基於支援向量機之生醫訊號分析系統

Abstract

醫療資訊越來越發達的今日，許多生醫訊號都是以數位的形式被儲存下來。然而隨著生醫訊號資料量的遽增，我們需要有一個方便使用且準確率高的生醫資訊系統，從中找尋出有用的資訊，供醫生作為診斷參考。 在本論文，我們設計一個以支援向量機為主的生醫訊號分析系統。我們利用支援向量機優異的資料分類功能，對於儀器所產生的各類生醫訊號做預測，評估受測者是否罹患疾病。在服務導向架構SOA下，我們將支援向量機的功能架構在分散式環境，利用中介軟體Web Services 整合異質的平台、服務與資料庫。利用此系統來預測受測者是否罹患癲癇等疾病，並將疑似有病患者的生醫訊號顯示供醫生做進一步的確認。目前系統對於癲癇病患的腦波訊號判斷準確度達98.73%。

As medical information technique advancing, many biosignals are recorded in digital form, generated dramatic amount of data as well. In order to extract useful diagnostic references from the biosignals, it is essential to establish a user-friendly and high accuracy biomedical information system for doctors. In the study, we design a Biosignal Analysis System based on Support Vector Machine (SVM). To predict various types of biosignals, e.g., Electroencephalogram (EEG) for epilepsy, the SVM approach is an appropriate data classifier to evaluate the disease. We implement the System by applying Service-Oriented Architecture (SOA), Web Services techniques. The methodologies are suitable for integrating heterogeneous platforms, protocols and applications. The doctors can use the System and examine the biosignal of suspected patients. Currently, the accuracy of the EEG signal for epilepsy is over 98.73%.