本論文提出一心臟衰竭疾病的快速辨識系統，基於快速篩檢的目的，我們提出了使用短時間的分析方法，其中分析的資料長度為五~十五分鐘。
我們的辨識系統主要從兩個方法來分析，一是支持向量機(SVM)、二為多輸入神經網路(MINN)。支持向量機為過去相關的研究常用之技術，多輸入神經網路為此研究新提出的方法。
第一個方法主要可以分為三個部分: 特徵擷取、特徵選取、及分類。我們使用了短時間的心律變異度分析，得到時域及頻域的線性特徵，此外、還有ㄧ些非線性的分析方法如多尺度熵分析、去趨勢波動分析等。為了進一步找出這些特徵與疾病的關聯性，我們使用了統計方法來驗證每一個特徵在充血性心衰竭與對照組間是否有顯著差異。在排除ㄧ些對於辨識沒有充分有效的特徵以及歧異點後，我們使用了前饋式特徵選取方法選出最佳的特徵組合，最後運用支持向量機的原理來尋找最好的支持超平面，分類健康與心衰竭的病患。
第二個方法同樣可分為三個部分: 資料前處理、模型建置、及分類。由第一個方法得到的啟發，我們發現兩組心電圖資料在時域及頻域上的特徵皆有顯著的差異性，因此我們不僅使用時域的心跳間期訊號，也將其作傅立葉變換取出頻域訊號作為神經網路的輸入端，並且使用不同的神經網路如稀疏自動編碼器(SAE)、長短期記憶網路(LSTM)、卷積神經網路(CNN)建置模型，最後再由多輸入神經網路架構建立輸入端，對模型進行訓練，並得到分類結果。本篇論文中也探討了支持向量機及神經網路的結果與比較，也比較了不同的資料長度的表現，也使用了麻省理工學院(MIT)的資料庫做進一步的測試，最後得到以卷積神經網路的架構下，在五百點(大約五~七分鐘)的心跳間期可達到訓練集準確率 93.36%，測試集準確率為 86.74%。

In this study, a fast detection system for congestive heart failure (CHF) based on multiple input neural network was proposed. To rapid screen, we offered a short-term analysis method to achieved this goal, which data size was 5 to 15 minute ECG data.

The proposed detection system was compared by two approaches: one was support vector machine (SVM), the other was multiple input neural network (MINN). The SVM method was majority applied in other related researches, and The MINN method was the novel approach that was first proposed in this thesis.

First approach was composed of three portions: feature extraction, feature selection, and classification. We used short-term heart rate variability (HRV) analysis to extract time and frequency domain features. Some non-linear methods were also used in this thesis such like multiple entropy analysis (MSE), and de-trended fluctuation analysis (DFA). Furthermore, we used statistical methods to verify whether there existed significant differences in these features between CHF and Control groups. We chose the sequential forward selection (SFS) to search for the best feature subset in the short-term analysis. Afterwards, the SVM was applied to search for the best support hyperplanes in order to classify CHF and Control groups.

Second approach could be composed of three portions similarly: data pre-processing, model-building, and classification. We employed deep learning neural network to train the features by input RR interval signal in both time and frequency domain. We used different type of neural network such as sparse-auto-encoder (SAE), long-short-term memory (LSTM), and convolution neural network (CNN) to build the model. Afterwards, through the multiple input neural network structure to train the model and get the classification result.

The comparisons of two approaches were discussed in this thesis, also we compared different data length performance. To verify the classification ability in other database, MIT-BIH database were also applied to test our system. With our current model, the recognition accuracy between CHF and Control is up to 93.36\% for training set, and 86.74\% for testing set.

