Fig. 3

Enhanced cell tracking of VAL cells in 3d migration chambers using pseudofluorescence. (A). Representative image pairs from the dataset used for training, including TL images (up), and manually annotated binary masks (bottom) (n of images included = 159). (B). Comparison of the signal-to-noise ratio (SNR) between cells observed via TL, CFP labeled cells (CFP), and pseudofluorescence (PF). The SNR value of each population derives from the mean of three independent observations. (C). Comparison of the intensity variation (intensity at time t / intensity at time 0, average of all the cells) over time of pseudofluorescence (magenta) and CFP (blue), showing the effect of photobleaching on CFP. (D). Representative micrographs showing the transformation into pseudofluorescence of cells which are poorly visible in TL and outside the focal plane. (E). Comparison of automatic tracking accuracy (Track duration) using TL, pseudo-fluorescence, CFP-labeled cells and pseudofluoresence calculated on the CFP positive cells. Values of each population were calculated from the mean of three independent observations. Track duration is expressed in % w.r.t the total video duration. (F). Representative micrographs showing tracks of cells obtained using PF (magenta lines) and CFP signals (cyan lines). Red arrow indicates a cell not detected using CFP due to photobleaching. (G). Representative micrograph showing tracking errors (red arrow, glitch between cells in close proximity) when using PF with respect to TL. (H). Results obtained on a 3D migration chamber (color-coded tracks, blue = 0 s, red = 8 hours). (I). Visual comparison between manually tracked cells (black lines), and automatically tracked cells using pseudofluorescence channel (magenta lines). num. Tracks = 40, 96, 59. (J). Quantitative evaluation of multiple object tracking accuracy. Statistics performed with Mann-Whitney t-test relative to FP as control, * = P < 0.05 *** = P < 0.001