Published in A R DIGITECH

Transcription

1 MEDICAL DIAGNOSIS USING TONGUE COLOR ANALYSIS Shivai A. Aher*1, Vaibhav V. Dixit*2 *1(M.E. Student, Department of E&TC, Sinhgad College of Engineering, Pune Maharashtra) *2(Professor, Department of E&TC, Sinhgad College of Engineering, Pune, Maharashtra) Abstract The human tongue plays an important role in detecting various diseases. The concept of detecting the disease by observing the tongue is well known in traditional Ayurveda as well as Chinese medicine. Features of the tongue if extracted based on its color, texture and geometry help in knowing the disease. Color feature extraction is one of the most important parameter useful in diagnosing diseases. The proposed method extracts the color features of the tongue using the color gamut in which 12 colors represent features of the tongue. Keywords Color feature extraction, tongue color gamut. I. INTRODUCTION The tongue is considered to be the mirror of viscera. It can be used for the diagnosis of disease with its color feature to be of utmost importance. Since ancient times, Ayurveda as well as Chinese medical practitioners have examined various features of the tongue based on its color, coating, texture and shape. But diagnosing the correct disease requires years of experience and is fully of uncertainty. In accordance with Ayurveda [6] the front i.e. the tip one-third of the tongue relates to lungs, heart, chest and neck. The central area relates to liver, spleen, stomach and pancreas. The rear one-third of the tongue i.e. the base relates to the lower abdominal organs small intestine and the colon. Any disorder in the organs corresponding to that area is reflected by discoloration or sensitivity in that area. In the Chinese context [3] the tongue is divided into its tip, margins, center and root. The tongue tip reflect changes in the heart and lungs, margins reflect changes in the liver and the gallbladder. Spleen and stomach are related by the center of tongue while kidneys, intestines and bladder correspond to the tongue root. Hence the tongue depicts the overall health of an individual and color features play a very important role in determining it. The proposed method therefore extracts the color features of the tongue systematically. II. RELATED WORK In Fatemah et. al. [1] color feature extraction is performed using the dynamic color distribution entropy of neighborhoods (D-CDEN) which effectively describes the spatial information of colors. D-CDEN takes account of the image content by attending to neighborhoods of pixel for every color bin, instead of drawing concentric circles. This method is employed mainly to retrieve an image. Color feature extraction done by S. R. Kodituwakku et. al. [2] deals with the comparison of individual color descriptor and also the comparison between combinations of color descriptor. Color histogram, color moments and color coherent vector (CCV) are considered for retrieval. The retrieval performance of 1

2 the descriptors are measured. Combination of color descriptors show better retrieval rate compared to individual descriptors. Automated feature extraction method by Ratchadaporn et. al. [3] deals with the color-space based feature extraction of tongue for clinical classification of ZHENG (Traditional Chinese medicine syndrome) using supervised machinelearning algorithm. Machine-learning techniques establish a relation between tongue features and ZHENG. In the proposed method the foreground pixel of the tongue are extracted and then assigned to the 12 colors representing the gamut. It gives a better representation of the tongue color features with human health which is described as follows. III. PROPOSED METHOD The tongue color gamut is first explained and further every foreground pixel of the tongue is compared with the 12 colors in the gamut. The color closest to the pixel of the tongue is assigned with the color. This makes up for the color features of the tongue. A. Tongue Color Gamut After plotting the foreground pixels it was seen that the majority of tongue pixels lie inside the black boundary of the chromaticity diagram. For better representation of the color gamut the 12 colors are plotted using RGB color space. Figure 1: Tongue color gamut representation using CIE 1931 Color Chromaticity Diagram. [5] The tongue color gamut [4] is a set of all the colors which appear on the human tongue. The 12 colors where selected by plotting each foreground pixel in the tongue image onto the CIE 1931 Color Chromaticity diagram. The Chromaticity diagram represent all the possible colors in the visible spectrum. Figure 2: Twelve colors in the Tongue color gamut with label. [4] In Figure 1 RGB points are plotted and connected using lines which forms a triangle. On the RG line there is a point Y (Yellow) marked. The RB line consists the point P (Purple) and similarly C (Cyan) 2

