Biomedical Engineering Reference
In-Depth Information
Table 5.6:
Repeatability and agreement study within observers
and methods, respectively. ObsI and ObsII stand for each observer,
and MB indicates the model-based technique. The table shows
µ
±
SD of the difference of the measurements in mm. The standard
deviation has been corrected for repeated measurements in the
agreement study [33]
Neck (mm)
Width (mm)
Depth (mm)
ObsI
−
0
.
07
±
1
.
09
0
.
94
±
1
.
87
−
0
.
65
±
2
.
41
ObsII
−
0
.
51
±
0
.
86
−
0
.
34
±
1
.
35
0
.
18
±
1
.
50
ObsI vs ObsII
−
0
.
03
±
1
.
22
0
.
34
±
1
.
91
−
0
.
70
±
2
.
45
ObsI vs MB
−
0
.
47
±
1
.
05
0
.
23
±
1
.
86
−
0
.
69
±
2
.
12
ObsII vs MB
−
0
.
44
±
0
.
91
−
0
.
11
±
1
.
43
0
.
00
±
1
.
55
5.4.2.2
Bland-Altman Study
The Bland-Altman plot is a statistical method of comparison of two clinical
measurement techniques. The agreement between the two techniques can be
quantified using the standard deviation of the differences between observations
made on the same subjects. Bland-Altman graphs show the distribution of the
differences by plotting the mean against the differences of paired measurements.
The information of the Bland-Altman graphs can be summarized by providing
the bias (
µ
) and standard deviation (SD) of the differences of the measurements.
The limits of agreement, defined as
µ
±
1
.
96
·
SD, provide an interval within
which the 95% of the differences between measurements are expected to lie.
When repeated measurements from two techniques are available, a corrected
standard deviation is computed [33]. A very similar analysis to the limits of
agreement approach can be applied to quantify the repeatability of a method
from replicated measurements obtained from the same measurement technique.
The results of the Bland-Altman study are shown in Table 5.6. Figure 5.10
shows the Bland-Altman graphs.
5.5
Discussion
Classic GAC approaches were unsatisfactory for segmenting the cerebral vascu-
lature from CTA and more sophisticated speed functions introducing statistical
