- 學位論文
探討血壓訊號於心房顫動期間之特徵
The investigation of blood pressure signal characteristics during atrial fibrillation
摘要
房纖維性顫動是一種常見的心律不整疾病。大約有4~8%的中風是因為無症狀的心房顫動所造成的，所以如何早期發現、治療心房顫動成為一個重要的議題。目前市面也有許多血壓量測裝置同時提供檢測心房顫動的功能，主要檢測方式是利用心房顫動期間心跳在時間領域上訊號的不規則性來判斷是否為心律不整事件。利用這種方式的確可以有效分辨竇性心律、不規則心律，但是要利用這個方法判斷是否為心房顫動所造成不規則心律仍然有其不準確性。本研究主要目的是分析心房顫動期間侵入式血壓訊號的特徵，進而利用模糊專家系統達到病症分類的效果。 本研究共蒐集37位因為心房顫動疾病，需要進行導管電氣燒灼根治手術的病患，分別記錄病患在心房顫動時期、竇性心律時期的侵入式血壓波形和心電訊號波形。分析參數包括侵入式血壓間期、振幅的不規則性、射血間期與RR間期的比例及訊號亂度分析。最後選擇32位病患來建構以模糊理論為基礎的病症分類模型。利用5位病患來測試分類系統的效能。 在結果部分顯示由於病患心房顫動的類型，在不同心律狀態下心率變異有所差異，導致在分類時不是所有參數都適用每位病患。射血間期指數僅適用於A類型心房顫動患者，亂度取樣熵值則是針對B類型心房顫動病患，而間期和振幅不規則指數適用於所有病患。不同類型的病患使用單一參數固定分類閥值時分類效果的靈敏性和特異性分可達到78%，82%以上。而利用模糊專家系統建構的分類模型，又針對不同訊號量測位置分別統計分類效能。中心血壓訊號分類模型的靈敏性和特異性為73%，100%；而周邊血壓訊號分類結果靈敏性和特異性為93%，100%。中心血壓分類的靈敏性較低的原因在於測試病患中在竇性心律時有早發性心房收縮的現象，所以分類的效能會受到影響。 本研究完成侵入式血壓特徵點偵測，得知在不同心律狀態下分類參數分佈的差異性，並且利用模糊專家系統建構分類模型達到病症分類的目的。
並列摘要
Atrial fibrillation (AF) is the most common arrhythmia. From 4% to 8% of strokes are caused by asymptomatic AF. Thus, how to early detect and treat AF becomes an important issue. There are many blood pressure measuring devices provide capability to detect AF. Most of the detection methods use the heart rate irregularities to determine whether there is an arrhythmic event. In this way, it can only detect arrhythmic but can’t decide whether the arrhythmic event is caused by AF. The main purpose of this study is to analyze the signal characteristics of invasive blood pressure waveform during AF and then to use fuzzy expert system to achieve disease classification. Thirty-seven AF patients were recruited in the study. All patients were admitted for catheter ablation operation. Invasive blood pressure signal and ECG were recorded during AF and sinus rhythm. Parameters including irregularity index of time intervals, irregularity index of amplitude, ejection period divided by RR interval and signal entropy were analyzed. Thirty-two patients were selected to establish the classification model based on fuzzy expert system. Five other patients were selected to test the system performance. The results indicated that not every parameter was suitable for classification because differences in heart rate response in different type of AF. Ejection period index applies only to patients with A type AF and sample entropy value is for patients with B type AF. On the other hand, irregular index of time intervals and amplitude are suitable for all patients. Using single parameter and fixed threshold for different patient types, the result of classification achieved sensitivity and specificity of 78% and 82%, respectively. On the other hand, using classification model based on fuzzy expert system, and separate according to different measurement position, the sensitivity and specificity of central blood pressure signal classification model were 73% and 100%, respectively. While the sensitivity and specificity of peripheral blood pressure classification were 93% and 100%, respectively. The study accomplishes characteristic point detection in invasive blood pressure and confirmed that classification parameters distribution differently in different heart rate states. Lastly, a classification model was constructed based on fuzzy expert system to achieve the purpose of disease classification.