A Novel Prediction Approach for Myocardial Infarction Using Data Mining Techniques

Transcription

1 A Novel Prediction Approach for Myocardial Infarction Using Data Mining Techniques M. Snehapriya 1, Dr. B. Umadevi 2 1 Research Scholar, 2 Assistant Professor & Head, P.G & Research Department of Computer Science, Raja Doraisingam Govt. Arts College, Sivagangai, TamilNadu, India Abstract: In today s modern world cardiovascular disease is the most lethal one. This disease attacks a person instantly that might create unexpected consequences for the human life. So diagnosing patients correctly on time is the most challenging task for the medical fraternity. The heart disease treatment is quite high and not affordable by most of the patients particularly in India. The research scope is to develop an early prediction treatment using data mining technologies. Now a day every hospital keeps the periodical medical reports of cardiovascular patients through some hospital management system to manage their healthcare. Unfortunately most of the systems rarely use the huge clinical data where vital information is hidden. The huge amount of data in varied forms but this data is seldom visited and remains untapped. So, in this direction lots of efforts are required to make intelligent decisions. The diagnosis of this disease using different features or symptoms is a complex activity. In this paper various data mining technologies are applied to make a proactive approach against failures in early predictions diagnosis of the disease. Keywords: CART, Decision Tree, ANN, Naive Bayes, KNN, Cardio Vascular Disease. 46 I. INTRODUCTION The disease which found genetically in all irrespective age group. The root cause for the disease is neither hereditary nor due to disorders in heart or blood vessels. It is also called as cardiovascular disease (CVD). Cardiovascular disease includes coronary artery diseases (CAD) such as angina and myocardial infarction (commonly known as a heart attack). The Other CVDs[1] include stroke, heart failure, hypertensive heart disease, rheumatic heart disease, cardiomyopathy, heart arrhythmia, congenital heart disease, valular heart disease, carditis, aortic aneurysms, peripheral artery disease, thromboembolic disease, and venous thrombosis. It is identifies that 90% of CVD are preventable. The Prevention involves improving risk factors through: healthy eating, exercise, avoidance of tobacco smoke and limiting alcohol intake. Treating risk factors, such as high blood pressure, blood lipids and diabetes is also beneficial. There are as many researches underwent to make an early prediction. A patient cannot identify himself that he was infected by the cardiovascular disease due to so many reasons. The data mining techniques which provide an analogy about the patients disease status. The paper is structured as follows. The section one describes about introduction. Section 2 deals about background study and its related works. The methodology of the research work is explained in section three. In section four portraits the Experiment results. Finally the paper is concluded in last section. II. BACKROUND STUDY AND RELATED WORKS Numerous studies have been done that have focal point on diagnosis of heart disease. They have used different attributes and applied different data mining techniques for diagnosis and achieved different probabilities for different methods. Rajwant Kaur et al [2] provided a study of different data mining techniques that many risk factors that cause heart disease and it is very difficult to understand and categorized. Most of the time Heart Diseases are detected when a patient reaches at last stage of disease. The Risk Factors help to analyse the disease in advance. They collected 50 patients database and used SVM Classifier with Genetic Algorithms. Recently, the author Moloud Adbar [3] applied and compared data mining techniques to predict the risk of heart diseases. They applied five algorithms including Neural Network, SVM, KNN and Logistic Regression. B.V. Baiju and R. J. Remy Janet [4] used continuous data instead of categorical data. They prefer Naïve Bayes since when the data is high and the attributes are independent of each other. The technique Naïve Bayes is used to find the result. Currently, another study conducted by Yangquan Lyu [5] has been based on the evaluation model of coronary artery disease by using data mining algorithm. The research outcome provides a new dynamic model, which makes it possible to assess lifetime, suggests linear time-invariant approach to assess CHD. The research work done by K.Rajeswari [6] uses the neural network technique for heart disease examination. The examination influences main feature in identifying patients with Ischemic heart disease. Presently, the authors P. Venkatesan and N. R. Yamuna [7] has been made study to compare the decision tree algorithms in classifying tuberculosis patient s response under randomized clinical trial condition. They were used three decision tree approaches such as C4.5, Classification and regression trees (CART) and Iterative dichotomizer 3 (ID3) for the classification of response. Recently, M. Vijayavanan, V. Rathikarani and Dr. P. Dhanalakshmi [8] has been used Probabilistic neural network (PNN) technique to analyse the ECG signal

