臺灣博碩士論文加值系統

心血管疾病(Cardiovascular diseases, CVDs)在人類十大死因中比例最高，為整體 16%。從 2000 年至 2019 因缺鐵性心臟病致死的人數增加 200萬人。然而一般心血管疾病不容易診斷，需使用心電圖(ECG)儀器進行時間不短的檢測。為了協助醫師能高效判別心電圖病徵，供醫師們參考的高準確度自動判別病徵數據，提出使用一維卷積神經網路(1D-CNN)、二維卷積神經網路(2D-CNN)、長短記憶神經網路(LSTM)三個神經網路模型，以整體學習(Ensemble Learning)合作判斷病徵，提升平均準確判別。。以及提出使用一維資料擴增處理、一維資料轉二維資料二值化轉換處理，彌補心電圖標籤數量不足以及優化二維資料處理速度。經實驗數據表現，在 80%訓練 20%預測資料使用方式下， 1D-CNN 為 99.2%、2D-CNN 為98.6%、LSTM 為 99.1%，整體合作判別為99.5%。經 AAMI 協會建議的標準下的數據表現，判別 VEB 病徵的準確率Acc為 99.3%、敏感度Sen為 98.7%、特異度Spe為 99.4%以及陽性預測率Ppr為 95.1%。判別 SVEB 病徵的準確率Acc為98.4%、敏感度Sen為 97.1%、特異度Spe為 98.5%以及陽性預測率Ppr為73.1%。本研究數據與其他文獻相較，顯示出以整體學習合作判斷，具有良好的成效。

Cardiovascular diseases (CVDs) have the highest proportion of the top ten causes of death in humans, at 16% overall. From 2000 to 2019 the number of deaths from iron deficiency heart disease increased by 2 million. However, cardiovascular disease in general is not easy to diagnose, and electrocardiogram (ECG) instruments are required for detection for a long time. In order to help physicians efficiently determine ECG symptoms, high-accuracy automatic identification of symptom data for the reference of physicians, it is proposed to use one-dimensional convolutional neural network (1D-CNN), two-dimensional convolutional neural network (2D-CNN), Three neural network models of long and short memory neural networks (LSTM) cooperate to determine the symptoms by ensemble Learning, and improve the average accuracy of the judgment. In addition, it is proposed to use one-dimensional augmentation data processing and one-dimensional data to two-dimensional data binarization conversion processing to make up for the insufficient number of ECG labels and optimize the processing speed of two-dimensional data. According to experimental data, In 80% training 20% predictive data usage, 1DCNN is 99.2% and 2D-CNN is 98.6%. The LSTM is 99.1%, and the overall cooperation judgment is 99.5%. According to the criteria recommended by the AAMI Association, for symptoms VEB, Acc(Accuracy) 99.3%, Sen(Sensitivity) 98.7%, Spe(Specificity) 99.4% and Ppr(Positive predictive rate) 95.1%, respectively. For symptoms SVEB Acc 98.4%, Sen 97.1%, Spe 98.5% and Ppr 73.1%, respectively. Compared with other literature, the data in this study showed good results based on cooperation with Ensemble Learning.

摘要 I
Abstract II
致謝 III
目錄 IV
圖目錄 VI
表目錄 VIII
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 2
1.3 研究貢獻 3
1.4 論文架構 4
第二章 相關研究 5
2.1 心電圖簡介 5
2.2 資料集 6
2.3 習知心電圖自動分類相關研究 7
2.3.1 人工神經網路 8
2.3.1.1 激活函數 9
2.3.2 一維卷積神經網路 11
2.3.3 二維卷積神經網路 14
2.3.4 遞歸神經網路與長短期記憶神經網路 16
2.3.5 混合模型 18
2.3.6 整體學習 19
第三章 多重神經網路架構設計 24
3.1 ECG訊號資料前處理 24
3.1.1 小波去噪 25
3.1.2 波形取樣 27
3.1.3 一維波形資料擴增處理 28