3 is marked on line GB. For the RGB color space the center is denoted by W (White). From each of R (Red), B (Blue), Y (Yellow), P (Purple), C (Cyan) a straight line is drawn to W (White). The points at which these lines intersect the tongue color gamut form a new color to be added to denote the 12 colors. This results in the addition of LR (Light Red), LP (Light Purple), LB (Light Blue) which are the midpoints between the lines from the black boundary to the center W (White). DR (Dark Red) is chosen as no other point occupies that region. The colors GY (Gray) and BK (Black) belong to grayscale hence they are not mentioned in Figure 1. B. Color Feature Extraction In the feature extraction process the RGB [5] values of foreground pixels of the tongue image need to be found out. Once the RGB values are known the CIELAB coordinates can be calculated by transferring RGB coordinates to CIEXYZ coordinates. The conversion is given below: Color [R G B] [L A B] C (Cyan) [ ] [ ] R (Red) [ ] [ ] B (Blue) [ ] [ ] P (Purple) [ ] [ ] DR (Dark red) [ ] [ ] LR (Light red) [ ] [ ] LP (Light purple) [ ] [ ] LB (Light blue) [ ] [ ] The CIEXYZ are converted to CIELAB via where (1) (2) X0, Y0, and Z0 terms mentioned in (2) represent the CIEXYZ tristimulus values of the reference white point. The 12 colors from the tongue color gamut are then compared with the LAB values and are further assigned with the color which is closest to it using the Euclidian distance. Once the foreground pixels of the tongue are evaluated, the total of each color is added up and then divided by the total number of pixels. Similar ratio is calculated for all the 12 colors and these together form the tongue color feature vector. The color feature vector is given by v, where v = [c 1, c 2, c 3, c 4, c 5, c 6, c 7, c 8, c 9, c 10, c 11, c 12 ] and c i stands for sequence of color. The color features of the tongue are thus extracted, where the original image is divided into its 12 color components. BK (Black) [ ] [ ] GY (Gray) [ ] [ ] W (White) [ ] [ ] Y (Yellow) [ ] [ ] IV. EXPERIMENT AND DISCUSSION Table 1. RGB and CIELAB co-ordinates for the 12 colors in tongue color gamut 3

4 V. Acknowledgement I give a sincere vote of thanks to my guide for having faith in me and providing me with all the necessary infrastructure and support. I would thank our Principal for giving me the opportunity and lastly my friends and family who have helped me keep my spirits high. REFERENCES [1] Fatemeh Alamdar, Mohammad Reza Keyvanpour, A New Color Feature Extraction Method Based on Dynamic Color Distribution Entropy of Neighborhoods, IJCSI, Vol. 8, Issue 5, No.1, Sept 2011 [2] S. R. Kondituwakku, S. Selvarajah, Comparison of Color Features for Image Retrieval Indian Journal of Computer Science and Engineering, Vol. 1,No. 3, [3] Ratchadaporn et. al., Automated Tongue Feature Extraction for ZHENG Classification in Traditional Chinese Medicine, Research article, Hindawi Corporation, Vol. 2012, Article ID , Mar 2012 CONCLUSION The proposed method extracts the color features of tongue which further can be used for detection ofdiseases such as Diabetes, kidney disorders, nephritis etc. It is easy to understand and efficient method with significantly lesser computations. All the 12 colors in the color gamut are successfully extracted. [4] Bob Zhang, V. K. Vijaya Kumar, David Zhang, Detecting Diabetes Mellitus and Non-proliferative Diabetic Retinopathy using Tongue Color, Texture and Geometry Features, IEEE Trans. On Biomedical Engg., Vol. 61, No. 2, Feb 2014 [5] Bob Zhang et. al., Tongue Color Analysis for Medical Application, Research article, Hindawi Corporation, Vol. 2013, Article ID , Mar 2013 [6] 4

5 Figure 3: (a) Original Image, (b)-(m) Tongue features extracted in accordance with 12 colors in Tongue color gamut. 5