Flow measurements in a blood-perfused collagen vessel using x-ray micro-particle image velocimetry

Abstract

Blood-perfused tissue models are joining the emerging field of tumor engineering because they provide new avenues for modulation of the tumor microenvironment and preclinical evaluation of the therapeutic potential of new treatments. The characterization of fluid flow parameters in such in-vitro perfused tissue models is a critical step towards better understanding and manipulating the tumor microenvironment. However, traditional optical flow measurement methods are inapplicable because of the opacity of blood and the thickness of the tissue sample. In order to overcome the limitations of optical method we demonstrate the feasibility of using phase-contrast x-ray imaging to perform microscale particle image velocimetry (PIV) measurements of flow in blood perfused hydrated tissue-representative microvessels. However, phase contrast x-ray images significantly depart from the traditional PIV image paradigm, as they have high intensity background, very low signal-to-noise ratio, and volume integration effects. Hence, in order to achieve accurate measurements special attention must be paid to the image processing and PIV cross-correlation methodologies. Therefore we develop and demonstrate a methodology that incorporates image preprocessing as well as advanced PIV cross-correlation methods to result in measured velocities within experimental uncertainty.

Figure 1. Diagram of blood-perfused collagen vessel bioreactor.

Figure 2. Schematic of x-ray PIV in collagen vessel bioreactor.

Figure 3. Sample images of blood flow in the collagen microchannel.

A: Raw image after reorientation and cropping to vessel diameter. B: Processed image (sliding mean background subtraction, high-pass filter, and mask). Image B has been brightness- and contrast-enhanced for visualization.

Figure 4. Effect of image preprocessing on PIV accuracy.

Mean velocity profiles obtained using standard PIV correlation applied to raw images (dashed blue lines) and processed images (solid black lines) are plotted with the volume-adjusted Poiseuille solution (dashed red lines) for each experimental case. The uncertainty based on measured flow rate and vessel diameter is plotted for the theoretical solution (upper and lower dashed red lines). The 95% confidence interval on the mean for the experimental data is not distinguishable from the mean. Left to right: glycerine-perfused PTFE, blood-perfused FEP, and blood-perfused collagen. Data shown is averaged in time and along the length of the vessel and is normalized for each experimental case by the vessel diameter and theoretical maximum velocity.

Figure 5. Comparison of image signal-to-noise ratio before and after image preprocessing.

The mean SNR of all images (>2000) is plotted for each data set.

Figure 6. Effect of PIV correlation method on PIV accuracy.

Mean velocity profiles obtained using standard cross-correlation (solid black lines) and robust phase correlation with 2.8 pixel diameter (dashed green lines) applied to processed images are plotted with the volume-adjusted Poiseuille solution (dashed red lines) for each experimental case. The uncertainty based on measured flow rate and vessel diameter is plotted for the theoretical solution (upper and lower dashed red lines). The 95% confidence interval on the mean for the experimental data is not distinguishable from the mean. Left to right: glycerine-perfused PTFE, blood-perfused FEP, and blood-perfused collagen. Data shown is averaged in time and along the length of the vessel and is normalized for each experimental case by the vessel diameter and theoretical maximum velocity.

Figure 7. Contours of SCC and RPC correlation planes.

The contour indicates relative correlation magnitude. Only the central 50% window is shown for RPC correlation planes.

Figure 8. Effect of sum-of-correlation on PIV accuracy.

Mean velocity profiles obtained using instantaneous (dashed green lines) and sum-of-correlation (solid blue lines) RPC applied to processed images are plotted with the volume-adjusted Poiseuille solution (dashed red lines) for each experimental case. The uncertainty based on measured flow rate and vessel diameter is plotted for the theoretical solution (upper and lower dashed red lines). The 95% confidence interval on the mean for the experimental data is not distinguishable from the mean. Left to right: glycerine-perfused PTFE, blood-perfused FEP, and blood-perfused collagen. Data shown is averaged in time for the instantaneous sets and along the length of the vessel for all sets and is normalized for each experimental case by the vessel diameter and theoretical maximum velocity.

Figure 9. Contours of RPC instantaneous and sum-of-correlation planes.

The contour indicates relative correlation magnitude. Only the central 50% window is shown.