Evaluation of automated label generation with SpinX. (a) Time-lapse image stills of cell membrane, spindle (SiR-Tubulin), and chromosomes. The segmentation results are outlined in green. (b) Graph shows computational run time for chromosome (green) and spindle (red) annotation generation using SpinX. Note a log scale has been applied on the y-axes. Each dot indicates an additional 100 images. (c) Bar graph shows image-wise accuracy for chromosome and spindle channels. An image is defined as mis-segmented if SpinX fails to segment the full object’s boundaries or if the properties (e.g., orientation) are incorrect. Accuracy percentages are calculated by the number of mis-segmented frames over the total number of frames. (d) Automated chromosome and spindle segmentation of cell trajectories over time without correction. Mis-segmented frames are highlighted in red. The time interval between each frame is 3 min. (e) The correlation matrix displays the calculated correlation coefficient ρ which denotes the matching between human segmentation versus automated segmentation. The histogram shows the distribution of ρ grouped by channels. (f) Time series graph shows chromosome-spindle formation at metaphase. Dash-line at angle φ = 90° represents perfect perpendicular properties between spindle and chromosomes. N = 4,473 images per channel. Data obtained from 88 movies of metaphase cells over at least six independent experiments.