舌診影像系統在疾病辨識之應用

The Application of Tongue Viewing System for Diseases Recognition

洪國峰(Kuo-Feng Hong) ； 廖偉丞(Wei-Cheng Liao) ； 翁清松(Ching-Sung Weng) ； 胡威志(Wei-Chih Hu) ； 蘇振隆(Jenn-Lung Su)

《Journal of Medical and Biological Engineering》 22卷S期 (2002/12) Pp. s53-s59

摘要

本文利用影像處理的技術，建立了一套舌診影像系統。針對31名子宮肌瘤病患、32名不孕症病患、31位中風病患、55位上消化道患者及33位健康人的舌影像進行舌特徵之分析。由分析結果得知，子宮肌瘤患者的淡紅舌比例較健康人低，且患者的舌左區與舌右區的溫度有明顯差異；不孕症患者的舌面積明顯比健康人來的大，且患者的整體舌溫比正常來得低；中風患者的整體舌苔百分比大於健康人，且病患的舌尖區溫度明顯比健康人低；上消化道病症患者其脾胃區之舌苔所佔的比例明顯地多於健康人，且病患之脾胃區溫度比健康人低。子宮肌瘤、不孕症、中風、上消化道疾病之統計辨識率分別為94%、94%、95%、85%；若以類神經網路來診斷，則為94%、94%、97%、93%，比統計的效果更佳。以ANOVA來分類，可將健康人與上消化道病患與中風病患區分出來，至於子宮肌瘤患者與不孕症患者並不能分類出來。

關鍵字

紅外線 ； 舌診 ； 影像處理 ； 舌特徵資料庫

並列摘要

A comprehensive tongue's image feature extracting system was presented in this paper. The system is designed to examine various symptoms or dysfunction that will register some specific characteristics at tongue. We have tested the system on 31 Myoma Uteri patients, 32 Infecundity patients, 31 Apoplexy patients, 55 Upper GI disorder patients, and 33 healthy persons. The results showed that Myoma Uteri patients which have smaller percentage of light red area on tongue than that of normal healthy persons. And, compared to normal healthy persons, the Myoma Uteri patients are showing a significant temperature difference between the right side and left side of tongue. The surface area of tongue for the Infecundity patients is significantly larger and the temperature over the tongue is lower than that of normal healthy persons. For Apoplexy patients, the percentage of coating area on the tongue is significantly larger and the temperature of tongue tip is lower than that of normal healthy person. For Upper GI disorder patients, the coating at the middle region of the tongue was larger and the temperature was lower than that of normal healthy person. Moreover, the true positive categorization of tested patient groups for the normal group, Myoma Uteri, Infecundity, Apoplexy, and Upper GI disorder patients groups are 94%, 94%, 95% and 85%, respectively. However, only normal group, Upper GI disorder patients groups, and Apoplexy can be identified thru the ANOVA tests. Furthermore, the correctness of using trained neural network to categorize the above four groups are 94%, 94%, 97%, and 93%, respectively.

並列關鍵字

Infrared ； Tongue viewing system ； Image processing ； Tongue-characteristic database