April 10, 2018Radiology
Coronary CT Angiography–derived Fractional Flow Reserve: Machine Learning Algorithm versus Computational Fluid Dynamics Modeling
Key result
Machine learning FFR matches FFR-CFD in detecting lesion-specific ischemia, both outperforming coronary CT angiography.
Why the study?
Does a machine learning algorithm for CT-derived FFR perform as well as computational fluid dynamics modeling in detecting lesion-specific ischemia?
Population
85 patients who had undergone coronary CT angiography followed by invasive FFR in a single-center…
Comparison
Fractional flow reserve derived from coronary CT… vs Fractional flow reserve derived from coronary CT…
Design
Cohort
Authors
Discussion
Loading...
Member takes
Implication
ML-based CT-FFR matches CFD accuracy with faster processing; leaves open prospective outcome validation before clinical adoption.
Structured PICO
Does a machine learning algorithm for CT-derived FFR perform as well as computational fluid dynamics modeling in detecting lesion-specific ischemia?
P
Population
85 patients (mean age 62 ± 11 years, 62% men) who had undergone coronary CT angiography followed by invasive FFR in a single-center retrospective study.
I
Intervention
Fractional flow reserve derived from coronary CT angiography based on a machine learning algorithm (FFR_ML)
C
Comparator
Fractional flow reserve derived from coronary CT angiography based on computational fluid dynamics (FFR_CFD), visual stenosis grading at coronary CT angiography, and quantitative coronary angiography (QCA), with invasive FFR as the reference standard
O
Outcome
Diagnostic performance (sensitivity, specificity, and area under the receiver operating characteristics curve) for detecting lesion-specific ischemiasurrogate
A machine learning-based approach for CT-derived FFR provides equivalent diagnostic accuracy to computational fluid dynamics modeling for detecting lesion-specific ischemia, but with significantly shorter processing times.
Cite This Study
Tesche et al. (2018) studied this question. The FFR derived from machine learning algorithm (FFR<sub>ML</sub>) showed similar performance to FFR<sub>CFD</sub> in detecting lesion-specific ischemia, both outperforming coronary CT angiography.