Table 5 Cell counting accuracy for different methods.

From: Segmentation, tracking, and sub-cellular feature extraction in 3D time-lapse images

Sequence

Ground truth

ACME10

MARS11

Supervoxel method12

Our method

Sequence 1

23

21.5 (3.2)

25.5 (2.2)

24 (1.1)

23.5 (0.9)

Sequence 2

30

41.1 (3.1)

35.1 (2.8)

32 (2.1)

30.1 (0.8)

Sequence 3

25

22.6 (2.1)

27.5 (3.2)

24 (1.5)

25 (0.5)

Sequence 4

18

18.8 (1.2)

18.5 (1.2)

18.2 (1.2)

18 (0.6)

Sequence 5

28

31.5 (2.9)

24.5 (2.3)

26.2 (1.1)

27.8 (1.0)

  1. For each time sequence, there is a fixed number of cells. Due to segmentation error, the algorithms can generate different number of cells for different time points of the sequence. The table shows average number of detected cells (standard deviation values in parenthesis) for the entire sequence.
Back to article page