Development of machine learning models for fractional flow reserve prediction in angiographically intermediate coronary lesions

Marco Lombardi ¹, Rocco Vergallo ^2,3, Andrea Costantino ⁴

¹Department of Cardiovascular Sciences, Fondazione Policlinico Universitario A. Gemelli IRCCS, Università Cattolica Sacro Cuore, Rome, Italy.
²Department of Internal Medicine and Medical Specialties (DIMI), Università di Genova, Genova, Italy.
³Interventional Cardiology Unit, Cardiothoracic and Vascular Department (DICATOV), IRCCS Ospedale Policlinico San Martino, Genova, Italy.
⁴Department of Biomedical Sciences, Humanitas University, Milan, Italy.
⁵Department of Cardiovascular Medicine, Tsuchiura Kyodo General Hospital, Tsuchiura, Japan.
⁶Department of Cardiology, Central Hospital of Internal Affairs and Administration Ministry, Postgraduate Medical Education Centre, Warsaw, Poland.
⁷Department of Cardiovascular Sciences, Policlinico Umberto I, Sapienza University of Rome, Rome, Italy.
⁸HartCentrum, Ziekenhuis Netwerk Antwerpen (ZNA) Middelheim, Antwerp, Belgium.
⁹Royal Brompton Hospital, London, UK.
¹⁰Oxford Heart Centre, John Radcliffe Hospital, Oxford University Hospitals, NHS Foundation Trust, Oxford, UK.
¹¹Department of Cardiology, Hospital Wroclaw, Wroclaw, Poland.
¹²Department of Cardiology, AZ West Hospital, Veurne, Belgium.

⁰Department of Cardiovascular Sciences, Fondazione Policlinico Universitario A. Gemelli IRCCS, Università Cattolica Sacro Cuore, Rome, Italy.

Catheterization and Cardiovascular Interventions : Official Journal of the Society for Cardiac Angiography & Interventions +

|

August 2, 2024

View abstract on PubMed

Summary

Machine learning models using clinical, angiographic, and optical coherence tomography (OCT) data can predict fractional flow reserve (FFR) in intermediate coronary lesions. These algorithms show moderate accuracy in identifying patients needing revascularization.

Area Of Science

Cardiovascular medicine
Medical imaging
Artificial intelligence in healthcare

Background

Fractional flow reserve (FFR) is the standard for guiding coronary revascularization decisions for angiographically intermediate coronary lesions (AICL).
Optical coherence tomography (OCT) provides detailed plaque morphology and lumen dimension data.
Predicting FFR non-invasively remains a clinical challenge.

Purpose Of The Study

To develop and validate machine learning (ML) models for predicting FFR.
To utilize clinical, angiographic, and OCT variables for FFR prediction.
To assess the performance of ML models in identifying significant AICL.

Main Methods

A pooled analysis of individual patient data from multicenter studies was performed.
Six supervised ML models were trained (n=351) and validated (n=151) using 25 variables.
Models were evaluated based on Area Under the Curve (AUC) and F1 score.

Main Results

ML models achieved AUCs ranging from 0.71 to 0.78 and F1 scores from 0.70 to 0.75.
Moderate sensitivity (0.68-0.77) and specificity (0.59-0.69) were observed.
Sensitivity analysis with a 0.75 FFR cut-off showed improved AUC (0.78-0.86) and specificity (0.71-0.84).

Conclusions

ML algorithms integrating clinical, angiographic, and OCT data can effectively predict FFR.
These models offer a potential non-invasive tool for managing AICL.
Further validation is warranted for clinical implementation.

Keywords:

artificial intelligence fractional flow reserve intermediate coronary lesions machine learning optical coherence tomography