Biomedical Engineering Reference
In-Depth Information
greater therapeutic effects. Furthermore, from the clinical doctors' perspective,
proper scheduling of surgery is essential for treating more patients in a given
facility.
To satisfy all the above-listed criteria, ''image-based cell quality assessment''
offers great potential for quality assurance in cell therapy. Image-based cell quality
assessment enables non-invasive, fully exhaustive, timely, and predictive evalu-
ation of cells.
Technologies that evaluate and assess cellular activities using cell image
measurement methods are being reported [
3
,
4
,
6
,
8
-
11
,
13
,
16
-
24
,
26
]. Takagi has
widely reviewed especially the non-invasive cell-imaging technologies, and has
introduced novel technologies. The reviews strongly indicate that cellular mor-
phologies significantly correlate with cellular activities. These findings underscore
the importance of cellular morphology monitoring in traditional cell culture
methods. Many textbooks have indicated that cellular morphology is an indicator
of cell quality, and therefore, it should be carefully monitored. In many facilities
that offer clinical tissue engineering therapy, cells are maintained and controlled
during the culture process by the experts' experienced ''visual expertise.'' Con-
sidering the successes of image-based cell assessments and the strong require-
ments in clinical cell therapy, analysis of datasets of morphological features and
biological phenomena are attractive for machine learning researchers. In recent
years, machine learning algorithms that have been widely applied in the field of
bioinformatics and systems biology (gene analysis, mRNA profile analysis, protein
data analysis, etc.) have been effectively used in image-based cellular analysis
studies [
1
,
5
,
15
]. However, as far as we recognize, studies using non-labeled
cellular images are still limited. Furthermore, applications of machine learning
algorithms to assist with cell quality assessments, especially focusing on the
requirements of cell therapy, are rare.
In this topic chapter, we present some of our successful modeling results that
support the effectiveness of machine learning applications in clinical tissue engi-
neering and cell therapy.
2 Strategy of Image-Based Cell Assessment Model Construction
The construction of a model for image-based cell quality assessment comprises 4
major steps: (1) Image data collection, (2) Image processing, (3) Experimental
data collection, and (4) Modeling (Fig.
1
).
In our studies, image data comprise phase contrast microscopic images, because
they represent the type of images most frequently used by cell biologists therefore
considered to contain historically-proven indicative information. For image data
collection, we use BioStation CT (Nikon Corporation, Tokyo, Japan), the fully
automated cell incubation and monitoring system for stable, periodical, and mass
time-lapse image data. For experimental data collection, we assayed the ''observed
cells'' using conventional cell biological techniques. From carefully selected
Search WWH ::
Custom Search