Acute Coronary Occlusion in NSTEMI Patients: Prevalence, Clinical Characteristics and the Potential Role of Artificial Intelligence.

Background and

Objectives

The electrocardiogram (ECG)-based STEMI/NSTEMI classification determines the urgency of invasive management in acute myocardial infarction. However, it often underestimates the presence of acute coronary occlusion (ACO) in patients presenting with non-ST-elevation myocardial infarction (NSTEMI). Artificial intelligence (AI)-assisted ECG interpretation has emerged as a potential tool to improve early recognition of ACO. This study aimed to determine the prevalence of ACO among NSTEMI patients, to compare clinical characteristics between patients with and without ACO and to explore the potential role of AI in earlier recognition of ACO. Materials and

Methods

All consecutive NSTEMI patients undergoing coronary angiography between September 2022 and December 2024 were included. Contrary to other studies that included TIMI flow grades 0-1, 0-2, or 0-3, ACO in our study was defined strictly as a culprit lesion with TIMI flow grade 0 at index coronary angiography. Clinical characteristics were compared between ACO and non-ACO patients. Admission 12-lead ECGs from ACO patients were retrospectively analysed using a clinically validated AI-based ECG interpretation model and classified according to the urgency of invasive management.

Results

Among 520 NSTEMI patients, 49 (9.4%) had angiographically confirmed ACO. Within the non-ACO group, 7.0% of patients had TIMI flow grade 1 on index coronary angiography (6.3% of the total population). Therefore, 15.7% of the study population had TIMI flow grade 0/1. ACO patients were younger (60.9 ± 12.8 vs. 66.3 ± 12.0 years, p = 0.0065). Clinical characteristics did not differ between the groups, except for dyslipidemia, which was more prevalent in non-ACO patients (38.8% vs. 53.9%, p = 0.043). Revascularisation rates were higher in the ACO group (93.9% vs. 82.2%, p = 0.037). Culprit vessel distribution differed markedly between the groups (p < 0.0001). In multivariable logistic regression analysis, age was independently associated with ACO (OR 0.96, 95% CI 0.93-0.99, p = 0.007). AI-assisted ECG analysis was performed in 42 ACO patients; 57.1% were classified as requiring immediate invasive management.

Conclusions

A significant proportion of NSTEMI patients have ACO. AI-assisted ECG interpretation may support earlier identification of ACO, although its clinical impact requires further validation. Future studies are warranted to confirm these findings.