Digital Identification and Quantification of Vessels Surrounding Implants

1
Digital Identification and Quantification of
Vessels Surrounding Implants Hector
Mobine Department of Bioengineering, University
of California, San Diego. La Jolla, CA
Introduction Tissue based oxygen and
glucose sensors require analyte transport by
blood vessels. The implanted sensor can be
visualized using a window chamber preparation
implemented by our group. The chamber allows the
functional visualization of the implant area.
Visual studies are aimed at assessing the extent
of flow of the vessels surrounding the implant,
as well as studying the capillary distribution
around the area. A major part of the challenge in
developing a relationship between transport
mechanisms and sensor performance is the
quantification of capillary density in the
vicinity of the sensor. The information
obtained by evaluating vascular growth around an
implant allows the researcher to evaluate the
needs of a sensor in relation to its proximity to
vessels, and the sensors effect on vascular
growth. The algorithms developed included
studying and reconstructing the color-space
associated with the image in order to reduce
noise, shadows and lighting artifacts.
Segmentation algorithms were tested to study
optimal methods for producing capillary counts,
including thresholding, edge detection.
Methods Acquisition/Preprocessing
High resolution color images were acquired using
a (1200x1600) digital camera. A 2D Wiener filter
was used to remove noise.
Border Location The most significant obstacle
to quantification was the identification and
mapping of small vessels. A simple thresholding
algorithm was applied to define the vessel path.
Edge Detection A Sobel edge detection
algorithm provided the best distinction between
the vasculature and surrounding tissue.
Segmentation Simple neighborhood based
operations evaluate the neighborhood of the pixel
in question to adjust the threshold scheme.
Enhancement The image quality was enhanced by
applying histogram equalization to increase image
contrast.
Figure 1 Process Flow
Chart

• Quantification
• Once the final vascular image was formed, the
  image was skeletonized and the lengths of the
  vessels were counted.
• figure 2
  figure 3
• original image
  thresholded image
• Results
• As evident in figure 3, a digital
  identification of the vasculature was obtained
  successfully. The resulting vessel count lengths
  obtained through the program proved to be
  consistently within 20 of human visual tracings.
  Although the procedure was successful the
  question remains as to whether the imaging
  captures vessels unseen by the human eye or is
  just detecting noise. Improvements of the
  algorithm are essential for the program to better
  identify vessels especially in poorly lit images.
  
• Conclusion
• Digital vessel identification and
  quantification is possible with adequate
  filtering and edge detection methods.
• Vessel quantification is accurate and can be
  used to map vessel growth.