Qualitatively, the 2 approaches showed agreement in the bulk flow patterns. However, the velocity computed with CFD was outside the DPIV uncertainty estimates, which can be attributed to the intrinsic difference and modeling assumptions of the 2 approaches. The findings reveal that wider bifurcation angles and double-stenting procedures are both characterized by increased areas of low flow and recirculation. Additionally, inferior performance in terms of viscous and wall shear stresses was observed in double-stented cases.

We present advanced particle image velocimetry (PIV) processing, post-processing, and uncertainty estimation techniques to support the validation of computational fluid dynamics analyses of medical devices. This work is an extension of a previous FDA-sponsored multi-laboratory study, which used a medical device mimicking geometry referred to as the FDA benchmark nozzle model. Experimental measurements were performed using time-resolved PIV at five overlapping regions of the model for Reynolds numbers in the nozzle throat of 500, 2000, 5000, and 8000. Images included a twofold increase in spatial resolution in comparison to the previous study. Data was processed using ensemble correlation, dynamic range enhancement, and phase correlations to increase signal-to-noise ratios and measurement accuracy, and to resolve flow regions with large velocity ranges and gradients, which is typical of many blood-contacting medical devices. Parameters relevant to device safety, including shear stress at the wall and in bulk flow, were computed using radial basis functions. In addition, in-field spatially resolved pressure distributions, Reynolds stresses, and energy dissipation rates were computed from PIV measurements. Velocity measurement uncertainty was estimated directly from the PIV correlation plane, and uncertainty analysis for wall shear stress at each measurement location was performed using a Monte Carlo model. Local velocity uncertainty varied greatly and depended largely on local conditions such as particle seeding, velocity gradients, and particle displacements. Uncertainty in low velocity regions in the sudden expansion section of the nozzle was greatly reduced by over an order of magnitude when dynamic range enhancement was applied. Wall shear stress uncertainty was dominated by uncertainty contributions from velocity estimations, which were shown to account for 90-99% of the total uncertainty. This study provides advancements in the PIV processing methodologies over the previous work through increased PIV image resolution, use of robust image processing algorithms for near-wall velocity measurements and wall shear stress calculations, and uncertainty analyses for both velocity and wall shear stress measurements. The velocity and shear stress analysis, with spatially distributed uncertainty estimates, highlights the challenges of flow quantification in medical devices and provides potential methods to overcome such challenges.

Micro-Particle Image Velocimetry (lPIV) measurements often suffer from poor image quality because of volume illumination effects, out of focus particles, and low seeding densities. As a result, measurements are typically ensemble averaged in time to improve the signal-to-noise ratio (SNR) of the resulting cross correlations. To achieve reliable, time-accurate lPIV measurements we need to improve the SNR of the recorded images and/or the SNR treatment of the resulting cross correlations. In this paper, we improve image quality and cross correlation SNR by comparing the use of confocal microscopy with spectral filtering. Steady-state spatiotemporally resolved data from widefield and confocal lPIV experiments were used and cross correlations were performed using standard techniques and the Robust Phase Correlation (RPC) method that employs a PIV-optimized spectral filter on the cross-correlation planes. The accuracy improvements were assessed by comparison against the time-averaged ensemble cross correlation, which currently represents the most accurate and accepted approach for steady-state lPIV measurements. Results show 24.77 % erroneous vectors for two-pass standard cross correlation with widefield imaging, which was reduced to 9.08 % erroneous vectors when using the RPC and confocal imaging. Furthermore, a 59.2 % reduction of error referenced to the ensemble correlation was observed when using RPC with confocal imaging over baseline cases. Improvements seen for RPC and confocal cases result from synergistically improving the correlation signal-to-noise ratio, resulting in correlation planes with sharper primary peaks and lower background levels.

No abstract

Particle Image Velocimetry (PIV) is currently the most widely used and well-established tool for non-invasive flow field velocity measurements, and a valuable method for validating computational fluid dynamics (CFD) models of medical devices. One of the critical steps in the CFD validation process is quantification of the experimental uncertainties. This work utilizes a new uncertainty estimation methodology developed by Charonko et al.1 for quantifying the PIV cross-correlation uncertainties. Uncertainties from experimental sources, including image magnification and acquisition timing, were propagated using Taylor series expansion for PIV data within the FDA benchmark Nozzle model2.