Research On Diagnostic Classification For Fissured Tongue Images In Computerized Tongue Diagnosis

Tongue diagnosis is one of the most valuable and widely used diagnostic methods in traditional Chinese medicine (TCM). However, due to its qualitative, subjective and experience-based nature, traditional tongue diagnosis has a very limited application in clinical medicine. Some researchers do automatically and objectively research on tongue diagnosis by using computerized technology.Despite considerable progress in the computerized tongue diagnosis, few are about tongue crack which is an important feature of tongue states whereas most researches focuse mainly on analysis for tongue coat and tongue proper. Even though there are few reports about feature extraction and analysis for tongue crack in computerized tongue diagnosis, these works are short of system and are shallower and at present there are hardly any reports about mapping from quantitative features of tongue crack to clinic disease.Tongue crack is an important feature of tongue states and it is obvious fissure with varied amounts, depths and shapes on tongue surface. Change of tongue crack states does not only reflect objectively and accurately changed circumstance of some typical diseases and TCM syndrome but also it can combine the other fearures of tongue states to increase higher system classification accuracy. Due to lower color contrast between tongue crack region and its surrounding tissues and the complexity of color distribution of tissues surrounding tongue cracks on color fissured tongue images, it is very hard and challenging for region extraction and feature analysis of tongue crack. So far these problems emerged in computerized tongue crack diagnosis have restricted badly use of tongue crack in computerized tongue diagnosis.To overcome these problems emerged in computerized tongue crack diagnosis, this dissertation does systematic and in-depth researches about these topics such as tongue crack enhancement, tongue crack region extraction, tongue crack feature description and tongue crack diagnostic classification oriented to diseases and TCM syndrome. Moreover aiming at these topics this dissertation proposes a series of settlements and schemes . Main contents of this dissertation are as follows:First, to change the actuality which doctors are unwilling and unable to see tongue crack in clinic, this dissertation proposes color transformation based on algebraic feature of three order unit matrix for color fissured tongue image. This algorithmic makes apparent color contrast between tongue crack and its surrounding tissues so that it can assist clinic doctors to diagnose fissured tongue image. To extract correctly and completely region of tongue crack this dissertation proposes detection of tongue crack based on distant gradient and prior knowledge. This algorithmic uses information of pixel color and gray change fully. This algorithmic of tongue crack region extraction offers credible technology which benefits computerized researches of tongue crack.Second, this dissertation proposes algorithm of diagnostic classification for physiological and pathological fissured tongue images. Firstly this dissertation extracts shape feature of tongue crack region using a method of extracting shape feature of tongue crack region based on technology of fractal geometry. Then this dissertation computes mean and standard deviation of three components of tongue image in RGB color space within tongue crack region and strings them to form a vector as color-texture feature of tongue crack. Finally this dissertation fuses shape feature of tongue crack region and color-texture feature of tongue crack to form tongue crack feature. Use tongue crack feature and linear support vector machine classifier to classify diagnostically physiological and pathological fissured tongue images. Experimental results demonstrate sensitivities of diagnosis and classification for physiological and pathological fissured tongue images are over 90% by using the approach proposed in this dissertation.Third, this dissertation proposes a diagnostic model of tongue crack and under the guidance of the model this dissertation proposes algorithm of diagnostic classification for fissured tongue images of moderate-severe chronic superficial gastritis, chronic atrophic gastritis and gastric ulcer. The algorithm extracts tongue state feature by using Gabor wavelet transformation and springy discriminant analysis and takes linear support vector machine classifier to classify clinically these three kinds of fissured tongue images. Experimental results demonstrate sensitivities of diagnosis and classification for these three kinds of fissured tongue images are over 80% by using the approach proposed in this dissertation.Finally, this dissertation proposes algorithm of computing pixel strength and gets strength image using it. This dissertation forms a gray-strength co-occurrence matrix and gets feature of tongue states by using the gray-strength co-occurrence matrix. Then this dissertation takes linear support vector machine classifier to classify clinically fissured tongue images diagnosed as yin deficiency syndrome and qi-blood deficiency syndrome according to TCM theory. Experimental results demonstrate sensitivities of diagnosis and classification for fissured tongue images diagnosed as yin deficiency syndrome and syndrome of qi-blood deficiency are over 85% by using the approach proposed in this dissertation.Even though TCM tongue diagnosis was researched by using computerized technology in 1980's for the first time, people did not use tongue crack in computerized tongue diagnosis until the initial stages of 21 century. Up to now, there are only few papers about the use of tongue crack in computerized tongue diagnosis, and most of them are incomplete and not systematically. Thus, in-depth study is necessary. This dissertation fills up some blank areas about tongue crack in computerized tongue diagnosis to a certain extent. Research fruits of this dissertation would accelerate practical research on tongue crack in computerized tongue diagnosis and also possess potential application in medical clinic practice simultaneously.