心肌梗塞分類建模

Using Envelope Representation for Myocardial Infarction Classification

陳麒宇(Chi-Yu Chen)

指導教授 ： 李育杰

國立交通大學/理學院/數學建模與科學計算研究所/碩士(2020年)

摘要

這篇論文，我們設計出了一套自動化診斷心肌梗塞的系統。有別於傳統十二導程心電圖紀錄，我們使用較方便穿戴的三導程心電圖紀錄作為心電圖裝置的診斷依據。利用envelope對於時間序列資料集上的表示，我們可以建構屬於健康狀態下以及不同類型心肌梗塞下的心電圖輪廓。心電圖紀錄的特徵可以用與這些心電圖輪廓的相似程度 來進行表示。透過表示結果，即可表明該紀錄是處於健康狀態或是心肌梗塞情況下。 在實驗中，我們使用德國聯邦物理技術研究院所提供的診斷心電圖數據庫來驗證我們的想法。建構心電圖輪廓時，我們發現儘管受試者在同一個類別下，心電圖記錄在R波上也會有所差異。因此，我們使用K-平均演算法在R波上將心電圖紀錄進行分群後 建構心電圖輪廓，使之降低因標準差過大而造成心電圖紀錄誤判的機會。在實驗結果中，區分心肌梗塞下的心電圖紀錄我們可以達到準確度95.31%、靈敏度97.01%、特異性87.50%。結果說明，envelope可以是一種心電圖的表示法，用來提取心電圖紀錄的特徵，幫助我們對健康狀態下以及心肌梗塞下的心電圖紀錄進行分類。

關鍵字

心肌梗塞 ； 心電圖 ； Envelope ； 時間序列 ； K-平均演算法

並列摘要

In this thesis, we have designed a system to diagnose myocardial infarction automatically. Different from the traditional 12-lead ECG record, we use a more convenient 3-lead ECG record as the diagnostic basis of the ECG device. Using an envelope representation for the time series dataset, we can build the ECG profiles under healthy control and different types of myocardial infarction. The ECG record’s features can represent by the similarity with these ECG profiles. The result indicates that the record is in healthy control or myocardial infarction. In our experiments, we used the diagnostic ECG database provided by Physikalisch-Technische Bundesanstalt (PTB) to validate our idea. When ECG profiles built, we found that although subjects were in the same class, the ECG record’s R-wave will be different. Therefore, we use the K-means algorithm to cluster the ECG record according to R-wave’s value and build their ECG profiles to reduce the possibility of misclassifying ECG records due to the significant standard deviation. In the experimental results, we can achieve an accuracy rate of 95.31%, a sensitivity of 97.01%, and a specificity of 87.50% to classify ECG records in myocardial infarction. The result shows that envelope can be a representation of the ECG. The ECG record’s features are represented by the ECG profiles, helping us to classify ECG records in healthy control or myocardial infarction.

並列關鍵字

Myocardial Infarction ； ECG ； Envelope ； Time series ； K-means

參考文獻

[1] Selcan Kaplan Berkaya et al. “A survey on ECG analysis”. In: Biomedical Signal Processing and Control 43 (2018), pp. 216–235.

[2] Masaki Kyoso and Akihiko Uchiyama. “Development of an ECG identification system”. In: 2001 Conference Proceedings of the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Vol. 4. IEEE. 2001, pp. 3721-3723.

[3] Nino Isakadze and Seth S Martin. “How useful is the smartwatch ECG?” In: Trends in cardiovascular medicine (2019).

[4] Ary L Goldberger et al. “PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals”. In: circulation 101.23 (2000), e215–e220.

[5] Benjamin E Jin et al. “A simple device to illustrate the Einthoven triangle”. In: Advances in physiology education 36.4 (2012), pp. 319–324.