Rapid virtual fractional flow reserve using 3D computational fluid dynamics.

Area of Science:

• Cardiovascular Imaging and Interventions
• Computational Fluid Dynamics
• Medical Technology

Background:

• Virtual Fractional Flow Reserve (vFFR) enhances Fractional Flow Reserve (FFR) assessment for guiding coronary interventions.
• Current 3D-computational fluid dynamics (CFD) methods for vFFR, while accurate, are computationally intensive and time-consuming.

Purpose of the Study:

• To evaluate a novel 3D-CFD software method utilizing graphics processing unit (GPU) computation for vFFR.
• To compare the speed, accuracy, and cost-effectiveness of the GPU-based method against the fastest existing central processing unit (CPU)-based 3D-CFD technique.

Main Methods:

• Investigated a novel GPU-based 3D-CFD software for vFFR computation.
• Compared the GPU method with a CPU-based 3D-CFD technique using 40 angiographic cases.
• Assessed metrics including computation time, accuracy, hardware cost, and energy consumption.

Main Results:

• The GPU simulation was significantly faster, achieving computation in a median of 31.7 seconds compared to 607.5 seconds for the CPU method (P < 0.0001).
• The novel GPU technique demonstrated high accuracy, achieving 99.6% relative accuracy compared to the CPU method.
• While initial hardware costs were higher for GPU (£4080 vs. £2876), the median energy consumption per case was substantially lower (8.44 Wh vs. 2.60 Wh, P < 0.0001).

Conclusions:

• The novel GPU-based 3D-CFD method offers up to a 28-fold acceleration in vFFR computation.
• This accelerated vFFR calculation achieves this speed improvement without a clinically significant compromise in accuracy.
• The findings suggest a more efficient and potentially cost-effective approach to vFFR analysis in clinical practice.