Machine Learning Approach to Automated Quality Identification of Human Induced Pluripotent Stem Cell Colony Images

Table 4

Results of structures 4–6 given in Figures 4–6 when different kernels were used. True positive rates (%) are given in parentheses and accuracy (%) can be found from the last column.

Kernel/class	Bad	Good	Semigood	ACC
		Structure 4
Linear	28 (68.3%)	52 (70.3%)	19 (32.8%)	57.2%
Polynomial	25 (61.0%)	45 (60.8%)	19 (32.8%)	51.4%
Polynomial	18 (43.9%)	42 (56.8%)	19 (32.8%)	45.7%
Polynomial	23 (56.1%)	35 (47.3%)	19 (32.8%)	44.5%
Polynomial	23 (56.1%)	28 (37.8%)	14 (24.1%)	37.6%
Polynomial	23 (56.1%)	24 (32.4%)	15 (25.9%)	35.8%
RBF	22 (53.7%)	48 (64.9%)	23 (39.7%)	53.8%
Structure 5
Linear	24 (58.5%)	59 (79.7%)	21 (36.2%)	60.1%
Polynomial	18 (43.9%)	44 (59.5%)	24 (41.4%)	49.7%
Polynomial	20 (48.8%)	46 (62.2%)	26 (44.8%)	53.2%
Polynomial	16 (39.0%)	41 (55.4%)	23 (39.7%)	46.2%
Polynomial	19 (46.3%)	37 (50.0%)	21 (36.2%)	44.5%
Polynomial	15 (36.6%)	38 (51.4%)	20 (34.5%)	42.2%
RBF	20 (48.8%)	52 (70.3%)	21 (36.2%)	53.8%
Structure 6
Linear	20 (48.8%)	38 (51.4%)	24 (41.4%)	47.4%
Polynomial	15 (36.6%)	38 (51.4%)	26 (44.8%)	45.7%
Polynomial	18 (43.9%)	34 (45.9%)	20 (34.5%)	41.6%
Polynomial	17 (41.5%)	37 (50.0%)	20 (34.5%)	42.8%
Polynomial	21 (51.2%)	28 (37.8%)	20 (34.5%)	39.9%
Polynomial	18 (43.9%)	22 (29.7%)	23 (39.7%)	36.4%
RBF	19 (46.3%)	43 (58.1%)	22 (37.9%)	48.6%