1 Introduction 1
1.1 Backgrounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Pathological Knowledge and Methodologies 7
2.1 Literature Survey and Comparison . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Pathological Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Cardiac Arrhythmias . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Congestive Heart Failure (CHF) . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Overview of ECG signal . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Clinical Database Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.1 Clinical Data Source . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.2 Database Information . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Analysis Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 Heart Rate Variability (HRV) . . . . . . . . . . . . . . . . . . . . . . 23
2.4.2 Linear Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4.3 Non-Linear Features . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4.4 Neural Network Algorithm . . . . . . . . . . . . . . . . . . . . . . . 34
3 Proposed Methodologies for Congestive Heart Failure Recognition System 37
3.1 Data Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.1.1 Prerequisite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.1.2 RR Interval Extraction . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Machine Learning (ML) Approach . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.1 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Feature Normalization . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2.3 Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.4 Classification with SVM . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3 Multi-input Neural Network (MINN) Approach . . . . . . . . . . . . . . . . 55
3.3.1 Model Establishment . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.2 Multi-input Configuration . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.3 Classification with MINN Model . . . . . . . . . . . . . . . . . . . 60
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4 Implementation Results and Discussion 65
4.1 Classification with SVM Model . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1.1 Statistical Results for NTUH Database . . . . . . . . . . . . . . . . . 67
4.1.2 Receiver Operating Characteristic (ROC) Analysis for NTUH Database 68
4.1.3 Classification Results for NTUH Database . . . . . . . . . . . . . . . 72
4.2 Classification with MINN Model . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.1 Feature Extraction for MINN Model . . . . . . . . . . . . . . . . . . 74
4.2.2 Overall System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3 MIT-BIH Database Experiments . . . . . . . . . . . . . . . . . . . . . . . . 83
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.4.1 HRV and Non-linear Measurement Analysis . . . . . . . . . . . . . 85
4.4.2 Comparison Between Two Approaches in This Study . . . . . . . . . 85
4.4.3 Classifier Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.4.4 CHF Detection Comparison with Literature . . . . . . . . . . . . . . 87
5 Conclusion and Future Works 91
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.2 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

[1] F. G. Yanowitz, “ECG learning center,” University of Utah School of Medicine,
[Online]. Available: http://ecg. utah. edu/.[Accessed 27 September 2012], 2001.
[2] G. M. Khan, “A new electrode placement method for obtaining 12-lead ECGs,”
Open heart, vol. 2, no. 1, p. e000226, 2015.
[3] C. G. C. U. National, “Chronic heart failure: National clinical guideline for diagnosis
and management in primary and secondary care: Partial update,” 2010.
[4] P. Malmivuo, J. Malmivuo, and R. Plonsey, Bioelectromagnetism: principles and applications
of bioelectric and biomagnetic fields. Oxford University Press, USA, 1995.
[5] M. Marouf, “High frequency noise approximation and adaptive reduction in the
ECG signals,” Ph.D. dissertation, 2018.
[6] J. E. Hall, Guyton and Hall textbook of medical physiology e-Book. Elsevier Health
Sciences, 2015.
[7] S.-D. Wu, C.-W. Wu, K.-Y. Lee, and S.-G. Lin, “Modified multiscale entropy
for short-term time series analysis,” Physica A: Statistical Mechanics and its
Applications, vol. 392, no. 23, pp. 5865 – 5873, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0378437113007061
[8] S.-D. Wu, C.-W. Wu, S.-G. Lin, K.-Y. Lee, and C.-K. Peng, “Analysis
of complex time series using refined composite multiscale entropy,” Physics
Letters A, vol. 378, no. 20, pp. 1369 – 1374, 2014. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0375960114002916
[9] Chabacano, “Diagram showing overfitting of a classifier,” 2008. [Online]. Available:
https://commons.wikimedia.org/wiki/File:Overfitting.svg#filelinks
[10] “Cardiovascular diseases (CVDs),” Sept. 2016. [Online]. Available: http://www.who.
int/mediacentre/factsheets/fs317/en/.
