A New Tongue Diagnosis Method Based On Adaptive Outline Extraction And Multiple Features Synthesis

Traditional Chinese medicine has a long history, and it is the precious treasure of Chinese culture. But with the spreading of western medicine, its scientific basis and popularity is seriously challenged. Now its theoretical study and inheritance attracts more and more attention of scholars. Tongue is an important method of "look", one of four traditional Chinese medicine diagnosis methods. By observing the tongue, a physician can understand the inner condition of his patient. Due to its extensive experience needed for physicians, arcane theories and the lack of clinical evidence, tongue diagnosis is very restricted for application. In recent years, with the growing of country’s attention and support, the standardization of traditional Chinese medicine and its scientific theory study has been greatly developed, medical automation makes a big progress, and various types of medical equipment continue to emerge showing the continues expanding for Chinese market in this area. This thesis, going with tongue diagnosis standardization and automation, came up with a computer-assisted tongue diagnosis method. It combined the advanced lighting and image acquisition technology from optical science, mature image processing technology such as edge detection and blocking from pattern recognition field of computer science and tongue diagnosis method from medical field. By taking advantages of all combined fields, a relatively complete tongue diagnosis system with specialized and standardized has been built. By partnering with academic researchers from electric light major, we tried to build a scientific and complete front tongue image acquisition system. By using light with fixed lighting curve, high-precision lens with fixed parameters, standard color cards and other methods, we tried the best to ensure the accuracy and standardization of the acquired tongue image. With the study of various types of edge detection algorithm, we used a complementary binding method of edge detection algorithm with Sobel operator and Canny operator, and improved the method with introduction of the tongue color characteristics, so that the algorithm has strong adaptability and accuracy. And then we obtained a complete outline with single pixel by our special searching and connecting algorithm. The tongue outline extraction algorithm is proved to be effective with experiments. After getting tongue outline, we divided the image and obtained tongue block portion, and then we compiled statistics of tongue pixels in accordance with the characteristics observed in traditional tongue diagnosis and collected information such as color, quantity, etc., so that the features used in the diagnosis have been quantified and extracted. Then with experience of physicians as well as experiments, we found conversion method transforming quantified features into diagnostic characteristics which physicians can understand intuitively, and then got tongue classification and diagnosis. Experiments showed that the algorithm extracting features such as blade color, coating color, thick and humidity with relatively high accuracy. Finally, we also tried to put out method into practice. The first step of application software development for mobile platforms such as smart phone has been completed. It supports people to do self-health checking anywhere. Then we will proceed to specialized equipment and ancillary cloud server development, expecting to provide diagnosis service with equivalent level of special physician which can help to solve the problem of medical resources shortage.