基於LSTM及運用EMD重複取樣之增強型 ECG SVEB 偵測方法

Enhanced ECG SVEB Detection using EMD with Resampling based on LSTM

盧成江(Agung Tri Arta Hariyanto)

指導教授 ： 王國禎

國立交通大學/資訊學院/資訊科學與工程研究所/碩士(2020年)

摘要

心房顫動 (AF) 是現今影響心律不整患者中最普遍的類型。透過對患者的ECG判讀和分類，尤其室上性異位搏動（SVEB）類別，可以幫助評估風險並有助於檢測AF。SVEB為AF中最難檢測的一種類型，然而當前的相關研究在單導聯ECG檢測SVEB的靈敏度、F1分數和G分數都顯示較低的分類預測結果，故此研究將使用單導聯ECG和針對患者本身的訓練模型，開發新的心律不整的搏動檢測方法，並根據AAMI標準對其進行分類，達成對SVEB高精準分類且仍滿足即時性的需求。我們提出的EMDR方法先用經驗模態分解 (EMD) 進行分解，並重採樣 (Resample) 第一個本質模態函數 (IMFs) 作為我們架構EMDR-LSTM (Long Short Term Memory) 的主要輸入。與其他採用兩個由一或二層LSTM的相關研究相對照，我們設計出一個新穎的深度學習模型架構。此架構僅為每一個輸入提供一個單層LSTM，這樣更適合我們的預處理方法EMDR，並可提升SVEB預測結果。EMDR-LSTM對增加SVEB分類預測結果有顯著的提升，據我們所知，此架構是第一個在LSTM應用是首個運用重採樣IMFs且基於AAMI標準對單導聯ECG進行AF分類的方法。與其他具代表性的方法相比，我們提出的EMDR-LSTM在所有資料集中準確率、靈敏度，陽性預測值、F1分數和G分數均顯示是有最好的分類預測結果。

關鍵字

心電圖（ECG）分類 ； 基於經驗模態分解重新採樣（EMDR） ； 長期短期記憶（LSTM）

並列摘要

Atrial Fibrillation (AF) is the most common arrhythmia type that affects patients today. Detecting and classifying patient ECG beats, especially the supraventricular ectopic beats (SVEB) class, can help to assess if patients have high possibilities of AF/flutter in the future. Detecting the SVEB class considered more difficult than the other classes. As a result, current related works show low classification prediction results, in terms of sensitivity, F1 score, and G score, for detecting the SVEB class in a single lead ECG. This work focuses on designing an arrhythmias beats detection method using single-lead ECG data with a patient-specific training model design, and does classification based on the AAMI standard. This work aims at achieving high classification prediction results in the SVEB class and still meets the real-time ECG classification requirement. The proposed method uses Empirical Mode Decomposition (EMD) with resampling (EMDR), which resamples the first Intrinsic Mode Functions (IMFs) as a main input, for our EMDR-LSTM (Long Short Term Memory) architecture. In contrast to the related works that use two separate models with one or two LSTM layers for each input, we design a novel LSTM model architecture that only uses a single model with one LSTM layer for each input. The proposed LSTM architecture is more suitable for our preprocessing method, EMDR, in order to enhance the SVEB classification prediction results. The proposed EMDR-LSTM has a significant effect to enhance SVEB classification prediction results. To the best of our knowledge, the proposed EMDR-LSTM is the first one that uses resamples first IMFs in LSTM that classifies arrhythmia using a single-lead ECG record based on the AAMI standard. Compared to representative related works, our proposed EMDR-LSTM achieves the highest classification prediction results in terms of accuracy, sensitivity, positive predictivity, F1, and G scores for SVEB class in all datasets.

並列關鍵字

Electrocardiogram (ECG) classification ； mpirical mode decomposition with resampling (EMDR) ； long short-term memory (LSTM)

參考文獻

[1] H. H. Publishing, “Atrial fibrillation: Common, serious, treatable.” [Online]. Available: https://www.health.harvard.edu/heart-health/atrial-fibrillation-common-serious-treatable. [Accessed: 10-Jul-2020].

[2] S. Saadatnejad, M. Oveisi, and M. Hashemi, “LSTM-Based ECG Classification for Continuous Monitoring on Personal Wearable Devices,” IEEE Journal of Biomedical and Health Informatics, vol. 24, no. 2, pp. 515–523, 2020.

[3] S. Kiranyaz, T. Ince, and M. Gabbouj, “Real-Time Patient-Specific ECG Classification by 1-D Convolutional Neural Networks,” IEEE Transactions on Biomedical Engineering, vol. 63, no. 3, pp. 664–675, 2016.

[4] M. Kachuee, S. Fazeli, and M. Sarrafzadeh, “ECG Heartbeat Classification: A Deep Transferable Representation,” 2018 IEEE International Conference on Healthcare Informatics (ICHI), 2018.

[5] A. L. Goldberger, L. A. N. Amaral, L. Glass, J. M. Hausdorff, P. C. Ivanov, R. G. Mark, J. E. Mietus, G. B. Moody, C.-K. Peng, and H. E. Stanley, “PhysioBank, PhysioToolkit, and PhysioNet,” Circulation, vol. 101, no. 23, 2000.