Reliability and validity of tongue color analysis in the prediction of symptom patterns in terms of East Asian Medicine Academic research paper on "Medical engineering"

JTCM

Online Submissions: http://www.journaltcm.com info@journaltcm.com

JTradit Chin Med 2016 April 15; 36(2): 165-172 ISSN 0255-2922 © 2016 JTCM. All rights reserved.

CLINICAL STUDY

Reliability and validity of tongue color analysis in the prediction of symptom patterns in terms of East Asian Medicine

Young-Jae Park, Jin-Moo Lee, Seung-Yeon Yoo, Young-Bae Park

Young-Jae Park, Department of Diagnosis and Biofunctional Medicine, College of Korean Medicine, Kyung Hee University, Seoul 134-727, Korea; Department of Diagnosis and Biofunctional Medicine, Kyung Hee University Hospital at Gangdong, Seoul 134-727, Korea Jin-Moo Lee, Department of Oriental Gynecology, College of Korean Medicine, Kyung Hee University, Seoul 134-727, Korea; Department of Women Health Clinic, Kyung Hee University Hospital at Gangdong, Seoul 134-727, Korea Seung-Yeon Yoo, Young-Bae Park, Department of Diagnosis and Biofunctional Medicine, College of Korean Medicine, Kyung Hee University, Seoul 134-727, Korea Correspondence to: Young-Jae Park, Department of Diagnosis and Biofunctional Medicine, Kyung Hee University Hospital at Gangdong, Seoul 134-727, Korea. bmpomd@khu.ac.kr Telephone: +82-2-440-7229 Accepted: July 20,2015

Abstract

OBJECTIVE: To examine whether color parameters of tongue inspection (TI) using a digital camera was reliable and valid, and to examine which color parameters serve as predictors of symptom patterns in terms of East Asian medicine (EAM).

METHODS: Two hundred female subjects' tongue substances were photographed by a mega-pixel digital camera. Together with the photographs, the subjects were asked to complete Yin deficiency, Phlegm pattern, and Cold-Heat pattern questionnaires. Using three sets of digital imaging software, each digital image was exposure- and white balance-corrected, and finally L* (luminance), a* (red-green balance), and b* (yellow-blue balance) values of the tongues were

calculated. To examine intra- and inter-rater reliabilities and criterion validity of the color analysis method, three raters were asked to calculate color parameters for 20 digital image samples. Finally, four hierarchical regression models were formed.

RESULTS: Color parameters showed good or excellent reliability (0.627-0.887 for intra-class correlation coefficients) and significant criterion validity (0.523-0.718 for Spearman's correlation). In the hierarchical regression models, age was a significant predictor of Yin deficiency (P = 0.192), and b* value of the tip of the tongue was a determinant predictor of Yin deficiency, Phlegm, and Heat patterns (P = - 0.212, - 0.172, and - 0.163). Luminance (L*) was predictive of Yin deficiency (P = - 0.172) and Cold (P = 0.173) pattern.

CONCLUSION: Our results suggest that color analysis of the tongue using the L*a*b* system is reliable and valid, and that color parameters partially serve as symptom pattern predictors in EAM practice.

©2016 JTCM. All rights reserved.

Key words: Medicine, East Asian Traditional; Complementary therapies; Tongue inspection; Yin deficiency; Symptom pattern

INTRODUCTION

East Asian medicine (EAM), tongue inspection (TI) is a process which involves visually inspecting the body of the tongue for color, shape, moisture, and movement,

