Comparative diagnostic performance and stability of deep learning- and CFD-based CT-FFR across vessels, cardiac phases, and centers - Scorecard - MDSpire
Advertisement
Comparative diagnostic performance and stability of deep learning- and CFD-based CT-FFR across vessels, cardiac phases, and centers
Clinical Scorecard: Evaluation of Diagnostic Efficacy and Consistency of Deep Learning and CFD Approaches for CT-FFR Across Different Coronary Vessels, Cardiac Phases, and Clinical Settings
At a Glance
Category
Detail
Condition
Coronary artery disease (CAD)
Key Mechanisms
CT-derived fractional flow reserve (CT-FFR) using deep learning (DL) and computational fluid dynamics (CFD)
Target Population
Symptomatic patients suspected of CAD
Care Setting
Multi-center clinical evaluation
Key Highlights
DL and CFD CT-FFR showed high diagnostic performance with AUCs of 0.90 and 0.89, respectively.
Both methods demonstrated strong correlation with invasive FFR (rho = 0.71 for DL; rho = 0.68 for CFD).
Stable performance across different coronary vessels, cardiac phases, and clinical centers.
Comparable classification rates in gray-zone lesions (DL: 86.4%, CFD: 84.6%).
Guideline-Based Recommendations
Diagnosis
Use invasive FFR as the reference standard for identifying ischemia-producing lesions.
Management
Consider CT-FFR for functional assessment in patients with CAD.
Monitoring & Follow-up
Evaluate the stability of CT-FFR performance across different clinical settings.
Risks
Be aware of potential inaccuracies in the gray zone around FFR thresholds.
Patient & Prescribing Data
220 patients (277 vessels) from two clinical centers.
CT-FFR can help reduce unnecessary invasive procedures by providing functional assessment.
Clinical Best Practices
Utilize both DL and CFD CT-FFR methods for comprehensive evaluation of coronary lesions.
Ensure high-quality CCTA images are available for accurate CT-FFR calculations.
Consider patient-specific factors when interpreting CT-FFR results.
A single-center editorial described real-world integration of artificial intelligence–based coronary artery calcium scoring into routine cardiac CT workflow, with researchers reporting rapid report availability and high agreement with manual reference standards while emphasizing continued radiologist oversight.
