臺灣博碩士論文加值系統

根據衛福部研究 , 2016年心臟疾病高居台灣十大死因第二名 , 心臟疾病死亡人數每年以約七百位增加中 , 因此如何提高心臟相關疾病診斷準確度為當務之急 。 本研究結合深層類神經網路及心電圖資料來發展出心電圖左心室肥厚症分類器 , 該分類系統會先利用資料預處理技術來填補心電圖的遺失值並根據深層類神經網路的需求來轉化心電圖資料 , 接著套入本研究設計的六層深類神經網路來進行左心室肥厚症的預測 , 而在每次學習迭代中 , 會以L2-Regularization 及Dropout 兩種方式來避免過度學習 。 整體系統稱為 Electrocardiographic left ventricular hypertrophy classifier (ELVHC) , 而實驗結果顯示該預測模型的準確度達73% 、 敏感度為66% 、 特異度為78% , 傳統心電圖臨床方法的敏感度僅達三成 , 本研究提出的方式於敏感度及準確度方面有顯著的提高 。

Heart disease ranks second in Taiwan’s top ten cause of death in 2016 and the number of deaths in heart disease increases by about 700 people each year. Therefore,
how to improve the accuracy of heart disease diagnosis is urgent. This study combined the deep neural network and ECG data to develop an ECG left ventricular hypertrophy classifier. The classification system uses data preprocessing technology to impute missing values of ECG data and transform ECG data for the requirement of the deep neural network. Then, input the data into the six-layers deep neural network designed in this study was used to predict left ventricular hypertrophy. L2-Regularization and Dropout are used to avoid overfitting in each model iteration. The system has been named the Electrocardiographic left ventricular hypertrophy classifier (ELVHC). The experimental results show that the accuracy of the prediction model is 73%, the sensitivity is 66%, and the specificity is 78%. The deep neural network model that the study proposed is significantly higher than two clinical methods in the sensitivity and the accuracy.

誌謝. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
摘要. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3. Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4. Architecture of study . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2. Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1. The description of the artificial neural network and the random forest
algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1. Artificial neural network . . . . . . . . . . . . . . . . . . . . . 8
2.1.2. Feedforward . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3. Back-propagation . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.4. Random forest algorithm . . . . . . . . . . . . . . . . . . . . . 14
2.2. The definition of the electrocardiogram and left ventricular hypertrophy 15
2.2.1. Electrocardiogram . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.2. Left ventricular mass index . . . . . . . . . . . . . . . . . . . 19
2.2.3. Left ventricular hypertrophy (LVH) . . . . . . . . . . . . . . . 21
2.2.4. Two clinical criteria for diagnosis LVH in the electrocardiogram 22
2.3. Related application in ECG and cardiovascular disease . . . . . . . . 22
3. Experimental Materials and Methods . . . . . . . . . . . . . . . . . . 26
3.1. Experimental Material . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2. Data-preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1. K-Nearest Neighbor . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2. Z-score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.3. One-Hot-Encoding . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3. Overfitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.1. L2-Regularization . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.2. Dropout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4. Model performance evaluation . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1. Confusion matrix . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5. The algorithm of deep neural network . . . . . . . . . . . . . . . . . . 34
4. Experimental Environment and Results . . . . . . . . . . . . . . . . . 36
4.1. Architecture of model building . . . . . . . . . . . . . . . . . . . . . . 36
4.2. Experimental environment . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.1. Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.2. Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3. The result of random forest models . . . . . . . . . . . . . . . . . . . 38
4.4. The result of deep neural network models . . . . . . . . . . . . . . . . 40
4.5. Compare with clinical criteria . . . . . . . . . . . . . . . . . . . . . . 42
5. Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . 45
5.1. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2. Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1. Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

[1] “Taiwan ministry of health and welfare.” https://www.mohw.gov.tw/cp-16-33598-1.html.
[2] W. B. Kannel, “Left ventricular hypertrophy as a risk factor: the Framingham experience.,” Journal of hypertension. Supplement: official journal of the International Society of Hypertension, vol. 9, no. 2, pp. S3–8; discussion S8–9,1991.
[3] P. D. Myers, B. M. Scirica, and C. M. Stultz, “Machine Learning Improves Risk Stratification After Acute Coronary Syndrome,” Scientific Reports, vol. 7,
p. 12692, Oct. 2017.
[4] K. O. Gupta and P. Chatur, “Ecg signal analysis and classification using data mining and artificial neural networks 1,” 2012.