and the tongue coating for color, thickness, and distribution.1 According to EAM, the tongue is directly connected to the internal organs through the heart, liver, spleen, lung, kidney, and urinary bladder meridians.2 As the tongue is a sensitive mirror for changes in internal organs, TIs serve as an important guide in determining pathological patterns. For example, pale redness of the tongue proper refers to a blood deficiency or Cold pattern, whereas dark redness or blueness of the tongue proper refers to an aggravated Heat or Cold pattern.3 Thickness of the tongue coating is associated with a Phlegm or food retention pattern.3 The pattern (or symptom pattern) refers to a group of symptoms, defined in terms of EAM theory, whereby a medical condition is diagnosed and treated in EAM practice. Although TI is an important diagnostic method to identify symptom patterns, expertise is a matter of subjective experience, and the results of the TI are influenced by environmental factors, such as light source or brightness.4 Moreover, the results of traditional TIs are obscure, expressed in the form of a three or four rank scale, such as "pale," "normal," "red" and "dark red," which is problematic when attempting to quantify and establish efficacy.2,3 Some studies have been conducted to quantify the results of traditional TIs and are categorized into two types: the development of medical instruments and the facilitation of mathematical algorithms. Kim et al 5,6 developed a digital tongue imaging system and reported that the thickness of the tongue coating was reliable and well-matched to clinician assessment. Yamamoto et al7 reported that hyperspectral imaging system results for the tongue images were a feasible surrogate for expert visual tongue analysis. In terms of the facilitation of algorithms, Pang et al8 and Watsuji et al 9 reported that Bayesian networks and fuzzy theory were useful for determining diseases or symptom patterns, respectively. However, these instruments and algorithms have some limitations in terms of broad application because the instruments are not widely used, and there is a need for special knowledge to apply the algorithms. Therefore, an assessment tool for TI that is popular, easy-to-use, and can present quantitative parameters is needed. Recently, the usage of mega-pixel digital cameras has been widely adopted in various medical fields, including dentistry, ophthalmology, and dermatology.10-12 Moreover, parameters of color systems, such as RGB (red, green, blue) and L*a*b* (luminance, red-green balance, yellow-blue balance), can be calculated using digital imaging software, such as Adobe Photoshop and Picture Color Analyzer.13 The L*a*b* system is one of the standard color models used to describe all visible colors and presents information regarding luminance.13 Some studies have performed reliability tests when conducting traditional TI or applying a TI medical device.14-17 However, few studies

have addressed the intra-rater and inter-rater reliabilities of color parameters of the L*a*b* system using a popular digital camera and digital imaging software. Therefore the first purpose of our study was to calculate tongue L*, a*, and b* parameters using a digital camera and digital imaging software and to examine whether this color analysis method has satisfactory intra- and inter-rater reliabilities. It is also important to examine criterion validity, that is to say, to examine whether the color analysis method results using the L*a*b* system are consistent with those evaluated by experts.6 As such, the criterion validity of each color parameter was also examined, together with intra- and inter-rater reliabilities.

The second purpose of our study was to examine whether color parameters served as predictors of symptom patterns which is a subcategory of a disease or disorder in EAM. Although some studies have reported the relationship between color parameters of the TI and symptom patterns,7-9 few studies have addressed the association of color parameters of the tongue with validated pattern questionnaire measures. Recently, the Yin deficiency questionnaire (YDQ), the Phlegm pattern questionnaire (PPQ), and the Cold-Heat pattern questionnaire (CHPQ) were developed and validated.18-20

Together with the relationship between tongue color parameters and symptom patterns, we examined aging effect on the symptom patterns. It is generally accepted that the thickness of the tongue coating is related to the severity of Phlegm or Yin deficiency pattern.2 For example, tongue surface is tinged with white or yellow as Phlegm pattern develops, whereas it is tinged with red as Yin deficiency pattern develops.2 Park et al 21 recently reported that there were aging effects on the PPQ and YDQ scores in normal college students (n = 75). Therefore it may be important to examine whether aging effects on the Phlegm and Yin deficiency patterns can be replicated in patients, and to evaluate which among tongue color and age is better at determining symptom patterns.

In summary, in the present study, color parameters were calculated using digital images and color analysis software, and intra- and inter-rater reliabilities and criterion validity tests were conducted. Finally, the ability of the color parameters and aging for predicting symptom patterns was examined.

METHODS AND MEASURES

Subjects and data collection

This study was a chart review study, the entire process of which is depicted in Figure 1. The electronic medical records (EMRs) of all 200 outpatients that visited Women's Health Clinic of the Kyung Hee University Oriental Medical Hospital in Gangdong, Korea from April 2011 to February 2012 were

Figure 1 Study flow chart

RET: right edge of the tongue, TC: tongue center, TT: tip of the tongue, YDQ: Yin deficiency questionnaire, PPQ: Phlegm pattern questionnaire, CHPQ: Cold-Heat pattern questionnaire.

reviewed in this study. The female outpatients ranged in age from 15 to 74 years [(39 ±11) years], and their gynecological problems included menstrual dysfunction, infertility, climacteric syndrome, postoperative management, and post-labor management. Each patient was asked to complete the YDQ, PPQ, and CHPQ.18-20 In order to avoid time intervals between the recording of symptom patterns and determination of color parameters of the tongue, the questionnaires were completed and tongue image recordings were conducted on the same day for each patient. The Institutional Review Board of the Kyung Hee University Oriental Medical Hospital at Gangdong approved this chart review study protocol.