3.1.4 資料切割 30
3.2 一維卷積神經網路架構 31
3.3 長短期神經網路架構 32
3.4 一維轉二維資料處理 33
3.5 二維卷積神經網路架構 35
3.6 整合架構流程 36
第四章 實驗成果 39
4.1 隨機(8/2比例)實驗數據 39
4.1.1 以隨機(8/2比例)數據與先前文獻比較 41
4.1.2 一維資料擴增處理前後比較 42
4.1.3 一維轉二維資料處理方式比較 44
4.1.4 使用MLP架構結合模型比較 46
4.2 使用AAMI標準之實驗數據 48
4.2.1 AAMI 指定VEB預測資料實驗數據 51
4.2.2 AAMI 指定SVEB預測資料實驗數據 53
4.2.3 以AAMI標準與先前文獻之數據比較 56
第五章 結論與未來展望 58
參考文獻 60

圖目錄
圖 2.1 心臟系統與電氣運作示意圖 [19] 5
圖 2.2 心電圖波形示意圖 [20, 21] 5
圖 2.3神經網路架構與感知器示意圖 9
圖 2.4 S函數 10
圖 2.5 tanh 函數 10
圖 2.6 tanh 函數 11
圖 2.7 源自提出CNN概念LeNet-5架構 [40] 11
圖 2.8 一維卷積神經網路示意圖 12
圖 2.9 卷積FULL模式 13
圖 2.10 卷積SAME模式 13
圖 2.11 卷積VALID模式 13
圖 2.12 2局部取樣最大與平均池化 14
圖 2.13二維卷積神經網路示意圖 15
圖 2.14 二維資料擴增方式 [11] 15
圖 2.15 RNN單元時序關係以及架構 16
圖 2.16 單元在時序為t的LSTM構造 17
圖 2.17 1D-CNN-LSTM架構[46] 18
圖 2.18雙向(bidirectional)LSTM架構 [48] 19
圖 2.19整體學習-袋裝法 20
圖 2.20 (a)整體學習-提升法 (b)整體學習-堆疊法 20
圖 2.21 二重CNN模型 [53] 21
圖 2.22多重混合RNN架構 [16] 22
圖 2.23 1D-CNN與LSTM 二重異質模型 [15] 22
圖 3.1 整體多重模型評估架構 24
圖 3.2 ECG訊號資料前處理流程 25
圖 3.3 用於2級分解小波去噪處理 [55] 25
圖 3.4 小波轉換9層轉換 26
圖 3.5 小波去噪流程圖 27
圖 3.6 小波轉換9層反轉換 27
圖 3.7 (a)小波去噪前之波形 (b)小波去噪後之波形 27
圖 3.8 波形取樣示意圖 28
圖 3.9 ECG一維資料波形進行擴增資料處理 29
圖 3.10 隨機(8/2比例)資料切割方式 30
圖 3.11 AAMI標準資料切割方式 31
圖 3.12 1D-CNN 架構圖 32
圖 3.13 LSTM架構圖 32
圖 3.14 一維資料轉二維資料區塊 33
圖 3.15 一維資料轉成二維資料 33
圖 3.16 一維資料轉二維資料波形流程圖 34
圖 3.17 一維資料轉二維資料流程示意圖 35
圖 3.18 2D-CNN架構圖 36
圖 3.19 整合流程參數處理 37
圖 3.20投票判別流程 38
圖 4.1 15分類轉換為AAMI 5分類標籤對照 41
圖 4.2 1D-CNN 2D-CNN LSTM共同FC層模型 46
圖 4.3 1D-CNN 2D-CNN共同MLP層模型 47
圖 4.4 AAMI標準測試資料使用方式 49
圖 4.5 VEB二分類轉換流程 49
圖 4.6 SVEB二分類轉換流程 49
圖 4.7 (a)轉換為VEB二分類標籤 (b)轉換為SVEB二分類標籤 50
圖 4.8 二分類參數表 50
圖 5.1 R-R波峰區間訊號 59

表目錄
表 2.1 ECG 15分類與5分類波形特徵差異 7
表 3.1 一維資料平移擴增方式 29
表 3.2 ECG 15分類總資料筆數 30
表 4.1 隨機(8/2比例)-1D-CNN 15分類數據 39
表 4.2 隨機(8/2比例)-2D-CNN 15分類數據 40
表 4.3 隨機(8/2比例)-LSTM 15分類數據 40
表 4.4 隨機(8/2比例)-經評估架構流程15分類數據 40
表 4.5 隨機(8/2比例)-5分類轉換數據 41
表 4.6 隨機(8/2比例)-數據與先前文獻比較表 42
表 4.7 隨機(8/2比例)-1D-CNN 15分類數據-擴增前 43
表 4.8 隨機(8/2比例)-1D-CNN 15分類數據-擴增後 43
表 4.9 一維擴增數量以及模型準確率前後比較 43
表 4.10 一維轉二維資料-一張波形轉換時間 44
表 4.11 隨機(8/2比例)十分之一數據訓練-2D-CNN 256取樣點數據 45
表 4.12 隨機(8/2比例)十分之一數據訓練-2D-CNN 128取樣點數據 45
表 4.13 隨機(8/2比例)十分之一數據訓練-2D-CNN 64取樣點數據 45
表 4.14 隨機(8/2比例)數據訓練-1D-CNN 2D-CNN LSTM共同MLP模型數據 48
表 4.15 隨機(8/2比例)數據訓練-1D-CNN 2D-CNN LSTM共同MLP模型數據 48
表 4.16 使用AAMI-VEB 1D-CNN 15分類數據 51
表 4.17 使用AAMI-VEB 2D-CNN 15分類數據 52
表 4.18 使用AAMI-VEB LSTM 15分類數據 52
表 4.19 使用AAMI-VEB經整合架構流程後15分類數據 52
表 4.20 使用AAMI-VEB 5分類轉換數據 53
表 4.21 使用AAMI-VEB 2分類轉換數據 53
表 4.22 使用AAMI-SVEB 1D-CNN 15分類數據 53
表 4.23 使用AAMI-SVEB 2D-CNN 15分類數據 54
表 4.24 使用AAMI-SVEB LSTM 15分類數據 54
表 4.25 使用AAMI-SVEB經整合架構流程後15分類數據 54
表 4.26 使用AAMI-SVEB 5分類轉換數據 55
表 4.27 使用AAMI-SVEB 2分類轉換數據 55
表 4.28 AAMI訓練數量以及測試數量比較 55
表 4.29 與其他使用AAMI相關研究數據比較 56

