Computational and Mathematical Methods in Medicine / 2016 / Article / Tab 3 / Research Article
Machine Learning Approach to Automated Quality Identification of Human Induced Pluripotent Stem Cell Colony Images Table 3 Results of structures 1ā3 given in Figures
1 ā
3 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 1 Linear 29 (70.7%) 52 (70.3%) 24 (41.4%) 60.7% Polynomial 21 (51.2%) 42 (56.8%) 19 (32.8%) 47.4% Polynomial 20 (48.8%) 40 (54.1%) 27 (46.6%) 50.3% Polynomial 14 (34.1%) 38 (51.4%) 26 (44.8%) 45.1% Polynomial 19 (46.3%) 26 (35.1%) 26 (44.8%) 41.0% Polynomial 15 (36.6%) 26 (35.1%) 24 (41.4%) 37.6% RBF 28 (68.3%) 51 (68.9%) 24 (41.4%) 59.5% Structure 2 Linear 29 (70.7%) 54 (73.0%) 24 (41.4%) 61.8% Polynomial 20 (48.8%) 43 (58.1%) 16 (27.6%) 45.7% Polynomial 20 (48.8%) 46 (62.2%) 23 (39.7%) 51.4% Polynomial 13 (31.7%) 44 (59.5%) 15 (25.9%) 41.6% Polynomial 19 (46.3%) 37 (50.0%) 18 (31.0%) 42.8% Polynomial 15 (36.6%) 43 (58.1%) 12 (20.7%) 40.5% RBF 29 (70.7%) 53 (71.6%) 24 (41.4%) 61.3% Structure 3 Linear 27 (65.9%) 52 (70.3%) 23 (39.7%) 59.0% Polynomial 22 (53.7%) 46 (62.2%) 16 (27.6%) 48.6% Polynomial 20 (48.8%) 40 (54.1%) 24 (41.4%) 48.6% Polynomial 19 (46.3%) 38 (51.4%) 17 (29.3%) 42.8% Polynomial 26 (63.4%) 26 (35.1%) 18 (31.0%) 40.5% Polynomial 22 (53.7%) 26 (35.1%) 12 (20.7%) 34.7% RBF 25 (61.0%) 49 (66.2%) 21 (36.2%) 54.9%