Co-registered optical coherence tomography and X-ray angiography for the prediction of fractional flow reserve

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Cardiovascular disease (CVD) is the leading cause of global mortality, with coronary artery disease (CAD) accounting for 42% of deaths.
• Anatomical assessment of CAD has limitations, leading to the adoption of Fractional Flow Reserve (FFR) as a key functional diagnostic metric.
• Invasive FFR measurement is the gold standard but carries procedural risks and costs.

Purpose of the Study:

• To evaluate the efficacy of an ensemble deep neural network (DNN) model in predicting invasively measured FFR.
• To utilize raw anatomical data from optical coherence tomography (OCT) and X-ray coronary angiography (XA) for FFR prediction.
• To assess the model's performance on a challenging dataset with a high prevalence of borderline lesions.

Main Methods:

• An ensemble deep neural network (DNN) model was developed to predict FFR.
• The model was trained using co-registered anatomical data derived from OCT and XA.
• A dataset with 46% of lesions in the challenging FFR range of 0.75 to 0.85 was used for evaluation.

Main Results:

• The DNN model achieved an overall accuracy of 84.3% in predicting FFR.
• The model demonstrated a sensitivity of 87.5% and a specificity of 81.4%.
• Integrating both OCT and XA data significantly improved the model's predictive accuracy compared to using single modalities.

Conclusions:

• An ensemble DNN approach integrating OCT and XA data can accurately predict invasive FFR.
• This non-invasive method shows promise for improving the functional assessment of coronary artery disease.
• The findings suggest a potential for AI-driven analysis to enhance CAD diagnosis and management.