[1]"The top 10 causes of death." https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death (accessed.
[2]" Electrocardiograph." https://www.newton.com.tw/wiki/%E6%84%9B%E5%9B%A0%E6%89%98%E8%8A%AC (accessed.
[3]V. Mittal, D. Gangodkar, and B. Pant, "Exploring The Dimension of DNN Techniques For Text Categorization Using NLP," in 2020 6th International Conference on Advanced Computing and Communication Systems (ICACCS), 2020: IEEE, pp. 497-501.
[4]S. Sakti, S. Kawanishi, G. Neubig, K. Yoshino, and S. Nakamura, "Deep bottleneck features and sound-dependent i-vectors for simultaneous recognition of speech and environmental sounds," in 2016 IEEE Spoken Language Technology Workshop (SLT), 2016: IEEE, pp. 35-42.
[5]Y. Wu, "A Chinese-English Machine Translation Model Based on Deep Neural Network," in 2020 International Conference on Intelligent Transportation, Big Data & Smart City (ICITBS), 2020: IEEE, pp. 828-831.
[6]A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems, vol. 25, 2012.
[7]K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[8]Y. H. Hu, W. J. Tompkins, J. L. Urrusti, and V. X. Afonso, "Applications of artificial neural networks for ECG signal detection and classification," Journal of electrocardiology, vol. 26, pp. 66-73, 1993.
[9]K.-i. Minami, H. Nakajima, and T. Toyoshima, "Real-time discrimination of ventricular tachyarrhythmia with Fourier-transform neural network," IEEE transactions on Biomedical Engineering, vol. 46, no. 2, pp. 179-185, 1999.
[10]O. T. Inan, L. Giovangrandi, and G. T. Kovacs, "Robust neural-network-based classification of premature ventricular contractions using wavelet transform and timing interval features," IEEE transactions on Biomedical Engineering, vol. 53, no. 12, pp. 2507-2515, 2006.
[11]T. J. Jun, H. M. Nguyen, D. Kang, D. Kim, D. Kim, and Y.-H. Kim, "ECG arrhythmia classification using a 2-D convolutional neural network," arXiv preprint arXiv:1804.06812, 2018.
[12]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, 2015.
[13]E. D. Übeyli, "Recurrent neural networks employing Lyapunov exponents for analysis of ECG signals," Expert systems with applications, vol. 37, no. 2, pp. 1192-1199, 2010.
[14]S. Chauhan and L. Vig, "Anomaly detection in ECG time signals via deep long short-term memory networks," in 2015 IEEE International Conference on Data Science and Advanced Analytics (DSAA), 2015: IEEE, pp. 1-7.
[15]F. Y. Zhou, L. P. Jin, and J. Dong, "Premature ventricular contraction detection combining deep neural networks and rules inference," Artificial Intelligence in Medicine, vol. 79, pp. 42-51, Jun 2017, doi: 10.1016/j.artmed.2017.06.004.
[16]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, doi: 10.1109/jbhi.2019.2911367.