YDQ, PPQ, and CHPQ

The YDQ consists of ten Yin deficiency pattern-related items,18 the PPQ consists of 25 Phlegm pattern-related items,19 and The CHPQ consists of 10 Heat pattern-related and 10 Cold pattern-related items.20 The items on the YDQ, PPQ, and CHPQ are rated on a 5-point Likert scale: 1 = disagree strongly; 2 = disagree; 3 = neither agree nor disagree; 4 = agree; 5 =

agree strongly. The total scores of the YDQ, PPQ, and Cold and Heat subscales of the CHPQ were used to examine whether patterns of Yin deficiency, Phlegm, Cold, and Heat were related to the color parameters of the tongue.

Tongue image detection and analysis

In our study, tongue image detection and analysis were conducted following the protocol of Seo's study.14 The outpatients were seated on a chair in a room of normal brightness and asked to expose their tongues. When exposing their tongues, the subjects were asked to locate the white reference paper (10 x 15 cm) by the right edge of their tongue using their right hand (Figure 2). With the angle between the tongue and paper being parallel, the subjects' tongues were photographed using a digital camera (D70s, Nikon Co., Tokyo, Japan), and macro-lens (70 mm 1: 2.8 dg macro, Sigma Co., Kawasaki, Japan) at a resolution of 3000 x 2000 pixels. If the surface of the tongue is photographed using the direct light of the flash, the imaged color of the tongue may be affected by the saliva covering its surface. To minimize this bias,

Figure 2Two digital images of a female subject's tongue and the white reference color

A: before; B after correcting the white balance and exposure. The range of R, G, and B values before correcting for the white reference color was 191-196, and the range after correcting for the white reference color was 210-215.

indirect light, e.g., white ceiling-bounced flashlight (Speedlight SB-600, Nikon Co., Tokyo, Japan), was applied when photographing. After photographs were taken, the digital images saved in the compact flash memory were copied to the hard disk of a personal computer. Each digital image was imported to the Silkypix Developer Studio software (Ichikawa Soft Laboratory Co., Ltd., Chiba, Japan) to correct white balance and exposure variations, based on the white color of the reference paper. When correcting exposure, the red (R), green (G), and blue (B) parameters of the white reference paper were regulated within the levels of 210-215. However, the Silkypix Developer Studio presented the R, G, and B information of only a few pixel points, not for thousands of pixel regions. This would have resulted in sampling error, and therefore, each digital image was imported to the Picture Color Analyzer software to achieve the average R, G, and B values of thousands of pixel regions from the images.22 We cropped three regions of interest (ROIs) from the subjects' tongue images: the right edge of the tongue (RET), the tongue center (TC), and the tip of the tongue (TT). In terms of the TC, 40 000-45 000 pixels of ROIs covered with tongue coating were cropped. The RET and TT images were cropped at the size of 20 000-25 000 pixels without tongue coatings. Auto-calculating with the Picture Color Analyzer, the average R, G, and B values of the RET, TC, and TT ROIs were determined. Finally, the average R, G, and B values from the ROIs were transformed to L* [luminance], a* [balance between green (-) and red (+ )], and b* [balance between blue (-) and yellow (+ )] parameters using the 'color picker' tool in the Adobe Photoshop software (Adobe Systems Inc., San Jose, CA, USA).

Reliability and validity tests

Before conducting reliability and validity tests, a computer monitor was calibrated using Spyder Elite (Datacolor Co., Lawrenceville, NJ, USA). To assess inter-rater reliability, 20 samples among the 200 digital images were randomly selected. After 30 min of training in the use of the software programs, each rater was asked to crop the ROIs of RET and TC and to calculate the R, G, and B values of the ROIs using the

same monitor. Thereafter, the R, G, and B values were transformed to L*, a*, and b* values. Among these values, a* values were used to examine whether there were significant agreements between the raters. To assess intra-rater reliability, 20 digital images used to assess inter-reliability were copied, and a total of 40 digital images were randomly presented to the three raters. The three raters were asked to crop the ROIs of the RET and TC for the 40 samples and to calculate the L*, a*, and b* values. Among these values, a* values were used to examine whether there was significant agreement between the two sets of digital images for each rater.

