Image-Based Cell Quality Assessment: Modeling of Cell Morphology and Quality for Clinical Cell Therapy - Computational Modeling in Tissue Engineering - page 202

(2) Image processing

Raw image Binalization Noise reduction

(1) Image data
collection

Object analysis

etc.

Length Breadth Area Perimeter Circularity Hole area

All
individual
cells

(4) Modeling

Statistics of morphological features

Sample

Experimentally deifined cell quality

(3) Experimental data collection

Fig. 1 Schematic illustration of image-based cell quality assessments

Background
extraction

Background
subtraction

Object
recognition

Particle
reduction

Raw image

Binarization

Object
Recognition

Object
measurement

Object noise
reduction

Noise

Cells

Fig. 2 Schematic illustration of image processing for assessing morphological features

assays, teacher signals can be obtained. There are cases where the culturing
condition itself can be used as a teacher signal. For image processing, we cus-
tomized our original image processing filters combining image analysis software
and original programs in C and R languages (Fig. 2 ). Briefly, the raw image was
processed to have the minimum error compared to manual cell counts after
binarization. In this process, we applied an original combination of filters that were
optimized using 26 types of human cells, including tumor cell lines and primary
cells. From the objects extracted after the binarization, we measured 9-30 mor-
phological features based on the characteristics of cells, together with multi-
collinearity examinations and interviews with cell culture experts. Statistical