[17]湯松年, 許峯橋, 許寓翔, and 呂緯祥, "使用卷積神經網路和遞歸神經網路雙模型輔助心電圖診斷系統," 中原大學資訊工程學系.
[18]湯松年, 洪以恩, 蔡旻樺, and 陳守倫, "使用卷積神經網路輔助心電圖診斷系統開發," 中原大學資訊工程學系.
[19]Schematic diagram of the cardiac system and electrical operation https://en.wikipedia.org/wiki/Cardiac_pacemake (accessed.
[20]"Electrocardiogram(ECG)-howMed." http://howmed.net/physiology/electrocardiogram-ecg/ (accessed.
[21]"How the heart works." https://www.nhlbi.nih.gov/health-topics/how-heart-works (accessed.
[22]G. Moody and R. Mark. "MIT-BIH Arrhythmia Database." https://physionet.org/content/mitdb/1.0.0/ (accessed.
[23]G. B. Moody and R. G. Mark, "The impact of the MIT-BIH arrhythmia database," IEEE Engineering in Medicine and Biology Magazine, vol. 20, no. 3, pp. 45-50, 2001.
[24]李哲瑋, "利用1維卷積神經網及切割法可即時適應病患的心電圖分類," 碩士, 電子工程學研究所, 國立臺灣大學, 台北市, 2018. [Online]. Available: https://hdl.handle.net/11296/5qm8r8
[25]S. Liu, J. Shao, T. Kong, and R. Malekian, "ECG Arrhythmia Classification using High Order Spectrum and 2D Graph Fourier Transform," Applied Sciences, vol. 10, p. 4741, 07/09 2020, doi: 10.3390/app10144741.
[26]P. De Chazal, M. O'Dwyer, and R. B. Reilly, "Automatic classification of heartbeats using ECG morphology and heartbeat interval features," IEEE transactions on biomedical engineering, vol. 51, no. 7, pp. 1196-1206, 2004.
[27]P. De Chazal and R. B. Reilly, "A patient-adapting heartbeat classifier using ECG morphology and heartbeat interval features," IEEE transactions on biomedical engineering, vol. 53, no. 12, pp. 2535-2543, 2006.
[28]M. Lagerholm, C. Peterson, G. Braccini, L. Edenbrandt, and L. Sornmo, "Clustering ECG complexes using Hermite functions and self-organizing maps," IEEE Transactions on Biomedical Engineering, vol. 47, no. 7, pp. 838-848, 2000.
[29]R. V. Andreao, B. Dorizzi, and J. Boudy, "ECG signal analysis through hidden Markov models," IEEE Transactions on Biomedical engineering, vol. 53, no. 8, pp. 1541-1549, 2006.
[30]L.-Y. Shyu, Y.-H. Wu, and W. Hu, "Using wavelet transform and fuzzy neural network for VPC detection from the Holter ECG," IEEE Transactions on Biomedical Engineering, vol. 51, no. 7, pp. 1269-1273, 2004.
[31]M. Minsky and S. Papert, "An introduction to computational geometry," Cambridge tiass., HIT, vol. 479, p. 480, 1969.
[32]J. M. Nazzal, I. M. El-Emary, and S. A. Najim, "Multilayer perceptron neural network (MLPs) for analyzing the properties of Jordan Oil Shale 1," 2008.
[33]M.-C. Popescu, V. E. Balas, L. Perescu-Popescu, and N. Mastorakis, "Multilayer perceptron and neural networks," WSEAS Transactions on Circuits and Systems, vol. 8, no. 7, pp. 579-588, 2009.
[34]S. Osowski and T. H. Linh, "ECG beat recognition using fuzzy hybrid neural network," IEEE Transactions on Biomedical Engineering, vol. 48, no. 11, pp. 1265-1271, 2001.
[35]W. Jiang and S. G. Kong, "Block-based neural networks for personalized ECG signal classification," IEEE Transactions on Neural Networks, vol. 18, no. 6, pp. 1750-1761, 2007.
[36]T. Ince, S. Kiranyaz, and M. Gabbouj, "A generic and robust system for automated patient-specific classification of ECG signals," IEEE Transactions on Biomedical Engineering, vol. 56, no. 5, pp. 1415-1426, 2009.
[37]Y. LeCun et al., "Backpropagation applied to handwritten zip code recognition," Neural computation, vol. 1, no. 4, pp. 541-551, 1989.
[38]R. Pascanu, T. Mikolov, and Y. Bengio, "Understanding the exploding gradient problem," CoRR, abs/1211.5063, vol. 2, no. 417, p. 1, 2012.
