Segmentation of Sonicated Blood Cells and Ultrasound Contrast Agent Microbubbles from Time-Lapse Microscopic Images
No Thumbnail Available
Date
2018-02-17
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Blood cells and ultrasound contrast agent microbubbles behave strangely when subjected to
an ultrasound field. They migrate towards the nodes of the ultrasound wave and form circular
patterns. However, differentiation between blood cells and ultrasound contrast agent
microbubbles can be easily determined via time-lapse analysis of the pattern formation. Ultrasound
contrast agent microbubbles migrate towards the wave nodes very quickly and
form tightly packed clusters. In contrast, formation of tightly packed clusters in the case of
blood cells is unlikely based on the sonophore model theory of cells. Moreover, the interaction
of the microbubbles with the blood cells and the surrounding medium is not fully understood.
To study the behavior and interaction of sonicated blood cells and microbubbles
on the time-lapse microscopic images, there is a need to define a contour around the cells
and the microbubbles. To do so, first, the microbubbles and the cells should be segmented
from the rest of image content. In this regard, this thesis devised a scheme that combines features
of the Laplacian of the Gaussian detector and a modified form of the Contrast Limited
Adaptive Histogram Equalization (CLAHE) technique for effective analysis of time-lapse
microscopic images. The scheme is tested on three datasets (one synthetic and two real) and
its subjective and objective performance is found quite pleasing. Absence of ground truth
for the real datasets makes the evaluation of the segmentation scheme merely subjective.
Objective evaluation is only performed on the computer-simulated time-lapse images. The
average segmentation sensitivity, specificity and accuracy of the proposed algorithm are valued
around 0.96, 0.91 and 0.95 on the synthetic dataset out of a unit scale, respectively. The
result generated could be a crucial input for effective particle tracking and sizing studies.
Description
Keywords
Segmentation, Red Blood Cell, Ultrasound, Microbubbles, Image Processing, Contrast Agent, LoG, CLAHE