2 and to capture the distribution of the feature vectors for classification. Data mining has been played an important role in the intelligent medical systems. The relationships of disorders and the real causes of the disorders and the effects of symptom that are spontaneously seen in patients. Knowledge of the risk factors associated with heart disease helps health care professionals to identify patients at high risk of having heart disease. Statistical analysis and data mining technique to help out healthcare professionals in the diagnosis of heart disease. And also the statistical analysis has identified the disorders of the heart and blood vessels, and includes coronary heart disease (heart attacks), cerebrovascular disease (stroke), raised blood pressure (hypertension), peripheral artery disease, rheumatic heart disease, congenital heart disease and heart failure. The major causes of cardiovascular disease are tobacco use, physical inactivity, an unhealthy diet and harmful use of alcohol. The former research works of various authors were identified and classified the root causes of heart disease. They used different data mining techniques for their prediction. 47 III. METHODOLOGY Many risk factors that cause heart disease and it is very difficult to comprehend and categorized. Most of time heart disease is detect when a patient reaches at last stage. The risk factors help to analyze the disease in advance. The system is used to know whether the patient has risk of heart disease or not. The test signals are taken from the arrhythmia database of MIT-BIH [9]. The Figure 1 explains the block diagram of ECG signal process using data mining techniques. In this predictive modeling, there are several tasks are used, which are classification, regression and categorization. The most popular algorithms used to predict the model are Decision tree, Artificial Neural Networks (ANN), Naive Bayes (NB), K-Nearest Neighbor (KNN) and Support Vector Machine (SVM). This section deals with the proposed methodology. The K-Nearest Neighbour (KNN) algorithm is used as the back bone of the method. It is adapted into the process of the predicting the heart disease. The existing and proposed algorithms in predicting the disease will be described in the next section. A. Classification and Regression Trees: Classification and Regression Trees (CART) is a nonparametric decision tree algorithm. It produces either classification or regression trees, based on whether the response variable is categorical or continuous. It is a binary recursive partitioning procedure, which always split the node into only two nodes. The partitioning procedure is repeated for every node of the data until it becomes the terminal node. B. Support Vector Machine: Support Vector Machine (SVM) is a supervised machine learning algorithm which can be used for both classification and regression challenges. However, it is mostly used in classification problems. A step in SVM classification involves identification as which are intimately connected to the known classes[10]. This is called feature selection or feature extraction. Feature selection and SVM classification together have a use even when prediction of unknown samples is not necessary. The key sets are used to distinguish the classes. C. Probabilistic Neural Network: A probabilistic neural network (PNN) is a feed forward probability based neural network. The architecture of the PNN can be extended to any number K of classes. The input layer contains N nodes: one for each of the N input features forming the feature vector. These are fanout nodes such that each input feature node branch to all nodes in the hidden (pattern) layer so that each hidden node receives the complete input feature vector for each data. The hidden nodes are connected to one node called group nodes: one group for each of the K classes. Each node in a single group of hidden layer is corresponding to number of observations that are taken for training. D. Naïve Bayes: Naive Bayes is the basis for many machine-learning and data mining methods. The rule is used to create models with predictive capabilities. It provides new ways of exploring and understanding data. The Naïve Bayes[11] Classifier technique is mainly applicable when the dimensionality of the inputs is high. Despite its simplicity, Naive Bayes is more sophisticated classification methods. E. K-Nearest Neighbour : K-Nearest neighbor (KNN) [12] is a simple and nonparametric classifier. KNN is preferred when all the features are continuous. KNN is also called as casebased reasoning and has been used in many applications like pattern recognition, statistical estimation. Classification is obtained by identifying the nearest neighbor to determine the class of an unknown sample. KNN[12] is preferred over other classification algorithms due to its high convergence speed and simplicity. KNN classification has two stages. The first stage is to find the k number of instances in the dataset that is closest to instance S. The second stage is the k numbers of instances then vote to determine the class of instance S. The Accuracy of KNN depends on distance metric and K value. There are different ways of measuring the distance between two instances such as cosine, Euclidian distance. In order to evaluate the new unknown sample, KNN computes its K Nearest Neighbors and assign a class by majority voting.

