蔣依吾 國立中山大學資訊工程系

主持人於執行衛生署中醫藥委員會90至101年度之中醫舌診標準化整合研究計劃（CCMP90-RD-017, 90.06.01~ 90.12.31、CCMP91-RD-110, 91.01.01~ 91.12.31、 CCMP92-RD-102, 92.01.01~ 92.12.31、CCMP98-RD-031, 98.01.21~99.12.31、CCMP99-RD-105, 99.11.16~100.11.20、 CCMP101-RD-026, 101-06-01~101-12-15）中，業已發展一可自動分析舌部影像特徵之中醫舌診電腦化系統，並經中醫藥委員會申請專利，於2008年4月獲准發明專利（"舌診系統及其方法"，中華民國專利證書發明第I296110號），復於98、99年推動中醫藥療效評估及診斷標準化類研究計畫(CCMP99-RD-037,99.11.18~100.11.20)中進行自動化舌診系統與中醫師診斷一致性研究，證實中醫舌診電腦化系統具高度自身一致性及符合舌診專家判讀之外部一致性，可減少或消除傳統診斷上環境因素對舌診影響，對舌象特性作定性及定量分析，提供客觀之診斷標準。確立自動化舌診系統一致性後，於101年執行中醫診斷術語標準化研究(CCMP101-RD-026, 101-06-01~101-12-15)使用自動化舌診系統進行對特定疾病-乳癌之採樣分析研究，藉以觀察乳癌患者所具舌象進一步定訂判讀指標。本兩年期計畫於第一期計畫結束後，已共收集乳癌病患60例，正常人70例，透過自動化舌診系統進行對乳癌病患與正常人之舌象特徵研究與比較，其中兩項中重要成果為：(1)找出乳癌病患與正常人舌象特徵顯著差異部份，發現兩者間以津液面積(p=0.018)、舌苔肝膽右區(p=0.025)、舌苔腎區(p=0.000*)、舌苔厚薄(p=0.000*)、舌質肝膽右區(p=0.025)、舌質腎區(p=0.000*)、齒痕數量(p=0.019)、齒痕最小面積(p=0.011)、齒痕肝膽右區數量(p=0.000*)、齒痕心肺區數量(p=0.023)、朱點數量(p=0.009)、朱點最大面積(p=0.000*)、朱點肝膽右區個數(p=0.001)、朱點心肺區個數(p=0.000*)皆達顯著水準，代表上述舌象特徵表現會因有否罹患乳癌而有數據上之差別，可進一步藉此鑑別乳癌患者與正常人。(2)以Logistic Regression導出預測模型，其中舌苔脾胃(p=0.013)、舌苔最大面積(p=0.006)、舌頭整體長寬比(p=0.021)、齒痕肝膽左區數量(p=0.027)、朱點脾胃區個數(p=0.034)為影響發腫瘤病部位之因子，並且能以因子影響程度預測腫瘤發病於胸部左側或右側。延續上期計畫成果，為加強資料參考價值與準確度，本期計畫首要工作為擴大收集乳癌病患人數，經與院方合作預期對乳癌患者再增加100筆樣本數，對新獲取舌頭影像進行自動化舌診分析，並與先前資料整合分類。其後，採取統計方法，依院方提供各項西醫檢驗資訊(病理資訊有期數、發病部位、癌細胞有否轉移，驗血檢測項目則有CEA、CA153、Cr、GOT、GPT、WBC、PTL、Hb、Hct)比對舌象特徵，個別找出可鑑別之顯著差異項目，並從顯著差異部份進行預測模型建立，以便自舌象特徵經預測模型分析結果據以評估患者病況。將其評估結果與西醫檢驗資訊相比較，觀察是否有相違處，從而驗證預測模型之穩定性與準確率。本計畫旨在於以電腦科學化分析方式，藉由大量數據進行定量分析，歸納出乳癌患者舌象特徵與病況判讀標準，使中醫舌診於乳癌疾病有客觀化之認識，以達確立乳癌預後之中醫舌診指標目的。

