Table 1.
Related works on 3D cell tracking.
Figure 1.
A schematic representation of the tracking algorithm steps.
The gray boxes are significant improvements with regards to our preliminary work [2].
Figure 2.
Cell appearance in 3D collagen gels.
(a) Cell and collagen gel aspects in an original phase-contrast Z-slice with a zoom showing the collagen fibers. (b) Schematic presentation of the 3D time-lapse sequence acquisition. Detailed cell aspect in (c) the original phase-contrast volume (after contrast enhancement) and (d) the correlated volume, illustrated by several Z-slices, a XZ cut plane (vertical slice) and an intensity isosurface 3D view.
Figure 3.
Cell detection in a correlated volume.
Illustration of (a part of) a single Z-slice containing 3 cells in focus (pointed by arrows) submitted to the different segmentation steps (correlated phase-contrast, mask, softMax and the final segmented volume). The magnified region of interest shown in the upper left corner of each image is centered on the cell located in the middle of the image.
Figure 4.
Schematic illustration of tracker management when two cell paths intersect.
The timeline bellow the schematic frames presents the start and end points of the 4 trajectories obtained in this situation.
Figure 5.
Detection of new cells in a volume sequence.
The new cells result from (frame 2) a cell division or (frame 3) a cell entry into the observed volume. The trajectories are rendered on (1a–3a) the average intensity Z projections from the correlated volumes and (1b–3b) the corresponding intensity isosurfaces (3D view).
Figure 6.
Tracking of an elongated cell.
An elongated cell (blue dot) spawns a second parasite tracker (red dot) on a body extension (on the cell tail before its retraction when migrating). The secondary tracker is discarded as it converges onto the cell's original tracker within a couple of frames. The images show Z-slices centered on the cell in the phase-contrast (1st row) and correlated (2nd row) volumes. The timeline below the images is similar to that in figure 4.
Figure 7.
False cell detection due to phase-contrast interferences.
The Z-slices of the phase-contrast (top) and correlated (bottom) volumes are centered on phase-contrast interferences falsely detected as a cell (blue dot). The actual cells are visible in the phase-contrast slice as dark elongations touching.
Figure 8.
Number of active cell trajectories across a sequence.
Set1 collects trajectories which are generated from the 1st volume without additional use of the cell detection step in the next volumes. Set2 collects all the other cell trajectories which are initialized later in the sequence thanks to this detection step. The graph shows the number of active trajectories (i.e., which are not stopped because of a collision with another cell tracker) in each set, and their union, over time.
Table 2.
3D gel compositions in the different experiments.
Figure 9.
U373 cancer cell trajectories and the extracted features characterizing 3D migration behavior.
(a) These trajectories are observed on the complete sequences in the absence (p42h0) or presence (p24h18) of 24% of HA in the collagen gel (b) The features (labeled on the Y axes) quantify the migration abilities of U373 cells in 3D gels which included a progressive increase of HA amount (see Table 2). The data distributions are presented by their median (square), inter-quartile range (box) and non-outlier range (bars).