[11] “Latest statistics show heart failure on the rise; cardiovascular
diseases remain leading killer,” January 2017. [Online]. Available:
http://newsroom.heart.org/news/latest-statistics-show-heart-failure-on-the-rise;
-cardiovascular-diseases-remain-leading-killer
[12] “Congestive heart failure (CHF) diagnosis,” Sept. 2015. [Online]. Available:
http://www.healthcommunities.com/congestive-heart-failure/diagnosis.shtml
[13] M.-Y. Lee and S.-N. Yu, “Selection of
failure recognition using support vector machine-based criteria,” in 5th European
Conference of the International Federation for Medical and Biological Engineering.
Springer, 2011, pp. 400–403.
[14] L. Pecchia, P. Melillo, M. Sansone, and M. Bracale, “Discrimination power of short-term
heart rate variability measures for chf assessment,” IEEE Transactions on Information
Technology in biomedicine, vol. 15, no. 1, pp. 40–46, 2011.
[15] K. H. Yoon, T. Thap, C. W. Jeong, N. H. Kim, S. Noh, Y. Nam, and J. Lee, “Analysis
of statistical methods for automatic detection of congestive heart failure and atrial fibrillation
with short rr interval time series,” in 2015 9th International Conference on,
Innovative Mobile and Internet Services in Ubiquitous Computing (IMIS). IEEE, 2015,
pp. 452–457.
[16] W. Chen, G. Liu, S. Su, Q. Jiang, and H. Nguyen, “A CHF detection method based on
deep learning with rr intervals,” in 2017 39th Annual International Conference of the
IEEE, Engineering in Medicine and Biology Society (EMBC). IEEE, 2017, pp.
3369–3372.
[17] A. L. Goldberger, L. A. 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, pp. e215–e220, 2000.
[18] M. A. Gatzoulis, S. Balaji, S. A. Webber, S. C. Siu, J. S. Hokanson, C. Poile, M. Rosenthal,
M. Nakazawa, J. H. Moller, P. C. Gillette et al., “Risk factors for arrhythmia and
sudden cardiac death late after repair of tetralogy of fallot: a multicentre study,” The
Lancet, vol. 356, no. 9234, pp. 975–981, 2000.
[19] D. Conen, M. Adam, F. Roche, J.-C. Barthelemy, D. F. Dietrich, M. Imboden,
N. K¨unzli, A. von Eckardstein, S. Regenass, T. Hornemann et al., “Premature atrial
contractions in the general population: frequency and risk factors,” Circulation, pp.
CIRCULATIONAHA–112, 2012.
[20] J. M. Sims and V. Miracle, “Premature ventricular contractions,” Nursing, vol. 27, no. 8,
p. 43, 1997.
[21] P. A. Wolf, R. D. Abbott, and W. B. Kannel, “Atrial fibrillation as an independent risk
factor for stroke: the framingham study.” Stroke, vol. 22, no. 8, pp. 983–988, 1991.
[22] “Congestive heart failure - systolic topic review.” [Online].
Available: http://www.healio.com/cardiology/learn-the-heart/cardiology-review/
topic-reviews/systolic-congestive-heart-failure#Pathophysiology
[23] J. W. Hurst, “Naming of the waves in the ECG, with a brief account of their genesis,”
Circulation, vol. 98, no. 18, pp. 1937–1942, 1998.
[24] “Target heart rates,” 10 2016. [Online]. Available:
//www.heart.org/HEARTORG/HealthyLiving/PhysicalActivity/FitnessBasics/
Target-Heart-Rates UCM 434341 Article.jsp#.WQ OEVV97IV
[25] E. Burns, “ECG library: Comprehensive, free educational resource covering over 100
ecg topics relevant to emergency medicine and critical care,” 2010. [Online]. Available:
http://lifeinthefastlane.com/ecg-library/.
[26] V. R., J. A., K. T., H. H., and V.-P. L. M., “Reduced heart rate variability in hypertension:
associations with lifestyle factors and plasma renin activity,” J. Hum. Hypertens., vol. 17,
p. 171, 2003.
