自動化剔除人為動作干擾功能之電腦輔助胃電圖診斷系統開發

Development of a Electrogastrogram Computer Aided Diagnostic System with the Automatic Motion Artifact Deletion Function

傅家啟(Jachih J. C. Fu) ； 連漢仲(Han-Chung Lien) ； 劉邦彥(Ban-Yen Liu)

《大葉學報》 8卷1期 (1999/06) Pp. 79-88

https://doi.org/10.7119/JDYU.199906.0079

摘要

資訊及訊號處理自動化等技術之應用已廣泛應用於製造業及一般產業，然而於胃電圖醫療輔助診斷自動化之領域中，上述技術應用之程度仍低且需求甚殷。胃電圖因其為非侵入性量測，操作簡單安全且易被受測者接受，因此為一甚具有發展潛力的診斷工具，唯胃電圖訊號量測其間易受呼吸、人為動作及其它器官的訊號干擾，因此在醫療臨床診斷上，胃電圖原始訊號之判讀甚為困難。且現有一般胃電圖訊號分析與判讀工具對於臨床診斷及研究之功能有限，不敷醫療研究之要求。本論文針對現有軟體之缺陷，發展一更具彈性與更有效之胃電圖訊號處理與分析工具，同時結合類神經網路與相關訊號處理演算法以自動化剔除人為動作所引起之雜訊，並測試前述自動化功能之績效。

關鍵字

醫療自動化 ； 電腦輔助診斷 ； 胃電圖 ； 類神經網路

並列摘要

The potential uses of electrogastrograms (EGGs) in clinical diagnosis and analysis have been widely discussed in the medical research community. However, the practical uses of the EGG are currently limited by the its low signal to noise ratio. Physicians can not easily interpret the EGG because of motion artifacts during signal recording, which are responsible for most of the noise. Therefore, there is widespread disagreement about the usage of EGG signals for clinical analysis.EGG based diagnosis will probably remain in the exploratory phase until software tools are developed for automated removal of motion artifacts. Current commercially available EGG software can determine the beginning and ending points for motion artifacts only based on manual input, making the results subjective and the EGG analysis tedious and time consuming. Moreover, because the current EGG software is of the ”stand alone” type, it is not integrated with other standard signal processing software, making it hard to use in research.In this paper, we present a share-ware EGG analysis tool to cope with the motion artifact problem. The software is based on a neural network motion artifact removal algorithm. The share-ware design allows users to access and to change the computer source code according to their needs. In this paper, we describe the neural network model and verify its performance.

並列關鍵字

medical automation ； computer aided diagnosis ； electrogastrogram ； neural network