To assess criterion validity, the three raters were asked to rate the redness of the tongue proper and the thickness of the tongue coating on a 100mm visual analogue scale (VAS) for 20 randomly selected digital images. Thereafter, the relationship between the redness of the tongue proper and the thickness of the tongue coating as evaluated by the raters and the a* values of the RET and TC were examined.

Statistical analyses

Intra-class correlation coefficients (ICCs) were calculated to assess intra-rater and inter-rater reliabilities. ICCs were considered to indicate excellent reliability when greater than 0.75, fair to good reliability between 0.40 and 0.75, and poor reliability when less than 0.40.23 Spearman's correlations were calculated to assess criterion validity. Thereafter, four hierarchical regression models consisting of three steps were formed to examine the associations of Yin deficiency, Phlegm, Cold, and Heat patterns with age and the L*, a*, and b* values of the RET, TC, and TT. In the hierarchical models, the total scores of the symptom patterns were used as dependent variables, and all of the L*, a*, and b* values, and age were used as independent variables. In the second step of the models, the L*, a*, and b* values of the three ROIs were used as independent variables. To examine collinearity of the independent variables, the tolerance and variance inflation factor (VIF) were calculated. In the regression models, a tolerance below 0.1 or a VIF above 10 was considered to determine collinearity between the independent variables.24 All statistical

analyses were performed with SPSS 15 for Windows (SPSS, Chicago, IL, USA). Values are presented as mean ± standard deviation ( x ± s). In all of the regression models using stepwise methods and Spearman's correlations, the significance threshold was set at P <0.05.

RESULTS

The mean values of the symptom patterns questionnaire and color parameters of the three ROIs

are listed in Table 1. Table 2 lists the intra- and inter-rater ICCs of the color parameters of the tongue. Intra-rater ICCs ranged from 0.648-0.972, and inter-rater ICCs ranged from 0.627-0.887. This indicated that the color parameters of the tongue had good or excellent reliability.23

Table 3 lists Spearman's correlations between the average VAS scales of redness and thickness as evaluated by the three raters and the a* values of the tongue. The evaluated redness of the tongue proper

[Table 1 Descriptive characteristics of questionnaire scores and color parameters of the tongue |

Questionnaire score or color parameter Mean±SD

Pathological pattern questionnaire PPQ (score) 77±21

YDQ (score) 25±7

Heat subscale of the CHPQ (score) 26±6

Cold subscale of the CHPQ (score) 58±10

Color parameter Right edge of the tongue L* 55±7

a* 25±5

b* 18±4

Tongue center L* 63±6

a* 16±4

b* 11±4

Tongue tip L* 48±7

a* 33±5

b* 20±4

Notes: PPQ: Phlegm pattern questionnaire; YDQ: K«-deficiency questionnaire; CHPQ: Cold-Heat pattern questionnaire; L*; luminance, a*; balance between green (-) and red (+), b*; balance between blue (-) and yellow (+).

Table 2 Intra- and inter-rater ICCs of color parameters (L*, a* and b*) of the tongue

Region

Color parameter

Intra-rater ICC

Inter-rater ICC

Rater 1 Rater 2 Rater 3

L* 0.969 0.922 0.830 0.735

a* 0.954 0.817 0.648 0.627

b* 0.926 0.897 0.701 0.748

L* 0.970 0.966 0.904 0.843

a* 0.960 0.862 0.773 0.764

b* 0.972 0.923 0.869 0.887

Right edge of the tongue

Tongue center

Notes: ICC: Intra-class correlation coefficient; L*: luminance; a*: balance between green (-) and red (+); b*: balance between blue (-) and yellow (+).

Table 3 Spearman's correlations between the average redness, thickness, and color parameter (a*) of the tongue

Region

Color parameter

Average VAS scale as evaluated by the three raters

Redness of the tongue proper

Thickness of the tongue coating

Right edge of the tongue a* 0.718 -

P-value <0.001 -

Tongue center a* - -0.523

P-value - 0.018

Notes: a*: balance between green (-) and red (+).