[27] H.-H. Osterhues, G. Großmann, M. Kochs, and V. Hombach, “Heart-rate variability
for discrimination of different types of neuropathy in patients with insulin-dependent
diabetes mellitus,” Journal of Endocrinological Investigation, vol. 21, no. 1, pp. 24–30,
1998. [Online]. Available: http://dx.doi.org/10.1007/BF03347282
[28] H. L. K. MD and MPH, “Beta blockade, ventricular arrhythmias, and sudden cardiac
death,” The American Journal of Cardiology, vol. 80, no. 9, Supplement 2, pp.
29J – 34J, 1997. [Online]. Available: http://www.sciencedirect.com/science/article/pii/
S0002914997008369
[29] M. Kupari, J. Virolainen, P. Koskinen, and M. J. Tikkanen, “Short-term heart
rate variability and factors modifying the risk of coronary artery disease in a
population sample,” The American Journal of Cardiology, vol. 72, no. 12, pp. 897
– 903, 1993. [Online]. Available: http://www.sciencedirect.com/science/article/pii/
000291499391103O
[30] M. G. Kienzle, D. W. Ferguson, C. L. Birkett, G. A. Myers, W. J. Berg,
and D. Mariano, “Clinical, hemodynamic and sympathetic neural correlates
of heart rate variability in congestive heart failure,” The American Journal
of Cardiology, vol. 69, no. 8, pp. 761 – 767, 1992. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/000291499290502P
[31] “Heart rate variability,” Electrophysiology, Task Force of the European Society of
Cardiology the North American Society of Pacing, vol. 93, no. 5, pp. 1043–1065, 1996.
[Online]. Available: http://circ.ahajournals.org/content/93/5/1043
[32] U. Rajendra Acharya, K. Paul Joseph, N. Kannathal, C. M. Lim, and J. S.
Suri, “Heart rate variability: a review,” Medical and Biological Engineering
and Computing, vol. 44, no. 12, pp. 1031–1051, 2006. [Online]. Available:
http://dx.doi.org/10.1007/s11517-006-0119-0
[33] “Heart rate variability,” 2016. [Online]. Available: https://www.firstbeat.com/en/
science-and-physiology/heart-rate-variability/.
[34] E. von Borell, J. Langbein, G. Despr´es, S. Hansen, C. Leterrier, J. Marchant-Forde,
R. Marchant-Forde, M. Minero, E. Mohr, A. Prunier, D. Valance, and I. Veissier, “Heart
rate variability as a measure of autonomic regulation of cardiac activity for assessing
stress and welfare in farm animals — a review,” Physiology & Behavior, vol. 92,
no. 3, pp. 293 – 316, 2007, stress and Welfare in Farm Animals. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0031938407000157
[35] B. M. Saykrs, “Analysis of heart rate variability,” Ergonomics, vol. 16, no. 1, pp. 17–32,
1973.
[36] I. S. Uzun, M. H. Asyali, G. Celebi, and M. Pehlivan, “Nonlinear analysis of heart rate
variability,” in Proceedings of the 23rd Annual International Conference of the IEEE,
vol. 2. Engineering in Medicine and Biology Society, 2001. IEEE, 2001, pp. 1581–
1584.
[37] M. A. Busa and R. E. van Emmerik, “Multiscale entropy: A tool for understanding the
complexity of postural control,” Journal of Sport and Health Science, vol. 5, no. 1, pp.
44–51, 2016.
[38] H. Kasturiwale and S. N. Kale, “Influence of physiological factors on heart rate variability
in predicting cardiovascular diseases,” in 2017 2nd International Conference on
Communication and Electronics Systems (ICCES), Oct 2017, pp. 775–780.
[39] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy analysis of biological
signals,” Physical review E, vol. 71, no. 2, p. 021906, 2005.
[40] S. M. Pincus, “Approximate entropy as a measure of system complexity.” Proceedings
of the National Academy of Sciences, vol. 88, no. 6, pp. 2297–2301, 1991.
