利用容許區間方法判別 extreme delta brush 腦波波形

Abstract

在醫學文獻提到的抗NMDA受體腦炎（Anti-NMDA receptor encephalitis）病患的腦電圖，較常有一種名為extreme delta brush的特殊腦電波形，由於此腦電波形的發現，藉此增加此疾病的診斷方式。在醫學界，腦電圖（Electroencephalograph, EEG）雖然不會危害人體健康，但在判讀上會花費醫生長時間的診斷，導致醫生進而選擇更快速精確且可能危害人體的檢查方式。所以我們在本論文中想發展一套統計方法來偵測腦波中extreme delta brush的特殊波形。在本論文當中，我們利用容許區間方法，來對此特殊波形的資料的變異數建構出適當的容許區間，來利用容許界線來偵測有extreme delta brush特性的腦電圖。最後從我們十三筆實際腦電圖資料的分析結果發現，我們所提出的方法在判別extreme delta brush上，除了right frontal lobe epilepsy的腦電圖外，抗NMDA受體腦炎腦電圖佔extreme delta brush特性的比例比其他腦電圖都來得高。

From literature, we could discover that there is a special Electroencephalograph (EEG) pattern, known as extreme delta brush, in the EEG of Anti-NMDA receptor encephalitis patients. With such discovery, we could analyze this pattern in order to improve the diagnostic accuracy of the disease. Although EEG is a noninvasive method which does not cause any harm to human body, it is however very time consuming to analyze EEG data. In this thesis, we adopt a tolerance interval approach to develop a statistical method which can efficiently detect extreme delta brush patterns. Based on an appropriate tolerance limit for the variance of the extreme delta brush data, we can set a threshold for the variance of EEG data to detect the extreme delta brush pattern. We analyze 13 EEG datasets using the proposed method. The result shows that extreme delta brush pattern can be detected more often in two Anti-NMDA receptor encephalitis EEG datasets than in the EEG datasets of other diseases except in an EEG dataset of right frontal lobe epilepsy.