Image-Based Cell Quality Assessment: Modeling of Cell Morphology and Quality for Clinical Cell Therapy - Computational Modeling in Tissue Engineering - page 214

(a)

0.25

Keratinocyte

0.2

0.15

0.1

Fibroblast

0.05

0

Chalier's class interval (Length [pixels])

(b) 0.35

0.3

Keratinocyte

0.25

0.2

0.15

Fibroblast

0.1

0.05

0

Chalier's class interval (Z- score [-])

(c)

0.35

0.3

Keratinocyte

0.25

0.2

0.15

Fibroblast

0.1

0.05

0

Chalier's class interval (Z score [-])

Fig. 10 Parameter distribution comparing keratinocytes and fibroblasts. a Distribution of length
between two types of cells. There is huge overlapping area in one parameter distribution.
b Distribution of Z-score from the discriminant analysis using 19 morphological features between
two types of cells. There is still overlapping area even after the discriminant modeling.
c Distribution of Z-score from the discriminant analysis using 19 morphological features using
''noise reduced data''. Noise objects were excluded from the discriminant analysis by eliminating
the group N_s and N_l objects from the data. Such noise data c reduction greatly reduces
overlapping area from cell distributions