3 Step 5 Step 6 : The test feature of the ECG signal will categorized either normal or abnormal. : The data mining algorithms applied over the data to train and test. The performance of the predictive model is through measurement metrics. identified Fig. 1 Block Diagram of the ECG Signal Process Algorithm: Heart Disease Prediction Using KNN: Input : ECG signal from patient record. Output : Predicting the patient ECG signal into normal or abnormal. Step 1 : Read ECG signal as input data. Step 2 : Apply Butterworth filtering techniques for preprocessing to remove the noise from the ECG signal. Step 3 : The Noised input signal is converted into Denoised. Step 4 : The peak and valley internal time for the Denoised signal is calculated. IV. EXPERIMENT RESULTS Any patients who are admitted in the hospitals are undergoing various tests in accordance with their symptoms. In this sequence, heart diseases are also included. Based on the symptoms, as well as the reasons explained, he may expose to ECG subject to the current health condition. In this research data set has been collected from the arrhythmia database of MIT-BIH [13]. The simulations are performed and the results are explored in MATLAB. The ECG report normally helps the doctor to diagnose the status of the heart condition. The graph generated by the ECG machine may leads to some positive decision if it is clear or might be a chance to make an incorrect decision due to failure in the graph structure. In order to make a clear understanding of the graph, it is essential to remove the noise present in the graph Shown in Figure 3 Figure 4. The Figure 3 shows the unclear and clumsy outcome of the patient heart report. This type of report will not provide adequate support for significant decision making. The filtering techniques Butterworth filter is applied over the noisy outcome of the ECG. The filtered signal shows much better than the noisy signal. It is not only supports for understanding but also for decision making. Even though signal is clear but it does not hold that much clarity in its excellence. So it is more essential to denoise the signal which is shown in Figure 5. The denoised signal will help to produce the peak and valley detection shown in figure 6. The test feature is applied over the signal to conclude 48 Fig. 2 Sample EGC dataset whether normal or abnormal signal which is given in Figure 7. If the time duration is of range from 0.06 to 0.10 seconds, it is predict as normal. And if the time duration is of range from 0.08 to seconds it is predict as abnormal. The various data mining [14] techniques are applied to predict the normal and abnormal status for heart disease. The decision making metrics such as accuracy, specificity, error rate, Negative Predicative Value (NPV) and Positive Predictive Value (PPV) are computed for different mining algorithms. The outcomes of the results are given Table 1 for the normal patients and the comparative output is given in Figure 10. Similarly Table 2 contains the abnormal patient s outcomes and comparative output is given in Figure 11. From the Table I and Figure 10, the KNN gives a much better accuracy and also it produces minimum error rate

4 when comparing the other mining algorithms for normal patient. In Table II and Figure 11, the accuracy and error rate of KNN is outperformed among the other data mining algorithms[15][16] for the abnormal patient. Fig.3 Patient ECG Input Signal (Noisy) Fig. 4 Removal of Noise Using Butterworth Filter Fig.5 Denoised ECG Signal (After Applying Butterworth Filter) 49

5 Fig. 6 Peak and Valley Detection for a Denoised ECG Signal Fig.7 Test Feature Fig. 8 ECG is Normal 50

6 Percentage Fig. 9 ECG is Abnormal Table I. Comparison of Algorithms for Normal Signal Algorithm Accuracy Error Rate Sensitivity Specificity PPV (Positive Predictive Value) NPV (Negative Predictive Value) KNN CART SVM PNN NB Algorithm Accuracy Table II. Comparison Of Algorithms For Abnormal Signal Error Rate Sensitivity Specificity PPV (Positive Predictive Value) NPV (Negative Predictive Value) KNN CART SVM PNN NB Normal Signal - Comparison % % 80.00% 60.00% 40.00% 20.00% 0.00% KNN CART SVM PNN NB Algorithms Accuracy Errorrate PPV NPV 51 Fig.10 Comparison of Algorithms for Normal Signal

