Dr. Özkan Karaca

Study Title:
Artificial Intelligence-Based Fully Automated Quantitative Coronary Angiography Versus Optical Coherence Tomography Guidance for Coronary Stent Implantation (FLASH)

Conference: TCT 2024
Link https://www.tctmd.com/slide/artificial-intelligence-based-fully-automated-quantitative-coronary-angiography-versus

Introduction:
Percutaneous coronary interventions (PCI) performed under imaging guidance have shown more favorable outcomes compared to angiography-guided PCI, with the 2024 European Society of Cardiology (ESC) Class 1A recommendations advocating for its use in cases involving the left main coronary artery, bifurcations, and long lesions. However, the clinical use of intracoronary imaging is limited by various practical, logistical, and economic barriers. The clinical benefits of intracoronary imaging remain unclear, especially in less complex lesions.

Objective:
The FLASH study aimed to demonstrate that AI-based quantitative coronary angiography (AI-QCA) assisted PCI is non-inferior to optical coherence tomography (OCT)-guided PCI regarding post-procedural outcomes (minimal stent area).

Methods:
This study focused on patients with coronary artery disease (CAD), and 3,138 patients were assessed for eligibility from 13 centers. 400 patients were randomly assigned to groups (October 2022 - February 2024). PCI was performed with AI-QCA guidance in 200 patients and with OCT guidance in another 200 patients. Stent selection and optimization were performed according to the protocol using AI-QCA or OCT guidance, and final OCT evaluations were conducted in both groups. The primary efficacy endpoint was the minimal stent area (mm²) measured by the final OCT, while the primary safety endpoint was major adverse cardiac events (MACE) occurring during the procedure. Additionally, the secondary major endpoint included death, myocardial infarction (MI), and repeat revascularization within 6 months. The stent size was selected to be up to 20% larger than the distal reference diameter, and the stent length was determined to adequately cover the proximal and distal reference segments based on the lesion length provided by AI-QCA. Once the operator deemed the AI-QCA-assisted PCI completed, a post-PCI OCT evaluation was performed to assess the primary endpoint.

Results:
In the AI-QCA group, one patient was excluded due to an uninterpretable image. In the OCT group, two patients did not undergo the final OCT evaluation, and two patients had uninterpretable images, leading to a total of four exclusions. The post-PCI minimal stent area (MSA) was 6.3 ± 2.2 mm² in the AI-QCA group and 6.2 ± 2.2 mm² in the OCT group, with an absolute difference of -0.16 mm² (95% Confidence Interval: -0.59 to 0.28); the p-value for non-inferiority was < 0.001, and for superiority, it was 0.48. In terms of overall stent expansion, The AI-QCA group showed 78.7%, while the OCT group had 79.2%, with a p-value of 0.78. Stent under expansion was observed in 50.8% of the AI-QCA group and 54.6% of the OCT group, with a p-value of 0.48. Dissection rates were recorded at 15.6% for the AI-QCA group and 12.8% for the OCT group, with a p-value of 0.42. Untreated reference segment disease was observed in 15.1% of the AI-QCA group and 13.3% of the OCT group, with a p-value of 0.61. Stent malapposition was found to be higher in the AI-QCA group at 13.6% compared to 5.6% in the OCT group, which was statistically significant (p = 0.007). The mean MSA in malapposed segments was measured at 8.4 ± 1.9 mm², with no malapposed segments having an MSA of < 5 mm². When evaluating clinical outcomes at six months, the mortality rate in the AI-QCA group was 0.5% (1/199), while no deaths occurred in the OCT group (p > 0.99). Any repeated revascularization was 0.5% (1) in both groups (p > 0.99), and target vessel revascularization was observed in the OCT group at 0.5% (1) while none occurred in the AI-QCA group, with p > 0.99.

Conclusions:
AI-QCA-assisted PCI demonstrated similar efficacy to OCT-guided PCI regarding post-PCI MSA, both overall and in both proximal and distal stent segments. No significant differences were identified in stent under expansion, dissection, or untreated reference segment disease; however, stent malapposition was more frequently observed in the AI-QCA group. Procedural complications and clinical events at six months were low and comparable between both groups. The AI-QCA technology can effectively bridge conventional angiography and imaging-guided PCI, making it suitable for routine clinical practice.

Comment:
In the FLASH study, AI-QCA-assisted PCI proved to be equivalent to OCT-guided PCI in terms of minimal stent area, with similar procedural safety and six-month outcomes in less complex coronary artery disease. This technology presents a significant alternative, especially in resource-limited settings or in less complex cases where intravascular imaging may provide limited benefit. However, larger-scale studies with long-term clinical outcomes are necessary to clarify AI-QCA's role in routine interventional practice.