«4 > c

'm 1 Pi

no o o O O o ON t^ O 00 m to (N o t^ NO ON O o o

'"H '"H '"H ^ ^ ^ ^ '"H

ON ON o o o (N OA NO t^ m m ON t^ ON o o o

O O O C2 C2 c> o o ^

l/^ ON VD t^ l/^ O <N t^ (N CO NO OA m 00 m t^ t^ 00 ON l/A c>

O o O o c> c> C> c>

00 o (N l/^ o 00 o o 00 o t^ o C2 t^ ON C2 t^ o o m NO O m NO

Ö ö ö ö C2 Ö Ö c> c> Ö

OA o o o o m NO o 00 l/^ O O ON t^ m o o o o o

^ ^ ^ ^ ^ ^ ^ ^

00 t^ CA o o O CA ON Hf 00 o OA NO ON OA o NO OA o o o o

Ö Ö Ö C> Ö c> O '"H ^

OA °° o 00 00 l/A NO l/^ ON m NO ON oc^ m

o O o o c> c> C2 C2

o m ON m o o o c^ o NO o :: ::

o O O o c> c> o c>

NO o o O o o ON t^ O 00 m «A o t^ NO t^ ON O O o

^ ^ ^ ^ ^ ^ ^

OA o o O (N OA t^ NO t^ OA m m OA ON OA ON l/A o o C2

O O o O O

no o Ö

(N on o

NO o l/A l/A C2 NO 00 O o C2

c> C2 C2 C2 C2

m t^ 00 ON o t^ Csl l/^ o

^ ^ r^ ^

l/A ON l/A l/^ l/^ o 0^

o C2 C2 c>

00 ri t^ t^ o o OA o c>

C2 C2 C2 c>

o 00 c^ OA O t^ l/^ OA

C2 C2 c> c>

J2 ÜJ J2

had strong positive correlations with the a* values of the RET (r = 0.718). The thickness of the tongue coating as evaluated by the raters had moderate negative correlations with the a* values of the TC (r = — 0.523). As the tongue coating becomes thicker, the tongue proper becomes concealed, and the apparent redness of the tongue proper will decrease.25 The results indicated that the color calculating method proposed in this study had significant criterion validity.

Table 4 lists regression models for the total scores of the YDQ, PPQ, Cold, and Heat subscales of the CHPQ. In the YDQ regression model, age was predictive ofYDQscore (P = 0.192). This indicated that age should be considered when predicting the Yin deficiency pattern based on the color parameters of the tongue. The b* value of the TT was predictive of Yin deficiency, Phlegm, and Heat patterns (P =

— 0.212, — 0.172, and — 0.163, respectively). This indicated that the b* value among the color parameters was a determinant and robust predictor of symptom pattern and fatigue. This also indicated that the tongues of subjects with Yin deficiency, Phlegm, and Heat patterns tended to lean towards blueness. L* was predictive of Yin deficiency (P =

— 0.172) and Cold (P = 0.173) patterns. This indicated that the tongues of subjects with a Yin deficiency pattern tended to be dark, whereas those with a Cold pattern tended to be light.

DISCUSSION

In this study, color analysis of TI using three sets of software exhibited good or excellent levels of intra-rater (0.648-0.972 of) and inter-rater (0.627-0.887) ICCs. These results suggest that the L*a*b* parameters of the TI using a cost-effective color analysis method are both reliable and valid. We found that age was a significant predictor of YDQ score. Aging phenomena, including lowered cognizance and sexual function, are related to Yin or Yang deficiency in the kidney.2'26 Park et al21 reported the aging effects on the Phlegm and Yin deficiency patterns in normal young students. However, our study results showed that age had an on Yin deficiency alone, and that the predictive power of the aging effect was similar to that of tongue color parameters. Considering Park et al 21 and our study results, it appears that aging effect on the Yin deficiency pattern is robust and should be considered when predicting the Yin deficiency pattern for patients with gynecological problems as well as for the normal young population. Another finding of our study is that the blue of the TT was a determinant predictor of Yin deficiency, Phlegm, and Heat patterns. As the TT region is not covered with tongue coating, the determinant factor in TT color is blood supply condition. Blood supply

