心房顫動自動偵測

Automatic Detection of Atrial Fibrillation

黃瀞(Ching Huang)

指導教授 ： 林康平

中原大學/電機資訊學院/電機工程研究所/碩士(2016年)

https://doi.org/10.6840/cycu201600623

摘要

心房顫動是屬於心臟心律不整的一種疾病，會導致心臟心房跳動過快且不規則，因此會影響心臟流出的血液量，也可能造成血栓，心房顫動是導致中風的主要原因之一。

本論文為針對心臟之心房顫動問題設計開發一種新的方法，本論文中用來檢測心房顫動的工具是利用日常醫院心臟檢查的12導程心電圖訊號。心電圖上之P波異常可以得知心房跳動過快且不規則的情形，本論文之方法，會將此P波異常的情形定義為顫動波。心電圖上之R-R間期的不規則，可以得知心臟跳動過快與節律不協調，所以本論文的心房顫動偵測方法，主要是利用偵測心房顫動發生時，R-R間期的不規則及顫動波的特性，進行演算法的電腦程式實現。演算法架構可分為兩個部分，分別為偵測心跳的變動程度與分辨顫動波的特徵。第一部份所計算偵測的心跳變動程度，主要是在觀察R-R間期變動的嚴重程度，同時利用心跳差的方式，判斷心跳的規律性。第二部份的顫動波偵測，主要是藉由顫動波的振幅與速率兩項特徵，判斷心房內的電氣活動以及心房速率。一般12導程心電圖資料記錄時間為10秒，本論文之設計必須利用10秒的心電圖訊號診斷出心臟疾病。本論文會以10秒單導程心電訊號進行心房顫動偵測，並利用三種不同的心電圖資料庫進行本論文所提方法之驗證，這三種具有指標性的心電圖資料庫，分別為MIT-BIH心律不整資料庫、MIT-BIH心房顫動資料庫、MIT-BIH正常竇性心律資料庫，測試所使用的資料長度，是將這些資料庫中的所有訊號切割為10秒鐘，並重疊每段10秒訊號中的最後5秒，作為心房顫動偵測演算法所適用的心電訊號。

本論文所提出的心電訊號對於心房顫動偵測方法在三個心電圖資料庫的驗證下，其結果分別為MIT-BIH心律不整資料庫的靈敏度為92.5%與特異度為95.6%，MIT-BIH心房顫動資料庫的靈敏度80.1%為與特異度為95%，MIT-BIH正常竇性心律資料庫因無心房顫動的訊號，無法計算靈敏度，僅計算特異度，其特異度為99.5%，此結果相較於其他的心房顫動偵測方法，可以發現僅利用10秒心電圖訊號長度的需求，為本論文之優勢。未來希望能進一步與智慧型手機結合，並寫成APP能被廣泛簡易使用，如此將可隨時隨地即時偵測，也可與雲端配合儲存訊號，協助醫生判斷，協助病患盡早發現盡早治療。

關鍵字

心房顫動、心電圖

並列摘要

Atrial fibrillation (AF) is a condition of Arrhythmia, which results in tachycardia with an irregular or abnormal rhythm so the heart cannot effectively pump blood to the rest of body and thrombus may occur, which is a primary factor to cause stroke. In this study, a new method was designed, and a 12 lead electrocardiogram used for heart examination in hospitals was applied for AF examination. The abnormal P waves on ECG, which was defined as fibrillation wave in this study, exhibits tachycardia with an irregular or abnormal rhythm. The irregularity of R-R interval on ECG reveals tachycardia with abnormal rhythm. In this study, the characters of irregularity of R-R interval and fibrillation wave appearing during AF were applied for AF detection and for algorithm in computer programing. The algorithm was divided into two parts. One is for detecting of variation of heart rate, and the other one is for distinguishing the characters of fibrillation wave. The detection of variation of heart rate was applied to observe the irregularity of R-R interval while judging the regularity of heart rate from heart rate difference. The detection of fibrillation wave was applied to judge the atrial electrical activity and atrial rate by observing the amplitude of vibration and rate of fibrillation wave. Usually, the recording time was 10 seconds by a 12 lead electrocardiogram. In this study, heart disease was diagnosed by 10 seconds recording of ECG signals.
One lead electrocardiogram with 10 seconds recording was applied for AF detection, and three different ECG database were used for verification; including, MIT-BIH arrhythmia, MIT-BIH atrial fibrillation and MIT-BIH database normal sinus rhythm. All signals in the database were cut into 10 seconds recording, and the later 5 seconds of which were overlapped for the AF detection algorithm.

ECG signals from the study were verified by the three ECG database, and the results showed that the sensitivity of MIT-BIH arrhythmia was 92.5% and the specificity of which was 95.6%, the sensitivity of MIT-BIH atrial fibrillation was 80.1% and the specificity of which was 95 %. However, no AF signals were obtained from MIT-BIH database normal sinus rhythm for sensitivity but specificity only, which was 99.5%. Therefore, this showed that ECG signals recording for 10 seconds were an advantage of this study. The study results are expected to intergrade with cell-phone as an App to be used widely and detect problems simultaneously; in addition, to save signals in cloud and to assist doctors for fast diagnosis for the patients’ benefits.

並列關鍵字

Atrial fibrillation (AF), Electrocardiogram(ECG)

參考文獻

[9] Petrutiu, Simona, et al. "Atrial fibrillation and waveform characterization. " IEEE engineering in medicine and biology magazine 25.6 (2006): 24-30.

[13] Sahoo, Sujit Kumar, et al. "Detection of atrial fibrillation from non-episodic ecg data: a review of methods." 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, 2011.

[14] Wallmann, D., et al. "Frequent atrial premature contractions as a surrogate marker for paroxysmal atrial fibrillation in patients with acute ischaemic stroke."Heart 89.10 (2003): 1247-1248

[15] B. Pourbabaee and C. Lucas, “Automatic detection and prediction of paroxysmal atrial fibrillation based on analyzing ecg signal feature classification methods,” in 2008 Cairo International Biomedical Engineering Conference, CIBEC 2008, Cairo, Egypt, 2008.

[16] Hickey, Brian, Conor Heneghan, and Philip De Chazal. "Non-episode-dependent assessment of paroxysmal atrial fibrillation through measurement of RR interval dynamics and atrial premature contractions." Annals of Biomedical Engineering 32.5 (2004): 677-687.