Microscopic Slice Image Processing and Automatic Labeling - Biomedical Signal Analysis: Contemporary Methods and Applications - page 374

350

expert 1
expert 2

300

15

250

200

10

150

5

100

50

0 500 1000 1500 2000

0

0 50 100 150 200 250 300 350

counted cells

ZANE (counted cells/image)

(a) inter-observer variability (b) ZANE vs. manual counting

1

1

2

2

ZANE
clusterin g
watershed
man+foc
manual

ZANE
clustering
watershed

3

3

4

4

5

5

0 20 40 60 80 100

0 20 40 60 80 100

counted cells algorithm runtime (s)

(c) comparison of counting methods (d) runtime comparison

Figure 13.11
Cell-counting results. (a) gives the counting results of two experts using a
microscope (and changing the focus plane). (b) presents a comparison of ZANE
versus manual microscope-based counting within various mice. (c) compares the
ZANE counting performance with other proposed counting method as well as
manual counting by microscope, counting only cells in the digital image within one
mouse. (d) lists the corresponding algorithm runtimes. In (b) and (c), a single
expert made all manual counts.

set (from two mice). The slope of the fitted line is about 0 . 9, which
implies a counting error of around 10%.
The comparisons of ZANE with other counting algorithms also favor
our proposed algorithm; indeed, the more advanced shape selectivity is