in the TT is carried out through the terminal branches of the lingual artery.27 The blood of the lingual capillaries, thereafter, departs from the tongue substance through the lingual veins and finally flows together to the internal jugular vein. Liu et al 28 previously studied the relationships between the hemodynamic characteristics of the TT region and different TT colors, using photophlethysmographic study. Liu's study27,28 results showed that peripheral blood flow rate in the TT was much lower in the blue tongue group than in the red tongue group, whereas the coefficient of the dicrotic notch, i.e., the peripheral vascular resistance-related parameter, in the TT was higher in the blue tongue color group than in the red tongue group. Liu's study28 also found that in the pale tongue color group, blood flow rate was lower than that in the red tongue color group, whereas no differences in the coefficient of the dicrotic notch between the pale and red tongue groups were found. Considering Liu's study and our study results, it is possible that in Yin deficiency, Phlegm, and Heat patterns, decreased blood flow rate and increased peripheral resistance in the terminal branches of the lingual arteries hindered oxyhemoglobin from flowing into the terminal branches of the lingual arteries, and at the same time, increased peripheral resistance in the TT region hindered the lingual veins from returning to the internal jugular vein. Finally, stagnated lingual venous flowing may have contributed to the shift of the TT color to 'blueness.' In the Cold pattern, blood flow rate decreased, whereas venous return was in the normal condition, and therefore low blood supply accompanied with normal venous return may have induced the color shift of RET to 'paleness.' Together with tongue color determinants, it is not clear which factor intervened with the regional differences in tongue color in the four symptom patterns, and therefore it would be challenging to examine the hemodynamic mechanism of the lingual arteries in inducing the regional differences in tongue color, according to symptom patterns. Criterion validity test results showed that the decrease in redness of the TC region was related to the increase in the thickness of the tongue coating, consistently with Han's study results.25 However, none of the TC color parameters were predictive of symptom patterns. For this reason, it appears that some subjects habitually brushed away the tongue coating when brushing their teeth, which may have resulted in the predictive bias of the color parameters of the TC on the PPQ score. Although our study results showed that color analysis of the tongue using the L*a*b* system was both reliable and valid, and that the color parameters served as predictors of symptom pattern, the data were entirely obtained from female subjects with gynecological problems. Therefore, further studies are needed to examine the effects of gender and disease differences.

In conclusion, we proposed a color analysis method using a mega-pixel digital camera and three sets of digital imaging software for the purpose of quantifying traditional TIs. Our study results of 200 female subjects and three raters showed that the color analysis method had significant reliability and criterion validity. In the four hierarchical regression models, age was a significant predictor of Yin deficiency, and the blueness of the TT was predictive of Yin deficiency, Phlegm, and Heat patterns. The dark TT and light RET regions were predictive of Yin deficiency and Cold patterns, respectively. Therefore it is concluded that color analysis of the tongue using the L*a*b* system is reliable and valid, and that color parameters partially serve as symptom pattern predictors.

REFERENCES

1 Anastasi JK, Currie LM, Kim GH. Understanding diagnostic reasoning in TCM practice: tongue diagnosis. AlternTher Health Med 2009; 15(3): 18-28.

2 Zou J, Tian D, Cong L, et al. Diagnostics of Traditional Chinese Medicine: inspection of the tongue. Shanghai: Shanghai University of Traditional Chinese Medicine, 1999: 102-153.

3 Lee BK, Park YB, Kim TH. Textbook of Oriental medicine: Tongue inspection. Seoul: Sungbosa Press, 1988: 70-90.

4 Chiu CC. A novel approach based on computerized image analysis for traditional Chinese medical diagnosis of the tongue. Comput Methods Programs Biomed 2000; 61 (2): 77-89.

5 Kim J, Jung Y, Park K, Park JW. A digital tongue imaging system for tongue coating evaluation in patients with oral

malodour. Oral Dis 2009; 15(8): 565-569.

6 Kim J, Han GJ, Choi BH, et al. Development of differential criteria on tongue coating thickness in tongue diagnosis. Complement Ther Med 2012; 20(5): 316-322.

