Conclusion and Discussion
5.2 Future Work
This thesis has reported some preliminary results of investigating the microcirculatory behaviors. Although the scheme and methods have already enabled us to estimate the capillary blood velocity, there still exist some issues requiring further investigation.
1. At the preprocessing stage, the image contrast needs to be further enhanced in order to achieve better results of realignment. However, the instrument for microscopic image acquisition is limited to manual adjustment. In the future, contrast-related parameters will be adjusted and optimized automatically.
2. We may use the intensity of illumination (average of R, G, and B values) to recognize capillaries and blood flow.
3. In this thesis, the Zhang-Suen algorithm was applied to image skeletonization.
Although this algorithm was computational efficient, the performance was not satisfactory that also affected the selection of centers of observation windows. As a result, it is important to develop advanced algorithm for image skeletonization.
4. Computerized automation algorithm needs to be developed for determining the observation windows.
Our research group has been investigating the Zen-meditation life system, in both physiological and mental aspects. Study on microcirculatory phenomena was originally aimed to investigate the effects caused by Zen meditation. In addition, Zen-meditation effects on microcirculatory system can be compared with those observed from other physiological signals such as heart rate variation (HRV), electrocardiograph (ECG), galvanometric skin resistance (GSR), respiratory signals,
etc.
There are two other subjects that also draw our attention.
1. Register the same capillary
For long-term monitoring, it is important to always measure the capillary blood velocity for the same capillary in one subject. Accordingly, we may explore the long-term effects on microcirculatory characteristics caused by Zen meditation. We are thus looking for the possibility of registering one unique capillary.
2. Morphic recognition
Health conditions of a subject can be revealed by the morphic patterns of the capillaries. In addition, microcirculatory behaviors have been used as a diagnostic reference to examine such potential diseases as 1) a connective tissue disease or macroglobulinemia signified by increasing diameters and varying contours of the capillaries, 2) low blood pressure revealed by a rapid decrease in the number of capillaries, and 3) critical infections (during the critical periods of several diseases) indicated by a slowing down or cessation of the capillary blood velocity (velocity <
198μm/sec) as well as the elongation of the diameter of the capillaries. Scheme and algorithm for quantifying the capillary morphologies [31] will be developed in the future to measure the capillary diameters, to recognize their contours, to count the number, to detect hemorrhage, etc.
Finally, the nailfold capillary microscope is very important to the prediction and diagnosis of a number of diseases. Development of prognostic and diagnostic schemes significantly contributes to preventive medicine that has been in urgent need in comparison with therapeutic medicine.
References
[1] Aubin F., Dufour MP., Risold JC., Lucas A., Humbert P., “Nailfold capillary microscopy in human ummunodeficiency virus-infected patients: a case-control study,” Microvascular Reasearch, vol. 58, pp.197－199, 1999.
[2] Martina B., Frach B., Surber C., Drewe J., Battegay E. and Gasser P., “Capillary blood cell velocity in finger nailfold: effect of Enalapril and Mibefradil in patients with mild to moderate hypertension,” Microvascular Research, vol.57, pp.94－99, 1998.
[3] Brumen V., Horvat D. and Bonić I., “Evaluation of serial application of capillaroscopy, photoplethysmography, and dermothermometry in diagnosis and prevention of radiolesions of peripheral microvessels,” Microvascular Research, vol.47, pp.270－278, 1994.
[4] VADER, SHERMAN, and LUCIANO, Human Physiology. 6th edition, McGRAW-Hill.
[5] Stücker M., Baier V., Reuther T., Hoffmann K. and Altmeyer P., “Capillary blood cell velocity in human skin capillaries located perpendicularly to the skin surface:
Measured by a new Laser Doppler Anemometer,” Microvascular Research, vol.52, pp. 188－192, 1996.
[6] 李錦奎，「體表微循環量測方法與應用」，國立台灣大學，碩士論文，民國 92 年。
[7] Norman M., Herin P. and Fagrell B., “An evaluation of skin capillary blood flow determinations in neonates using a computerized videophotometric method,”
Microvascular Research, vol.43, pp.276－284, 1992.
[8] 田牛，微循環方法學，初版，原子能出版社，1987。
[12] Einav S. and Berman H. J., “Fringe mode transmittance laser Doppler microscope anemometer: Its adaptation for measurement in the microcirculation,” Biomedical Engineering, vol.10, pp.393－399, 1988.
[13] 易政男，「微循環顯微影像之擷取與分析」，國立中央大學，碩士論文，
民國90年。
[14] Yau H. F., Yih J. N., Lee H. Y., Lai Y. K., Chang H. H., “In vivo observation of the blood flow in human capillaries under finger nailfold,” Medical and Biological Engineering, vol.22, pp.91－96, 2002.
[15] Intaglietta M., Silverman N. R. and Tompkins W. R., “Capillary flow velocity measurements in vivo and in situ by television methods,” Microvascular Research, Vol. 10, pp. 165－179, 1975.
[16] 胡育彰，「動態影像估測之醫學影像應用」，私立中原大學，碩士論文，
民國90年。
[17] Horn K. P. and Schunck B. G., “Determing optical flow,” Artificial Intelligence, pp. 185－203, 1981.
[18] Wayland H. and Johnson P. C., “Erythrocyte velocity measurement in microvessels by two-slit photometric method, ” Applied Physiology, vol.22, pp.333－337, 1967.
[19] Tsukada K., Minamitani H., Sekizuka E., Oshio C., “Image correlation method for measuring blood flow velocity in microcirculation: correlation｀window＇
simulation and in vivo image analysis,” Physiological Measurement, vol.21, pp.459－471, 2000.
