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

Table 6

Results of -NN with different weighting and measure combinations. True positive rates (%) are given in parentheses and accuracy (%) can be found from the last column.

Measure and weighting combination/class	Bad	Good	Semigood	ACC
Chebyshev/equal weights	18 (43.9%)	52 (70.3%)	26 (44.8%)	55.5%
Chebyshev/inverse weights	19 (46.3%)	53 (71.6%)	27 (46.6%)	57.2%
Chebyshev/squared inverse weights	17 (41.5%)	53 (71.6%)	27 (46.6%)	56.1%
Cityblock/equal weights	22 (53.7%)	55 (74.3%)	24 (41.4%)	58.4%
Cityblock/inverse weights	23 (56.1%)	52 (70.3%)	27 (46.6%)	59.0%
Cityblock/squared inverse weights	20 (48.8%)	51 (68.9%)	22 (37.9%)	53.8%
Correlation/equal weights	17 (41.5%)	59 (79.7%)	16 (27.6%)	53.2%
Correlation/inverse weights	16 (39.0%)	54 (73.0%)	14 (24.1%)	48.6%
Correlation/squared inverse weights	22 (53.7%)	53 (71.6%)	17 (29.3%)	53.2%
Cosine/equal weights	19 (46.3%)	57 (77.0%)	14 (24.1%)	52.0%
Cosine/inverse weights	19 (46.3%)	62 (83.8%)	15 (25.9%)	55.5%
Cosine/squared inverse weights	21 (51.2%)	50 (67.6%)	14 (24.1%)	49.1%
Euclidean/equal weights	24 (58.5%)	55 (74.3%)	29 (50.0%)	62.4%
Euclidean/inverse weights	25 (61.0%)	53 (71.6%)	27 (46.6%)	60.7%
Euclidean/squared inverse weights	20 (48.8%)	46 (62.2%)	26 (44.8%)	53.2%
Standardized Euclidean/equal weights	22 (53.7%)	54 (73.0%)	26 (44.8%)	59.0%
Standardized Euclidean/inverse weights	25 (61.0%)	53 (71.6%)	27 (46.6%)	60.7%
Standardized Euclidean/squared inverse weights	20 (48.8%)	46 (62.2%)	26 (44.8%)	53.2%
Spearman/equal weights	15 (36.6%)	59 (79.7%)	19 (32.8%)	53.8%
Spearman/inverse weights	17 (41.5%)	61 (82.4%)	19 (32.8%)	56.1%
Spearman/squared inverse weights	17 (41.5%)	56 (75.7%)	17 (29.3%)	52.0%