[41] J. W. Kantelhardt, E. Koscielny-Bunde, H. H. Rego, S. Havlin, and A. Bunde, “Detecting
long-range correlations with detrended fluctuation analysis,” Physica A: Statistical
Mechanics and its Applications, vol. 295, no. 3-4, pp. 441–454, 2001.
42] H. Sak, A. Senior, and F. Beaufays, “Long short-term memory recurrent neural network
architectures for large scale acoustic modeling,” in Fifteenth annual conference of the
international speech communication association, 2014.
[43] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional
neural networks,” in Advances in neural information processing systems, 2012,
pp. 1097–1105.
[44] C.-C. Chang and C.-J. Lin, “LIBSVM: A library for support vector machines,” ACM
Transactions on Intelligent Systems and Technology, vol. 2, pp. 27:1–27:27, 2011, software
available at http://www.csie.ntu.edu.tw/cjlin/libsvm.
[45] F. Chollet et al., “Keras,” https://github.com/fchollet/keras, 2015.
[46] M. A. Peltola, “Role of editing of r–r intervals in the analysis of heart rate variability,”
Frontiers in Physiology, vol. 3, p. 148, 2012.
[47] W. Navidi, Statistics for Engineers and Scientists, 4th ed. McGraw-Hill Education,
2015.
[48] N. Nachar, “The mann-whitney u: A test for assessing whether two independent samples
come from the same distribution,” Tutorials in Quantitative Methods for Psychology,
vol. 4, no. 1, pp. 13–20, 2008.
[49] M. P. Fay and M. A. Proschan, “Wilcoxon-mann-whitney or t-test? on assumptions for
hypothesis tests and multiple interpretations of decision rules,” Statistics surveys, vol. 4,
p. 1, 2010.
[50] A. Marcano-Cedeno, J. Quintanilla-Dom´ınguez, M. Cortina-Januchs, and D. Andina,
“Feature selection using sequential forward selection and classification applying artificial
metaplasticity neural network,” in IECON 2010-36th Annual Conference on IEEE
Industrial Electronics Society. IEEE, 2010, pp. 2845–2850.
[51] A. Marcano-Cede˜no, J. Quintanilla-Dom´ınguez, M. G. Cortina-Januchs, and D. Andina,
“Feature selection using sequential forward selection and classification applying artificial metaplasticity neural network,” in IECON 2010 - 36th Annual Conference on IEEE
Industrial Electronics Society, Nov 2010, pp. 2845–2850.
[52] D. P. Muni, N. R. Pal, and J. Das, “Genetic programming for simultaneous feature selection
and classifier design,” IEEE Transactions on Systems, Man, and Cybernetics, Part
B (Cybernetics), vol. 36, no. 1, pp. 106–117, Feb 2006.
[53] T. Mitchell, “Computational learning theory,” in Machine Learning. McGraw-Hill
Science/Engineering/Math, 1997, ch. 7, pp. 201–229.
[54] Shawe-Taylor, An Introduction to Support Vector Machines and Other Kernel-based
Learning Methods. Cambridge University Press, 2000.
[55] S. Tong and D. Koller, “Support vector machine active learning with applications to text
classification,” J. Mach. Learn. Res., vol. 2, pp. 45–66, Mar. 2002. [Online]. Available:
http://dx.doi.org/10.1162/153244302760185243
[56] M. Mark H. Ebell M.D., Evidence-Based Diagnosis A Handbook of Clinical Prediction
Rules. Springer New York.
[57] 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, pp. e215–e220, 2000. [Online]. Available:
http://circ.ahajournals.org/content/101/23/e215
[58] D. C. De Jong, Maj Marla J. Randall, “Heart rate variability analysis in the assessment
of autonomic function in heart failure,” Journal of Cardiovascular Nursing, vol. 20, pp.
186–195, 2005.