- 學位論文
結合舌診與脈診系統參數在臨床之初步應用
The Clinical Preliminary Application of Combined Tongue with Pulse System parameter
摘要
中醫辨證以望、聞、問、切四診為依據。舌診與脈診分別是望診與切診中重要的一項，目前舌診與脈診皆在量化上有了些許成果，但始終是分開進行並未將兩者結合。本研究目的將嘗試結合舌診與脈診參數，提升疾病的辨識能力。 本研究藉由拍攝高血壓及上消化道病患的舌像，利用JSEG演算法根據顏色、紋理將舌頭分割成區塊狀並進一步辨識厚薄苔。利用脈診儀量取脈波和訊號處理的方式，從時域上擷取五個特徵點P角、DV、V點、Lw-P角/U角和Heart rate搭配Fuzzy Rule-Based System (FRBS)分辨常、細、澀、數、弦、遲、滑七種脈象。針對共15位健康人、64位高血壓及35位上消化道疾病觀察舌像與脈象。透過訓練組及測試組之安排進行系統之評估，並與病歷進行對照了解系統之Accuracy(Ac)、Sensitivity(Se)、Specificity(Sp)、kappa(K) 值。 結果顯示，高血壓患者在舌根部位厚黃苔及薄黃苔百分比大於健康人；30~50歲高血壓患者以弦脈居多，細弦脈次之，不同於健康人的常脈。50歲以上較不容易以弦、細弦脈辨識健康人與高血壓病患。單獨以FRBS舌診參數或脈象參數及結合兩組參數辨識其Ac、Se、Sp、K值分別為(0.83、0.79、1、0.58)、(0.98、0.98、1、0.93)及(0.98、0.98、1、0.93)。上消化道25例訓練組患者舌中薄白苔與厚白苔的面積明顯多於健康人且會有數脈或遲脈的脈象。單獨以FRBS舌診參數或脈象參數及結合兩組參數辨識其Ac、Se、Sp、K值分別為 (0.74、0.8、0.6、0.38)、(0.71、0.5、1、0.45)及(0.83、0.93、0.7、0.64) 。而針對13例上消化道測試組測試其結果為(0.85、1、0.6、0.66)。此外針對21例高血壓、35例上消化道和5例健康人進行混和測試，其兩者結合辨識結果為(0.75、1、0.59、0.54)。其中同時有高血壓及上消化道疾病6例能辨識出5例 本研究將舌診與脈診參數結合應用於高血壓與上消化道辨識，合併的Sensitivity與kappa coefficient都比個別辨識的結果高，此結果與原本目的一致，確實結合兩者系統參數可以使得辨識能力更精確更接近中醫診斷以「四診合參」的想法。
並列摘要
There are four major ways of diagnosis in Chinese medicine practice; tongue-view diagnosis and pulse diagnosis are one of the important roles in inspection and palpation diagnosis. Currently, some progress for tongue-view and pulse diagnosis in quantification, but do not combine with each other. The purpose of study tried to integrate tongue-view and pulse system parameter for raising diagnosis rate of disease. Hypertension and Upper GI disorder tongue image were segmented to piece and identified thickness or thin coating by JSEG algorithm based on color and texture. Pulse signal which was measured by pulse system was identified to 7 pulse type(Normal、Fine、Choppy、Hurried、Wiry、Slow and Slippery) based on P-angle、DV、V-position、Lw-p angle/U-angle and Heart rate which was got from time domain and by Fuzzy Rule-Based System(FRBS). Tongue image and pulse type were observed for 15 healthy persons、64 hypertension patients and 35 Upper GI disorder. In order to evaluate this developed system, those patients are separated as training group and test group, and then compared recognition result with patient record to obtain Accuracy(Ac), Sensitivity(Se), Specificity(Sp), kappa(K) values for this system. The results showed that the thickness and thin yellow coating of root tongue of Hypertension patients were larger then healthy persons; Between 30 and 50 years old, the Wiry was most and Fine-Wiry was second for Hypertension patients, not as healthy person in Normal. Moreover, System could not identify Hypertension patients above 50 years old by Wiry or Fine-Wiry. Performance values (Ac, Se, Sp, and K) of this system for using tongue system parameter or pulse system parameter only and both system parameters are (0.83, 0.79, 1, and 0.58) (0.98, 0.98, 1, and 0.93), and (0.98, 0.98, 1, 0.93), respectively. The thickness and thin white coating of middle tongue of 25 training group Upper GI disorder patients were larger than healthy persons. The pulse type was Hurried or Slow for Upper GI disorder patients. Performance values (Ac, Se, Sp, and K) of this system for using tongue system parameter or pulse system parameter only and both system parameters are (0.74, 0.8, 0.6, 0.38), (0.71, 0.5, 1, 0.45), and (0.83, 0.93, 0.7, 0.64), respectively. In Upper GI disorder test group (13 cases), the recognition result can reach (0.85, 1, 0.6, 0.66) when both parameters are used. Moreover, 21 Hypertension, 35 Upper GI disorder, and 5 healthy cases were put together to test. The performance values were (0.75, 1, 0.59, and 0.54). 5 patients with Hypertension and Upper GI disorder could be identified from 6 cases. The parameter of tongue-view system and pulse system was combined to identify Hypertension and Upper GI disorder in this study. The result shows the performance is better when compared with system used tongue or pulse parameter only. The result agreed with the assumption of study that the recognition rate could be more accurate and closer to the principles of traditional Chinese medicine diagnosis that were based on inspection, auscultation, questioning and palpation.