[39]J. X. Gu et al., "Recent advances in convolutional neural networks," Pattern Recognition, vol. 77, pp. 354-377, May 2018, doi: 10.1016/j.patcog.2017.10.013.
[40]Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
[41]M. Kachuee, S. Fazeli, and M. Sarrafzadeh, "Ecg heartbeat classification: A deep transferable representation," in 2018 IEEE international conference on healthcare informatics (ICHI), 2018: IEEE, pp. 443-444.
[42]P. Rajpurkar, A. Y. Hannun, M. Haghpanahi, C. Bourn, and A. Y. Ng, "Cardiologist-level arrhythmia detection with convolutional neural networks," arXiv preprint arXiv:1707.01836, 2017.
[43]Y. Wu, F. Yang, Y. Liu, X. Zha, and S. Yuan, "A comparison of 1-D and 2-D deep convolutional neural networks in ECG classification," arXiv preprint arXiv:1810.07088, 2018.
[44]S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural computation, vol. 9, no. 8, pp. 1735-1780, 1997.
[45]R. C. Staudemeyer and E. R. Morris, "Understanding LSTM--a tutorial into long short-term memory recurrent neural networks," arXiv preprint arXiv:1909.09586, 2019.
[46]S. L. Oh, E. Y. K. Ng, R. S. Tan, and U. R. Acharya, "Automated diagnosis of arrhythmia using combination of CNN and LSTM techniques with variable length heart beats," Computers in Biology and Medicine, vol. 102, pp. 278-287, Nov 2018, doi: 10.1016/j.compbiomed.2018.06.002.
[47]G. Petmezas et al., "Automated Atrial Fibrillation Detection using a Hybrid CNN-LSTM Network on Imbalanced ECG Datasets," Biomedical Signal Processing and Control, vol. 63, Jan 2021, Art no. 102194, doi: 10.1016/j.bspc.2020.102194.
[48]Ö. Yildirim, "A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification," Computers in biology and medicine, vol. 96, pp. 189-202, 2018.
[49]L. K. Hansen and P. Salamon, "Neural network ensembles," IEEE transactions on pattern analysis and machine intelligence, vol. 12, no. 10, pp. 993-1001, 1990.
[50]" Ensemble Learning." https://ithelp.ithome.com.tw/articles/10271882 (accessed.
[51]L. Breiman, "Bagging predictors," Machine learning, vol. 24, no. 2, pp. 123-140, 1996.
[52]J. Brownlee, Ensemble Learning Algorithms With Python Make Better Predictions with Bagging, Boosting, and Stacking, 2022.
[53]X. Fan, Q. Yao, Y. Cai, F. Miao, F. Sun, and Y. Li, "Multiscaled fusion of deep convolutional neural networks for screening atrial fibrillation from single lead short ECG recordings," IEEE journal of biomedical and health informatics, vol. 22, no. 6, pp. 1744-1753, 2018.
[54]U. Orhan, M. Hekim, and M. Ozer, "EEG signals classification using the K-means clustering and a multilayer perceptron neural network model," Expert Systems with Applications, vol. 38, no. 10, pp. 13475-13481, Sep 2011, doi: 10.1016/j.eswa.2011.04.149.
[55]M. A. Kabir and C. Shahnaz, "Denoising of ECG signals based on noise reduction algorithms in EMD and wavelet domains," Biomedical Signal Processing and Control, vol. 7, no. 5, pp. 481-489, Sep 2012, doi: 10.1016/j.bspc.2011.11.003.
[56]J.-J. Ding. "Time frequency analysis and wavelet transform class note,the Department of Electrical Engineering." National Taiwan University (NTU). http://djj.ee.ntu.edu.tw/TFW.htm (accessed.
[57]J. O. S. III. "Spectral Audio Signal Processing." W3K Publishing. https://ccrma.stanford.edu/~jos/sasp/Upsampling_Stretch_Operator.html (accessed.
[58]A. Isin and S. Ozdalili, "Cardiac arrhythmia detection using deep learning," Procedia computer science, vol. 120, pp. 268-275, 2017.
[59]M. Sokolova and G. Lapalme, "A systematic analysis of performance measures for classification tasks," Information Processing & Management, vol. 45, no. 4, pp. 427-437, Jul 2009, doi: 10.1016/j.ipm.2009.03.002.
[60]Y. H. Hu, S. Palreddy, and W. J. Tompkins, "A patient-adaptable ECG beat classifier using a mixture of experts approach," IEEE transactions on biomedical engineering, vol. 44, no. 9, pp. 891-900, 1997.