以服務導向架構提供基於支持向量機的醫療保健資訊分析

Abstract

近年來網際網路的普及以及Web技術的不斷演進，以網頁作為醫療服務的平台的概念在近年來逐漸成型，並且也出現了以網頁服務為平台的醫院資訊系統。另一方面，由於醫療的進步，很多新型的疾病被診斷出來；然而這些疾病與第一線的篩驗數值往往沒有很好的準則可以判斷，太鬆造成病人沒辦法及時得到治療，太緊又會浪費醫療資源做太多檢查。這種複雜的醫療決策正是人工智慧與資料探勘可以發揮的地方。 我們利用支援向量機的資料分類功能，對於串聯質譜儀分析出的新陳代謝物濃度數據做分析，評估新生兒是否罹患甲基丙二酸血症(MMA)。利用支援向量機方法，檢測的敏感度可由傳統Cutoff Value方法的76%提升到超過96%。除此之外，我們也在特徵資料進入SVM處理之前，以取對數值與加入乘法反元素的方式加以處理，配合不同的kernel進一步將正確率進一步提昇至99%以上。在實作方面，我們以服務導向的做法，將支援向量機的功能架構在分散式環境,，除了容易整合異質的平台之外，也提供了服務外延的便利性。

The concept, web as a platform, has been adopted in Hospital information systems successfully in National Taiwan University Hospital (NTUH). With the advances in medical technology, most diseases can be diagnosed, but the prices of some tests are still too expensive. Most interpretations of test results rely on cutoff values; consequently, the choice of a good cutoff value is important: a low cutoff value gives many false-positive, but a high cutoff value will let patients lose early treatment chances. For overcome this situation, we design, implement a newborn screening system using Support Vector Machine (SVM) classifications. . In this system, the predicting accuracy of MMA could be improved from 76% (cut-off value approach) to over 96%. Moreover, we have add enhancements for features: multiply inverse and logarithm before SVM, and the accuracy can be more than 99% in MMA dataset with some kernels. In addition, National Taiwan University Hospital Information System (NTUHIS) have been developed and implemented to integrate heterogeneous platforms, protocols, databases as well as applications. To expedite adapting the heterogeneities, we deploy Service-Oriented Architecture (SOA) concepts to the newborn screening system based on web services.