[5] J. Adnan, N. Daud, A. Mokhtar, F. Hashim, S. Ahmad, A. Rashidi, and Z. Rizman, “Multilayer perceptron based activation function on heart abnormality
activity,” Journal of Fundamental and Applied Sciences, vol. 9, no. 3S, pp. 417–432, 2017.
[6] U. R. Acharya, H. Fujita, O. S. Lih, M. Adam, J. H. Tan, and C. K. Chua, “Automated detection of coronary artery disease using different durations of
ecg segments with convolutional neural network,” Knowledge-Based Systems, vol. 132, pp. 62–71, 2017.
[7] M. Boussaa, I. Atouf, M. Atibi, and A. Bennis, “Ecg signals classification using mfcc coefficients and ann classifier,” in International Conference on Electrical
and Information Technologies (ICEIT), pp. 480–484, IEEE, 2016.
[8] K.-l. Hsu, H. V. Gupta, and S. Sorooshian, “Artificial neural network modeling of the rainfall-runoff process,” Water resources research, vol. 31, no. 10, pp. 2517–2530, 1995.
[9] D. O. Hebb, The organization of behavior: A neuropsychological theory. Psychology Press, 2005.
[10] M. F. OSC, “Structure of a neuron.” https://owlcation.com/stem/Structure-of-a-Neuron.
[11] F. Rosenblatt, “The perceptron: a probabilistic model for information storage and organization in the brain.,” Psychological review, vol. 65, no. 6, p. 386,
1958.
[12] X. Glorot, A. Bordes, and Y. Bengio, “Deep sparse rectifier neural networks,” in International Conference on Artificial Intelligence and Statistics, pp. 315–323,
2011.
[13] G. Klambauer, T. Unterthiner, A. Mayr, and S. Hochreiter, “Self-normalizing neural networks,” in Advances in Neural Information Processing Systems,
pp. 972–981, 2017.
[14] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning representations by back-propagating errors,” Nature, vol. 323, pp. 533–536, Oct. 1986.
[15] A. Liaw, M. Wiener, et al., “Classification and regression by randomforest,” R news, vol. 2, no. 3, pp. 18–22, 2002.
[16] T. G. Dietterich, “Ensemble learning,” The handbook of brain theory and neural networks, vol. 2, pp. 110–125, 2002.
[17] L. Fraiwan, K. Lweesy, N. Khasawneh, H. Wenz, and H. Dickhaus, “Automated sleep stage identification system based on time–frequency analysis of a single
eeg channel and random forest classifier,” Computer methods and programs in biomedicine, vol. 108, no. 1, pp. 10–19, 2012.
[18] R. G. Kumar and Y. Kumaraswamy, “Investigating cardiac arrhythmia in ecg using random forest classification,” Int. J Comput. Appl, vol. 37, pp. 31–34,
2012.
[19] R. E. Klabunde, “Electrocardiogram graph.” http://www.cvphysiology.com/Arrhythmias/A009.
[20] Cables and L. Sensors, “The description of the 12-leads ecg.” https://www.cablesandsensors.com/pages/12-lead-ecg-placement-guide-with-illustrations.
[21] R. B. Devereux, D. R. Alonso, E. M. Lutas, G. J. Gottlieb, E. Campo, I. Sachs, and N. Reichek, “Echocardiographic assessment of left ventricular hypertrophy: Comparison to necropsy findings,” American Journal of Cardiology, vol. 57,
pp. 450–458, Feb. 1986.
[22] R. D. Mosteller, “Simplified calculation of body-surface area,” The New England Journal of Medicine, vol. 317, p. 1098, Oct. 1987.
[23] E. L. Schiffrin, Q. Pu, and J. B. Park, “Effect of amlodipine compared to atenolol on small arteries of previously untreated essential hypertensive patients,” American journal of hypertension, vol. 15, no. 2, pp. 105–110, 2002.
[24] M. Clinic, “Comparison of normal heart and lvh.” https://www.mayoclinic.org/diseases-conditions/left-ventricular-hypertrophy/multimedia/
left-ventricular-hypertrophy/img-20008677.
[25] M. Sokolow and T. P. Lyon, “The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads,” American heart journal, vol. 37, no. 2, pp. 161–186, 1949.
[26] P. N. Casale, R. B. Devereux, D. R. Alonso, E. Campo, and P. Kligfield, “Improved sex-specific criteria of left ventricular hypertrophy for clinical and computer
interpretation of electrocardiograms: validation with autopsy findings.,” Circulation, vol. 75, pp. 565–572, Mar. 1987.