7 Percentage Abnormal Signal - Comparison % 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% KNN CART SVM PNN NB Algorithms Accuracy Errorrate PPV NPV Fig. 11 Comparison of Algorithms for Abnormal Signal 52 V. CONCLUSION The symptoms of heart disease are familiar for everyone. But no one can predict when it may happen and also the prediction will not give any instance status i.e. normal or abnormal. Our research intension is to notify the patients that neither normal nor abnormal at its early stage. The application of data mining algorithms shows it better performance results among themselves. But the KNN gives much accuracy and minimum error rate rather than comparing with the rest of mining algorithms. Our research concludes that a patient has a great opportunity to know early that he is infected by cardiac disease. VI. REFERENCES [1] Rajwant Kaur, Sukhpreet Kaur, Prediction of Heart disease Based on Risk Factors Using Genetic SVM Classifier, IJARCSSE, Vol 5, Issue 12, December [2] Moloud Adbar, Sharareh R. Niakan Kalhori, Tole Sutikno, Imam Much Ibnu Subroto, Goli Arji, Comparing Performance of Data Mining algorithms in Prediction Heart Diseases, IJECE, Vol 5, No 6, December [3] B. V. Baiju and R. J. Remy Janet, A survey on heart disease Diagnosis and Prediction using Naïve Bayes in Data Mining, IJCET, Vol 5, No.2, April [4] L Yongquiang, H. Jiaming, W. Yiran, Y. Jijiang, T. Yida, W. Wenyao, A. Nazim, Dynamic evaluation model of coronary heart disease for ubiquitous healthcare, Computer Industry, 2015; 69: [5] K. Rajeswari, V. Vaithiyanathan, T. R. Neelakanttan, Feature Selection in Ischemic Heart Disease Identification using Feed Forward Neural Networks, International Symposium on Robotics and Intelligent Sensors 2012, Procedia Engineering, 2012, 41: [6] P. Venkatesan and N. R. Yamuna. Treatment response classification in randomized clinical trials: a decision tree approach, Indian Journal of Science and Technology, 6.1 (2013): [7] M. Vijayavanan, V. Rathikarani and Dr. P. Dhanalakshmi, Automatic Classification of ECG Signal for Heart Disease Diagnosis using morphological features, International Journal of Computer Science & Engineering Technology (IJCSET), ISSN : , Vol. 5, No. 04, Apr [8] R. G. Mark, P. S. Schluter, G.B. Moody, P.H. Devlin, D. Chernoff, An annotated ECG database for evaluating arrhythmia detectors, IEEE Transactions on Biomedical Engineering, 29(8):600, (1982). [9] M. Revathi, Review of Heart Disease Prediction using Data Mining Techniques, IJSTE, Vol 2, Issue 10. [10] R. Rupali Patil, Heart Disease Prediction System using Naive Bayes and Jelinek-mercer smoothing, International Journal of Advanced Research in Computer and Communication Engineering, Vol. 3, Issue 5, May [11] M. A. Jabbar, Prediction of heart disease using k- nearest neighbor and particle swarm optimization, Biomedical Research, 28.9 (2017). [12] G.B. Moody, R. Mark, The MIT-BIH Arrhythmia Database on CD-ROM and software for use with it, Computers in Cardiology, 17: (1990).

8 [13] B. Umadevi D. Sundar, Dr. P. Alli, An Optimized Approach to Predict the Stock Market Behavior and Investment Decision Making using Benchmark Algorithms for Naive Investors, Computational Intelligence and Computing Research (ICCIC), 2013 IEEE International Conference on( IEEE Xplore Digital Library), pg1-5. [14] B. Umadevi, D. Sundar, Dr. P. Alli, A Study on Stock Market Analysis for Stock Selection - Naïve Investors Perspective using Data Mining Technique, International Journal of Computer Applications, ( ), Vol 34 No.3, [15] Dr. B. Umadevi and M. Snehapriya, A Survey On Prediction Of Heart Disease Using Data Mining Techniques, International Journal Of Science and Research (IJSR), Vol 6, Issue 4, April