運用貝氏模式建構慢性下背痛風險評估系統

Construction of Bayesian Low Back Pain Risk Assessment System

江鴻屏

指導教授 ： 詹前隆

元智大學/資訊學院/資訊管理學系/碩士(2004年)

摘要

資訊科技協助人們在工程或其它領域的需求，而在醫療領域，資訊科技的應用也愈來愈普遍，如資料探勘(Data Mining)和資料倉儲(Data Warehouse)等技術，都已逐漸在疾病的預測或分析上產生貢獻。關於慢性下背痛，雖然不像癌症會有致死的機會，但是對病患的生活品質影響卻非常鉅大，更影響工作能力，造成病患與社會整體的負擔(TBI，Total Burden of Illness)。因此了解慢性下背痛風險因子，從而改變健康行為，降低或消除風險因子，預防其發展成慢性下背痛是非常關鍵的。
　　本研究以問卷方式收集病人慢性下背痛風險因子資料，並在資料探勘的眾多分類技術(如決策樹、類神經網路與貝氏分類模式)中，選擇貝氏模式中的自然貝氏模式做為本研究之核心模式，應用於慢性下背痛上，來對問卷樣本資料進行分類的動作，在約九百筆的樣本下進行訓練學習的過程，得到的模式敏感度為0.762、鑑別率為0.837，而整體資料正確率為0.81，而ROC curve之圖形區域面積為0.879，這結果顯示出自然貝氏分類對於慢性下背痛的分類，具有相當不錯的成效。
　　本研究透過資料探勘中的自然貝氏分類模式，應用於慢性下背痛領域進行疾病預測，並建構一套Web-based之慢性下背痛風險評估系統，將問卷輸入線上自動化，使用者透過網路連結自行評估慢性下背痛之發生機率，並可動態選取風險因子，來了解下背痛機率的變化情形，來得知風險因子與慢性下背痛之關係，與改變行為之後的效益，及早改變生活或工作習慣，降低慢性下背痛的發生機率。而藉由網路多媒體的呈現方式，達到更佳的教育方式。

關鍵字

下背痛 ； 風險評估 ； 資料探勘 ； 自然貝氏分類法

並列摘要

The computational capability of information technology helps people in many areas. In medical domain, the application of information technology such as data mining or data warehousing becomes more and more general. And they are useful in illness forecast and analysis. Low back pain (LBP) is unlike cancer that will cause death. But LBP does influence patients’ quality of life greatly. This causes total burden of illness (TBI). To help patient understand the risk factors and help patient change their behavior. This is the key to reduce the probability of LBP occur.
　This study collects LBP risk-factor data in questionnaire and chooses naïve bayes classifier to be a classification on LBP from many data mining technologies (decision tree, neural network, Bayesian classifier). We collect 900 subject data and import these data to naïve bayes classifier in training process. The sensitivity of this model is 76%, and the specificity is 84%. The area of ROC curve is close to 0.88. This result shows that naïve bayes classifier is a good model in LBP risk assessment.
　This study uses naïve bayes classifier to evaluate the probability of LBP. And construct a web-based LBP risk evaluative system. User can input their risk data to evaluate their probability of LBP. This help user understand their risk factor and the benefit of changing behavior. They also can change their risk factor dynamically to know that if they change their behavior early they can reduce the probability of LBP occur. This web-based system provides a friendly interface to teach people understand the importance of LBP.

並列關鍵字

Low Back Pain(LBP) ； risk assessment ； data mining ； na&iuml ； ve bayes classifier

參考文獻

[1] Alcouffe J., Manillier P., Brehier M., Fabin C., Faupin F., “Analysis by sex of low back pain among workers from small companies in the Paris area: severity and occupational consequences.”, Occupational & Environment Medicine, 56(10), pp.696-701,1999

[2] Ashford J. R., “An Approach to the Analysis of Data for Semi-Quantal Responses in Biology Response”, Biometrics, 15, pp.573-581, 1959

[3] Bayes T., “An essay towards solving a problem in the doctrine of chances.”, Philos Trans Soc Lon, 53, pp.370-418, 1763

[4] Becker M., “The Health Belief Model and Personal health Behavior.”, Thorofare, New Jersey, Charles B. Slack, 1974

[5] Berkson J., “Application of the Logistic Function to Bioassay”, Journal of American Statistical Association, 39, pp.357-365, 1944