[20] 陳國禔，「微循環顯微觀測系統之設計製作與觀測指標模糊理論應用初步 評估」，私立中原大學，碩士論文，民國90年。
[21] Rao K. R. and Hwang J. J., Techniques and Standards for Image,Video and Audio Coding. Prentice Hall, 2001.
[22] Otsu N, “A threshold selection method from gray－level histogram,” IEEE Transactions on system, man, and cybernetics, vol.9, pp.62－66, 1979.
[23] Shapiro L.G. and Stockman G. C., Computer Vision. Upper Saddle River, N.J.:
Prentice Hall, 2001.
[24] McAndrew A., Introduction to Digital Image Processing. 1st edition, Thomson, 2004.
[25] Hahn M., Heubach T., Steins A., and Jünger M., “Hemodynamics in nailfold capillaries of patients with systemic scleroderma: synchronous measurements of capillary blood pressure and red blood cell velocity,” Investigative Dermatology, vol.110 (6), pp.982－985, 1998.
[26] 余麗君、姜亞芳，病理生理學，初版，五南圖書出版有限公司，2003。
[27] Nelson P., Biological Physics: Energy, Information, Life. 1st edition, Freeman, 2004.
[28] Carr R. T. and Lacoin M., “Nonlinear dynamics of microvascular blood flow,”
Annals of Biomedical Engineering, vol. 28, pp.641－652, 2000.
[29] Liu C.Y., Wei C.C. and Lo P.C., “Variation Analysis of the Sphygmogram to Assess the Cardiovascular System under Meditation”, Evidence-based Complementary and Alternative Medicine (doi:10.1093/ecam/nem065).
[30] Westerhof N., Stergiopulos N., and Noble M. I. M., Snapshots of Hemodynamics.
Springer US, 2005.
[31] Jones B.F., Oral M., Morris C. W., and Ring E. F. J., “A proposed taxonomy for nailfold capillaries based on their morphology,” IEEE Transactions on Medical Imaging, vol.20 (4), pp.333－341, 2004.
Appendix A
Other matching criteria
1. Sum of Absolute Difference (SAD)
The SAD is a popular measure requiring less computational time than the other methods. It is defined as
| ( , ) ( , ) | (A.1)
i j
SAD=
∑∑
S i j −R i jWe denoted the reference image as R(i, j) and the other image as S(i, j), where i and j are the pixel coordinates of the image in the image space.
2. Cross-correlation Coefficient (CC) The cross-correlation coefficient is defined as
2
where X=S(i, j), Y=R(i, j), and NEP is the total number of pixels included in the summation calculation.
3. Standard deviation of gray-level ratio (SDGR)
This measurement computes the standard deviation of gray-level ratio (SDGR).
First, an N×N ratio image, Ratio(i, j), is derived by dividing the gray level of target image S(i,j) by the non-zero gray level of the reference image R(i,j) at the same pixel location. Then the standard deviation is computed for all the non-zero gray values in the ratio image.
Ratio(i, j) = {S(i, j) / R(i, j) if R(i, j) ≠0 (A.3) 0 if R (i, j) =0 ∀(i, j)
and SDPR=standard deviation of {Ratio (i, j) | 0 ≤ i, j < N and Ratio (i, j)≠0 }
Appendix B
Otsu’s Method:
Let the pixels of a given picture be represented in L gray levels [1, 2, …, L]. The number of pixels at level i is denoted by and the total number of pixels by
. In order to simplify the discussion, the gray-level histogram is normalized and regarded as a probability distribution:
ni Now suppose that we dichotomize the pixels into two classes and (background and objects) by a threshold at level k; denotes pixels with levels [1, …, k], and denotes pixels with levels [k+1, …, L]. Then the probabilities of class occurrence and the class mean levels, respectively, are given by
C0 C1 Are the zeroth- and the first-order cumulative moments of the histogram up to the kth level, respectively, and
1 is the total mean level of the original picture. It can be easily verify the following relation for any choice of k:
0 0 1 1 T, 0 1 1
ω μ ω μ+ =μ ω ω+ = (B.9) The class variances are given by
2 2 2 These require second-order cumulative moments (statistics).
In order to evaluate the “goodness” of the threshold (at level k), Otsu introduce the following discriminant criterion measures (or measures of class separability) used in the disciminant analysis: are the within-class variance, the between-class variance, and the total variance of levels, respectively. Then our problem is reduced to an optimization problem to search for a threshold k that maximizes one of the object functions (the criterion measures) in (B.12).
This standpoint is motivated by a conjecture that well-thresholded classes would be separated in gray levels, and conversely, a threshold giving the best separation of
classes in gray levels would be the best threshold.
The discriminant criteria maximizing λ,κ, and η, respectively, for k are, however, equivalent to one another; e.g., κ λ= +1and η λ λ= /( + in term of1) λ , because the following basic relation always holds:
2 2 2.
w B T
σ +σ =σ (B.16)
It is noticed that σw2 and σB2 are function of threshold level k, but σT2 is independent of k. It is also noted that σw2 is based on the second-order statistics
(class variances), whileσB2 is based on the first-order statistics (class means).
Therefore,ηis the simplest measure with respect to k. Thus Otsu adopt η as the criterion measure to evaluate the “goodness”(or separability) of the threshold at level k.
The optimal threshold k* that maximizes η, or equivalently maximizes σB2, is selected in the following sequential search by using the simple cumulative quantities (B.6) and (B.7), or explicitly using (B.2)-(B.5):
( )k B2( ) /k T2 and then the optimal threshold k* is determined by
2