Skip to main content
. 2020 Sep 2;10(24):11026–11048. doi: 10.7150/thno.44053

Figure 3.

Figure 3

Assessing ALICE's performance and capabilities. (A) Identification, localization and enumeration of 5 representative simulated cell phenotypes among 300 synthetic cells in a sample image along with output cell thumbnails for 4 channels: Nuc, M1, M2 and M3 that denote respectively, the nucleus and cytoplasm marker 1, 2, 3. (B) Mean percentage error and SD in the enumeration of 20 cell phenotypes (all n=100). (C) Sensitivity and specificity in the enumeration of 20 cell phenotypes (all n=100). Error bars represent the 95% CI. (D) Assessing ALICE cell enumerative performance against ImageJ and CellProfiler using 4 publicly available real image sets. (E) Characterizing the processing time with respect to image size, processing scheme and result export. Insets show representative image and size dimension. Error bars represent SD. *** - P < 0.001. (F-I) Passing-Bablok regression analysis for P3, P7, P11 and P15 (all n=100) with results of the remaining phenotypes shown in Figure S3. Black dash lines represent the identity line, red solid lines represent the fitted line and blue solid lines represent the 95% CI of the fitted line. Insets depict individual fluorescent channel images, τ denotes the Kendall's correlation coefficient and m denotes the slope. (J-M) Bland-Altman plots for the same 4 phenotypes (all n=100) with results of the remaining phenotypes shown in Figure S4. ΔCount denotes the difference between the 2 counts. Orange dash lines represent the mean difference between ALICE's count and the simulated ground truth, purple solid lines represent the 95% limits of agreement and brown dotted lines represent the 95% CI of the limits of agreements. The phenotypes are defined in Table S2.