[27] C. S. Dangare and S. S. Apte, “Improved study of heart disease prediction system using data mining classification techniques,” International Journal of
Computer Applications, vol. 47, no. 10, pp. 44–48, 2012.
[28] A. Janos and W. Steinbrunn, “Heart disease data set.” http://archive.ics.uci.edu/ml/datasets/Heart+Disease.
[29] U. I. M. L. Repository, “Statlog (heart) data set.” http://archive.ics.uci.edu/ml/datasets/statlog+(heart).
[30] T. Song, X. F. Qu, Y. T. Zhang, W. Cao, B. H. Han, Y. Li, J. Y. Piao, L. L. Yin, and H. Da Cheng, “Usefulness of the heart-rate variability complex for predicting cardiac mortality after acute myocardial infarction,” BMC Cardiovascular Disorders, vol. 14, p. 59, May 2014.
[31] K. Gregory Piatetsky, “Top data science and machine learning methods used in 2017.” https://www.kdnuggets.com/2017/12/top-data-science-machine-learning-methods.html.
[32] G. E. A. P. A. Batista and M. C. Monard, “An analysis of four missing data treatment methods for supervised learning,” Applied Artificial Intelligence,
vol. 17, pp. 519–533, May 2003.
[33] N. S. Altman, “An introduction to kernel and nearest-neighbor nonparametric regression,” The American Statistician, vol. 46, no. 3, pp. 175–185, 1992.
[34] D. Harris and S. Harris, Digital Design and Computer Architecture, Second Edition. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc., 2nd ed.,2012.
[35] D. M. Hawkins, “The problem of overfitting,” Journal of chemical information and computer sciences, vol. 44, no. 1, pp. 1–12, 2004.
[36] S. Han, J. Pool, J. Tran, and W. Dally, “Learning both weights and connections for efficient neural network,” in Advances in neural information processing
systems, pp. 1135–1143, 2015.
[37] A. Kumar, “How to diagnose underfitting/overfitting
of learning algorithm.” https://vitalflux.com/
machine-learning-diagnose-underfittingoverfitting-learning-algorithm/.
[38] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” The Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014.
[39] R. K. Riegelman, Studying a study and testing a test: how to read the medical evidence. Lippincott Williams & Wilkins, 2005.
[40] M. A. Bauml and D. A. Underwood, “Left ventricular hypertrophy: an overlooked cardiovascular risk factor,” Cleve Clin J Med, vol. 77, no. 6, pp. 381–7, 2010.
[41] D. Pewsner, P. Jüni, M. Egger, M. Battaglia, J. Sundström, and L. M. Bachmann, “Accuracy of electrocardiography in diagnosis of left ventricular hypertrophy in arterial hypertension: systematic review,” Bmj, vol. 335, no. 7622, p. 711, 2007.
[42] T. Martin, Y. Bhaskar, and K. Umesh, “Sensitivity and specificity of the electrocardiogram in predicting the presence of increased left ventricular mass index
on the echocardiogram in afro-caribbean hypertensive patients,” West indian medical journal, vol. 56, no. 2, pp. 134–138, 2007.
[43] P. Gosse, E. Jan, P. Coulon, A. Cremer, G. Papaioannou, and S. Yeim, “Ecg detection of left ventricular hypertrophy: the simpler, the better?,” Journal of hypertension, vol. 30, no. 5, pp. 990–996, 2012.
[44] G. Schillaci, F. Battista, and G. Pucci, “A review of the role of electrocardiography in the diagnosis of left ventricular hypertrophy in hypertension,” Journal
of electrocardiology, vol. 45, no. 6, pp. 617–623, 2012.
[45] J. K. Park, J. H. Shin, S. H. Kim, Y.-H. Lim, K.-S. Kim, S. G. Kim, J. H. Kim, H. G. Lim, and J. Shin, “A comparison of cornell and sokolow-lyon electrocardiographic criteria for left ventricular hypertrophy in korean patients,” Korean circulation journal, vol. 42, no. 9, pp. 606–613, 2012.
[46] N. Samesima, L. F. Azevedo, L. D. N. J. De Matos, L. S. Echenique, C. E. Negrao, and C. A. Pastore, “Comparison of electrocardiographic criteria for
identifying left ventricular hypertrophy in athletes from different sports modalities,” Clinics, vol. 72, no. 6, pp. 343–350, 2017.