine learning (ML) model that can analyze digital 12-lead ECG signals to accurately predict a true blocked coronary artery in patients with LBBB. The machine learning model will analyze raw digital ECG waveforms to detect subtle, microscopic patterns that might be missed by the human eye. To confirm the accuracy of the model, its predictions will be compared directly with invasive coronary angiography results, which is the gold standard reference method used to visualize blocked vessels. Additionally, the study aims to evaluate if the model can differentiate between a true heart attack caused by a blocked artery (Type 1 MI) and other non-occlusive conditions that cause elevated heart enzymes (Type 2 MI). Ultimately, the investigators intend to determine whether integrating this machine learning tool into emergency care can safely reduce the rate of unnecessary emergency invasive procedures for patients who do not have a true coronary blockage.

Study Type

OBSERVATIONAL

Enrollment

Conditions

Acute Myocardial Infarction (AMI)Left Bundle Branch Block

Interventions

Digital 12-Lead ECG Analysis and Invasive Coronary AngiographyOTHER

Standard 12-lead digital electrocardiogram (ECG) data recorded during the emergency department index visit will be analyzed using a developed machine learning model. The model's predictions will be compared against the results of standard invasive coronary angiography (the gold standard reference method) performed as part of routine clinical care.

Eligibility

Sex: ALLMin age: 18 Years

Medical Language ↔ Plain English

Inclusion Criteria: * Patients aged 18 years and older who present to the emergency department. Patients presenting with acute ischemic chest pain or clinical ischemia-equivalent symptoms (such as acute dyspnea, unexplained diaphoresis, or syncope). Patients with a confirmed Left Bundle Branch Block (LBBB) on their initial 12-lead electrocardiogram (ECG), which can be either newly developed or known/chronic. Patients who undergo invasive coronary angiography during their index hospital admission. Patients or their legally authorized representatives who provide written informed consent to participate in the study. Exclusion Criteria: * Patients under the age of 18. Pregnant or lactating women. Patients with poor-quality or uninterpretable digital ECG recordings due to severe artifact, missing leads, or technical errors. Patients who develop cardiopulmonary arrest before an initial diagnostic 12-lead ECG can be obtained in the emergency department. Patients transferred from another healthcare facility who have already undergone coronary angiography or revascularization. Patients who decline to participate or refuse to provide written informed consent.

Outcomes

Primary Outcomes

Diagnostic Performance for Occlusive Acute Myocardial Infarction

Evaluation of the developed machine learning model's diagnostic performance in predicting angiographically proven acute coronary occlusion (defined as TIMI 0-1 flow or equivalent true occlusion during catheterization). The primary metrics to evaluate this outcome will include the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC), Sensitivity, Specificity, Positive Predictive Value (PPV), and Negative Predictive Value (NPV).

Time frame: Within the emergency department index visit (typically within 24 hours of presentation).

Secondary Outcomes

Title: Differentiation Performance Between Type 1 MI and Type 2 MI

Evaluation of the machine learning model's performance (measured by AUC, sensitivity, and specificity) to distinguish between acute coronary occlusion (Type 1 MI) and non-occlusive ischemic myocardial injury or supply-demand mismatch presenting with elevated cardiac troponin (Type 2 MI).

Time frame: Within the hospital stay (up to 7 days).

Projected Reduction Rate of Unnecessary Angiographies

Simulation and post-hoc analysis to quantify the potential relative reduction in the rate of emergency invasive coronary angiographies among LBBB patients without true coronary occlusion by applying the model's diagnostic probability scores.

Time frame: Calculated at the study completion