Table 3

Validity of automatically extracted cell tracks

Image sequence no.

Cell detection rate median (min, max)

# of frames/# of required cell-to-cell associations

Swap errors

Lost or deleted

Track detection (correct/total)

%


1, untreated

0.98 (0.92, 1.0)

63/4960

11

1

68/80

85

2, spc

0.98 (0.92, 1.0)

60/6077

9

4

90/103

87

3, tgfβ

0.96 (0.90, 0.99)

64/4284

17

2

49/68

73

4, tgfβ & U0126

0.97 (0.90, 1.0)

58/3933

3

3

63/69

91

5, U0126 & spc

0.99 (0.95, 1.0)

60/5900

6

5

89/100

89


Automatic track detection consists of cell identification and track generation. The cell identification rate is measured over all individual images. A cell track was counted as swapped and thus false in two cases: either if two tracks "exchanged" their cells (which leads to a double swapping error) or if the merging during the cell division (backward tracking) happened with the wrong child cell. The total number of cell-to-cell associations for each video file is given in column 3 (e.g., video 1 consists of 80 tracked cells over 63 frames, requiring a total of 62 × 80 = 4960 cell-to-cell associations. Only 11 of those were incorrect (0.2%), demonstrating an association performance of 99.8% for sample video 1).

The proportion of tracks that were followed correctly across all frames (i.e. without any form of mis-association of cells) is given in column 6. The third video clearly shows the highest swapping error, which was expected as it contains the fastest cells and the lowest cell detection sensitivity (0.96).

Huth et al. BMC Cell Biology 2010 11:24   doi:10.1186/1471-2121-11-24

Open Data