Searching trees: live time-lapse cell-cycle progression modeling and analysis
A method of tracking a cell through a plurality of images includes selecting the cell in at least one image obtained at a first time, generating a track of the cell through a plurality of images, including the at least one image, obtained at different times using a backward tracking, and generating a cell tree lineage of the cell using the track.
1. A method of tracking a cell through a plurality of images comprising:
selecting at least one cell in at least one image obtained at a first time;
generating a track of the at least one cell through a plurality of images, including the at least one image, obtained at different times using a forward tracking and a backward tracking, wherein generating the track comprises distinguishing, using information generated for the at least one cell by the forward tracking, between a visually occluded cell in at least one of the plurality of images and a mitosis event detected by the backward tracking; and
generating a cell tree lineage of the at least one cell using the track.
2. The method of claim 1 , wherein generating the track of the cell through the plurality of images further comprises modeling a cell cycle duration of the cell.
3. The method of claim 1 , wherein generating the track of the cell through the plurality of images further comprises using the backward tracking in modeling a rate of mitosis of the cell.
4. The method of claim 1 , further comprising labeling, simultaneously, a plurality of cells, including the at least one cell, with different identities in the plurality of images.
5. The method of claim 1 , further comprising visualizing a progression of the cell over time through a graphical tree structure.
6. The method of claim 1 , further comprising building a tree of potential track hypotheses for the cell.
7. The method of claim 6 , further comprising inserting a dummy node in the tree to represent an occluded cell, wherein the dummy node represents the visually occluded cell that is visually occluded in the at least one image by another cell.
8. The method of claim 1 , further comprising analyzing an appearance cue of the cell accounting for a similarity of the cell to a pairwise cell over time.
9. The method of claim 1 , further comprising inferring high-level information from the track including at least one of a rate of mitosis of the cell and a velocity of the cell.
10. The method of claim 1 , further comprising:
determining the mitosis event detected by the backward tracking is a false positive upon distinguishing between the visually occluded cell and the mitosis event; and
removing the mitosis event from consideration for inclusion in the track of the cell.
11. A method of tracking cells in a time-lapse sequence of images comprising:
detecting a plurality of cells in at least one image of the time-lapse sequence of images;
generating, from the at least one image, a track for each of the plurality of cells through the time-lapse sequence of images using a forward tracking;
performing a backward tracking tracing each of the tracks back in time through the time-lapse sequence of images until detecting a potential respective mitosis event where two of the tracks claim a same cell;
labeling, simultaneously, the plurality of cells with different identities in the plurality of images using the forward tracking and the backward tracking;
generating a cell tree lineage of the plurality of cells using the tracks; and
modeling a rate of mitosis for the plurality of cells using the cell tree lineage,
wherein the labeling of the plurality of cells with different identities in the plurality of images using the forward tracking and the backward tracking comprises:
distinguishing, for each of the potential respective mitosis events, between a visual occlusion of one cell by another cell and an actual mitosis event; and
labeling, for at least one distinguished visual occlusion, each of the cells associated with the visual occlusion in at least one image of the time-lapse sequence of images.
12. The method of claim 11 , further comprising:
determining a confidence for each of the tracks; and
pruning at least one track associating cells in sequential images of the time-lapse sequence of images, the at least one track having a confidence less than a threshold.
13. The method of claim 12 , wherein the pruning is performed prior to the backward tracking.
14. The method of claim 12 , wherein the confidence of each of the tracks is determined using a combination of feature similarity scores for the respective cell over time.
15. The method of claim 11 , further comprising:
detecting an abnormal condition in the cell tree lineage; and
determining a risk associated with the plurality of cells.
16. The method of claim 11 , wherein generating the cell tree lineage comprises outputting a representation comprising a plurality of slices corresponding to the plurality of images arranged in order of time, wherein the presentation includes a visualization of a plurality of tracklets, each tracklet corresponding to a location of a given one of the plurality of cells over time.
17. The method of claim 11 , further comprising modeling a cell cycle duration of a first cell of the plurality of cells using a respective one of the mitosis events as a starting time of a cell cycle of the first cell.
18. The method of claim 11 , wherein the generation of the cell tree lineage of the plurality of cells includes adding a dummy node in the cell tree lineage representing the visually occluded cell in the at least one image.
19. The method of claim 11 , further comprising removing at least one miss detected mitosis event from consideration for inclusion in a first track of a first cell upon distinguishing between a visual occlusion of one cell by another cell and an actual mitosis event, wherein the first cell is the one cell or the another cell.