關鍵字：自動化舌診系統，乳癌。

The TCM indices of tongue diagnosis for the prognosis of breast cancer(2-2)

John Y. Chiang Department of Computer Science and Engineering National Sun Yat-sen University

In the Tongue standardized integration plan in TCM from 90 to 101（CCMP90-RD-017, 90.06.01~90.12.31、CCMP91-RD-110, 91.01.01~91.12.31、CCMP92-RD-102, 92.01.01~92.12.31、CCMP98-RD-031, 98.01.21~99.12.31、CCMP99-RD-105, 99.11.16~100.11.20、CCMP101-RD-026, 101-06-01~101-12-15），it has developt an TCM Tongue computerized system which analyzes the tongue characteristic automatically, and got the patent application in April,2008("the tongue diagnosis system and method",ROC patent certificate invention 1296110 number), and in the chinese medicine assessment and diagnosis standardized plan(CCMP99-RD-037,99.11.18~100.11.20), we did the research about The Study on the Agreement between Automatic Tongue Diagnosis System and Traditional Chinese Medicine Practitioners . We prove that the results of this study lead us to conclude that ATDS is effective in preventing influence of external factors and can provide TCM practitioners with objective and precise diagnostic data. After we prove the high Agreement between Automatic Tongue Diagnosis System and Traditional Chinese Medicine Practitioners, the implementation of medical diagnostic terminology standardization in 101 uses automated tongue diagnosis system to analyze the specific diseases - breast cancer so that we can observe the tongue image situation of breast cancer patients to set the judgment index. The first-years plan has collected 60 breast cancer patients and 70 normal people. We do the research about breast cancer patients and normal people through the automated tongue diagnosis system, The two important results are that (1) We found the outstanding differences of tongue image situation between breast cancer patient and normal people. The main differences are body fluid area (p = 0.018)、right area of Tongue hepatobiliary (p = 0.025), the kidney area of tongue(p = 0.000*), the tongue coating thickness (p = 0.000*), the hepatobiliary right area of tongue (p = 0.025), kidney area of tongue quality (p = 0.000*), the scalloped quantity (p = 0.019), scalloped minimum area (p = 0.011), number of right area of scalloped hepatobiliary (p = 0.000*), the number of scalloped cardiopulmonary area (p = 0.023), number of Zhu points (p = 0.009), Zhu points maximum area (p = 0.000*), the number of right zone of hepatobiliary of Zhu points (p = 0.001), Zhu point cardiopulmonary zone number (p = 0.000*). The above tongue characteristics mean the situations that breast cancer occurred, and the information would identify the breast cancer patients and normal people.(2) We use the Logistic Regression to predict model. The factors that would lead to the disease place are Tongue stomach(p = 0.013) , the the tongue maximum area (p = 0.006), the tongue overall aspect ratio (p = 0.021), number of scalloped hepatobiliary left (p = 0.027), Zhu points stomach characteristics of the zone number (p = 0.034). Moreover, we can use the degree of the factors to predict the place that disease occurred.( the left breast or the right breast)Following the results of last plan, in order to increase the value and accuracy of information, we will collect more breast cancer people in this plan. With the help of the hospital, we predict that we will increase 100 breast cancer people. By using the automated tongue diagnosis system, we combine and classify with the last data. Later, we compare the characteristic of tongue image situation according to the western medicine information(pathological information prediction、cancer place、Cancer cell metastasis and the blood test of CEA, CA153, Cr, GOT The GPT, WBC, PTL, Hb, Hct) by using the Statistics methods. We find the mainly outstanding difference and establish the predicting model from the mainly outstanding difference so that the information would show the patients situation due to the analysis of predicting model. After that, we compare the analysis results with the western medicine information and observe the truth to prove the stability and accuracy. The project uses the computerized science analysis way to sum the characteristic of tongue image situation of breast cancer patients and the standards of disease from the huge amount of data and analysis so that it is objective for TCM tongue diagnosis and breast cancer disease to reach the goal of The TCM indices of tongue diagnosis for the prognosis of breast cancer.