Biomedical Engineering Reference
In-Depth Information
correlation coefficient between the corresponding windows of two subsequent
frames to be calculated as:
MN
R
ij
p
ð
;
q
Þ¼
ð
M
p
Þ
ð
N
q
Þ
2
3
Mp
P
Nq
P
I
i
m
ð
;
n
Þ
I
j
m
ð
þ
p
;
n
þ
q
Þ
4
5
:
m¼
1
n¼
1
(7.8)
M
P
p
N
P
q
I
i
m
ð
;
n
Þ
I
j
m
ð
;
n
Þ
m¼
1
n¼
1
In addition to the common cross-correlation analysis, not only the peak position
but also the cross-correlation function at the peak can be evaluated. This latter
function value characterizes the decorrelation rate discussed above. Full cross-
correlation analysis requires a relatively large interrogation time, and only quasi-
real-time operation with a frequency rate of about 5 maps/s can be realized with the
hardware currently available.
The information obtained with cross-correlation analysis seems to be a little
excessive as it contains the direction of the averaged biospeckle displacement. For
such random fields as subskin blood flux it seems that either decorrelation or simple
autocorrelation analysis would be sufficient if only the value of the averaged blood
flux intensity needs to be extracted.
7.7.4 Efficiency of the Evaluation
The time-space cross-correlation analysis of the temporal evaluation of the dynamic
biospeckle patterns is shown to be a means of real time flow visualization of a living
tissue blood microcirculation. Digital processing of biospeckle patterns records
yields 2D maps exhibiting the temporal and spatial variations in the blood flow.
This might be used for biomedical diagnostic purposes to detect, say, a microscale
deviation from the normal case. The three methods of evaluation of the dynamic
speckle patterns have been tested. Both decorrelation and autocorrelation analysis
were realized in a real time mode, when entire treatment of the digital specklegram
was performed over the time interval between successive frames (40 ms). Results in
the form of 2D maps of subskin blood flux were visualized on the PC monitor at a
frequency of 25 Hz.
Full cross-correlation analysis of the dynamic biospeckle pattern needs rather
more PC time and, with the hardware currently available, only quasi-real time
operation could be achieved at a frequency of about 5 maps/s. The information
obtained with cross-correlation analysis seems to be somewhat excessive as it
contains the direction of the averaged biospeckle displacement. For random fields
such as subskin blood flux it appears that decorrelation or simple autocorrelation
analysis would be sufficient when only the value of the averaged blood flux intensity
Search WWH ::
Custom Search