7 Yamamoto S, Tsumura N, Nakaguchi T, et al. Principal component vector rotation of the tongue color spectrum to predict "Mibyou" (disease-oriented state). Int J Comput Assist Radiol Surg 2011; 6(2): 209-215.

8 Pang B, Zhang D, Li N, Wang K. Computerized tongue diagnosis based on Bayesian networks. IEEE Trans Biomed Eng 2004; 51(10): 1803-1810.

9 WatsujiT, Arita S, Shinohara S, Kitade T. Medical application of fuzzy theory to thediagnostic system of tingue ind-pection in Traditional Chinese Medicine. 1999 IEEE International Fuzzy Systems Conference Proceedings 1999; 1: 145-148.

10 Wee AG, Lindsey DT, Kuo S, Johnston WM. Color accuracy of commercial digital cameras for use in dentistry.

Dent Mater 2006; 22(6): 553-559.

11 Duangsang S, Tengtrisorn S. The central corneal light reflex ratio from photographs derived from a digital camera in young adults. J Med Assoc Thai. 2012; 95(5): 699-703.

12 Ratner D, Thomas CO, Bickers D. The uses of digital photography in dermatology. J Am Acad Dermatol 1999; 41(5): 749-756.

13 Park YJ, Park YB. Objective evaluation of the color of tongue substance using L*a*b* color coordinates. Orient Pharm Exp Med 2006; 6: 112-120.

14 Seo JH, Park YB, Park YJ. Reliable facial color analysis using a digital camera and its relationship with pathological patterns: a pilot study. Eur J Integr Med 2014; 6(3): 322-327.

15 Lagouvardos PE, Fougia AG, Diamantopoulou SA, Polyzois GL. Repeatability and interdevice reliability of two portable color selection devices in matching and measuring tooth color. J Prosthet Dent 2009; 101(1): 40-45.

16 Kim M, Cobbin D, Zaslawski C. Traditional Chinese Medicine tongue inspection: an examination of the inter-and intrapractitioner reliability for specific tongue characteristics. J Altern Complement Med 2008; 14(5): 527-536.

17 Lo LC, Chen YF, Chen WJ, Cheng TL, Chiang JY. The study on the agreement between automatic tongue diagnosis system and Traditional Chinese Medicine practitioners. Evid Based Complement Altern Med 2012; 2012: 505063.

18 Lee S, Park J, Lee H, Kim K. Development and validation of K«-Deficiency Questionnaire. Am J Chin Med 2007; 35: 11-20.

19 Park YJ, Park JS, Kim MY, Park YB. Development of a valid and reliable Phlegm pattern questionnaire. J Altern Complement Med 2011; 17(9): 851-858.

20 Ryu H, Lee H, Kim H, Kim J. Reliability and validity of Cold-Heat pattern questionnaire. J Altern Complement Med 2010; 16(6): 663-667.

21 ParkYJ, Park YB. Relationships between pathological patterns and vocal qualities estimated by acoustic parameters.

Chin J IntegrMed 2015; 21(5): 346-354.

22 Otaka I, Kumagai K, Inagaki Y, Shimoyama M, Saegusa K, Hara T. Simple and inexpensive software designed for the evaluation of color. Am J Opthalmol 2002; 133(1): 140-142.

23 Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979; 86(2): 420-428.

24 Han SS, Lee SC. Statistical analyses in the nursing and public health fields: Regression analysis. Seoul: Hyun-Moon Press, 2004: 323-359.

25 Zou J, Tian D, Cong L, Sun X, Wu G, Chen G (Eds). Diagnostics of Traditional Chinese Medicine: differentiation of syndromes according to pathologic changes of the viscera and their interactions. Shanghai: Shanghai University ofTraditional Chinese Medicine, 1999: 312-351.

26 Kimura Y, Ariji Y, Gotoh M, et al. Doppler sonography of the deep lingual artery. Acta Radiol 2001; 42(3): 306-311.

27 Liu WQ, Liu WH, Gu X, et al. Study of three color patterns and photoelectric blood volumetric graphs in tongue substance for healthy 182 subjects. TCM Res 2008; 20: 38-41.

28 Bolyle M, Butcher R. Haemodynamic monitoring learning package. http://www.aci.health.nsw.gov.au/_data/as-

sets/pdf_file/0008/221201/POW_SDL_Haemodynamic. pdf Last accessed, February 1, 2016.