Medical-AI trial: intravascular-ultrasound (IVUS)-based AI to improve outcomes after coronary stenting (INNOVATE-PCI) (NCT05807841)

1Stenting and the precision of IVUS

In angina and heart attack, where the coronary arteries feeding the heart narrow, catheter treatment that widens the narrowing and places a stent (PCI, percutaneous coronary intervention) is widely performed.

During it, beyond angiography (a silhouette of the vessel), intravascular ultrasound (IVUS), which views the inside of the vessel in cross-section from the catheter tip, allows precise assessment of plaque character, degree of narrowing, and how well the stent is apposed. But reading IVUS images takes expertise and is sometimes underused.

2Analyzing angiography and IVUS images

This study validates, in real practice, the power of ML models that analyze these angiography and IVUS images, as a prospective, multicenter observational study. Per the registry summary, it validates the diagnostic performance and clinical impact of models developed by ML based on coronary angiography and IVUS in real-world practice.

The primary outcomes are culprit-related target-vessel failure (TVF), related to the treated lesion, and nonculprit-related TVF, related to untreated lesions. TVF is a composite of adverse cardiac outcomes tied to the vessel (such as restenosis, repeat treatment, or heart attack). The size is 3,000, and it is recruiting.

3The lesion treated, and the lesion not treated

The primary endpoints here divide in two. The same measure, target vessel failure, means different things depending on which lesion it concerns.

Target vessel failure gathers the adverse cardiac outcomes relating to a treated vessel — restenosis, reintervention, myocardial infarction. IVUS is precise but takes skill to read, and turning that information into better outcomes takes experience. Across 3,000 cases, a prospective multicentre study in routine care aims to capture the clinical impact that constrained trials do not show.