Fig. 6.

Quantitative analysis of ACDC accuracy in detecting cilia across different experimental conditions. a Manual and automated detection of NIH3T3 primary cilia labeled with AcTub, Smo-GFP, and Arl13b. Manual true candidates (TC) represent the total number of true cilia. Automated TC represents the number of manual TC that was detected in automated analysis mode. Among the three reporters, Arl13b had the lowest FP rate (1%) and shared the lowest FN rate with AcTub (4%), resulting in the highest F1 score (F1 = 0.97). Automatically detected TC (dark green bar) plus the corresponding FNs (orange bar) equals the same count as Manual TC (light green bar). b Manual and automated detection of htert-RPE1 primary cilia labeled with AcTub, Smo-GFP, and Arl13b. Among the three reporters, Arl13b had the lowest FP rate (6%) and shared the lowest FN rate with AcTub (1%), resulting in the highest F1 score (F1 = 0.96). c Manual and automated detection of Arl13b-labeled primary cilia in htert-RPE1 cells at two different image magnifications (×60 vs. ×40). Precision and recall values were similar across both groups, and the F1 scores were identical for images taken at both magnifications (F1 = 0.96). Each condition (i.e., AcTub-labeled NIH3T3 cilia, Arl13b-labeled RPE cilia) was its own separate experiment that was conducted once, so as to demonstrate the cilia FP and FN rates of each condition. The image analysis process for the data set of each condition was repeated three times and each iteration yielded the same values for FP and FN rates. All main bars and error bars are reported as average absolute cilia count per image ± standard deviation per image