利用文件及影像檢索建立胃癌診斷與治療的案例式推理

Building a Model of Case-Based Reasoning for Diagnosis and Treatment of Gastric Carcinoma using Documents Mining and Image Retrieval

楊超然(Chao-Jan Yang)

指導教授 ： 劉立

臺北醫學大學/醫學院/醫學資訊研究所/碩士(2004年)

摘要

這些年來，雖然醫療科技與生物資訊有相當大的突破與進步，但在目前，癌症仍然是對人類健康與生命的最大威脅。是以如何建立一個能早期診斷癌症的機制，並將癌症的治療納入流程，利用前人之經驗加上新的研究結果來早期發現病患，並訂定治療計劃，以提升病患的治癒率，實為臨床工作者所必需孜孜不倦持續努力的方向。隨著資訊科技的快速發展，現代的醫療照護藉助科技的大力協助，獲得了空前未有的進展。高品質的醫療照護，除了基本的高品質醫療技術及醫療人員的愛心之外，高度複雜的病患資訊管理，已成為不可或缺的一部份。所以如何運用數位化的醫學資訊來幫助臨床工作者解決病患的問題，在目前以人為本的社會中，是相當重要的。在人工智慧的領域裡，在人工智慧的領域裡，案例式推理(Case-Based Reasoning, CBR)是一種解決問題(problem solving)的技術，藉由以前的經驗，來解決目前所遭遇到的問題。當問題領域中有清楚、簡明的知識表達；或案例的內容複雜、不易分割，及與經驗有關、重複性高的情況，案例式推理特別能發揮其功效。而在目前的案例式推理系統中，大部份都只有針對文件資料的部份進行。然而在醫學的領域中，影像資料所能提供的價值，遠超過其他的專業領域，若使文件與影像資料可合而為一，則對整體診斷成效會更有所助益。當然，新的醫療技術隨著時間也會有新的發展，更新的檢驗方法或技術也將會不斷的出現，如何將更新更準確的方法納入案例式推理中以增加系統的準確度，並減低因加入新索引而導致案例式推理系統重整所需花費的時間。故在系統中仍保留人工加入索引的機制，使有經驗的臨床工作者可以加入自己的想法及隨時加入新且更精確的索引，以保持整個系統的可塑性及活力。進而使病患可以在早期經由症狀與初步的胃鏡檢查結果，即可早期診斷出胃癌的存在，並由此及早訂定治療的計劃，及時治療；並可藉由以往診療的經驗，建立最適合病患治療的流程，以期降低醫療成本，並提高病患的存活率。

關鍵字

胃癌 ； 案例式推理文件探勘 ； 影像檢索

並列摘要

Although medical and biotechnology technology have made a great breakthrough and improved a lot in recent years, cancer is still the one of the major threats to mankind’s health and survival. According to the Cancer Registry Annual Report by the Department of Health, Executive Yuan, Taiwan, R.O.C., 56,323 people were attacked by cancer in 2002. On average, one got cancer every 9 minutes 20 seconds and people dying of gastric cancer amounted to 2,446. Gastric cancer ranks fifth of the 10 causes of cancer death. With early diagnosis and treatment, gastric cancer patients can mostly be cured and 95% of them will live for five years or more. If the gastric cancer is found in the later phase, however, the cure rate is almost zero. In the AI (Artificial Intelligence) field, Case-Based Reasoning (CBR) is a technology for problem solving. People can solve their current problems based on their previous experiences. Most of the existing CBR systems are applied to the processing of documented data., but in the medical field, image data can offer better values than other professional fields. Gastric cancer can be diagnosed not only from the clinical symptoms but also from gastroscopy images;therefore, the Study tries to bring the gastroscopy images into indexes and images and find out the most suitable and presentable attribute weight values. Of course, new medical technology will develop as time passes by and updated examination approaches or technologies will be discovered from time to time. It is important to know how to bring new and accurate approaches into CBR to increase the systematic accuracy and decrease the time you spend relaying CBR due to added indexes. The manual mechanism can still operate for index addition, and experienced clinical workers can add their own thoughts and new and accurate indexes any time to retain plasticity and vitality for the whole system. Furthermore, medical staff can also create a cure flow path suitable for patients to reduce the medical costs and increase the survival rate for the patients based on their medical experiences.

並列關鍵字

Gastric carcinoma ； Texting mining ； Content base Image Retrieval ； Case-Based reasoning

參考文獻

許懷仁，〈生物醫學文件探勘系統之架構設計與實作〉，國立成功大學資訊工程學系，碩士論文，民91.6。

S. Champati, W.F. Lu and A.C. Lin, “Automated Operation Sequencing in Intelligent Process Planning: A Case-Based Reasoning Approach＂, International Journal of Advanced Manufacturing Technology, Vol.12, No.1, pp.21-36, 1996.

Montazemi, A.R. and K.M. Gupta (1996), “An Adaptive Agent for Case Description in Diagnostic CBR Systems, “ Computers in Industry, Vol. 29, pp.209-224.

Kolodner, J. (1993), Case-Based Reasoning, Morgan Kaufmann Publishers, San Mateo. John, H., Dennis, R. and Flaura, K., “The Learning Classifier System: an Evolutionary Computation Approach to Knowledge Discovery in Epidemiologic Surveillance,＂ Intelligence in Medicine, Vol.19, pp.53-74, 2000.

Kononenko, I., “Machine Learning for Medical Diagnosis: History, State of the Art and Perspective,＂ Artificial Intelligence in Medicine, Vol. 23, Issue: 1, pp. 89-109, August 2001.