Detecting Atrial-ventricular blocks Arrhythmia based on -intervals on ECG Signals Makki Akasha abikier, Ibrahim Musa Ishag, Mohammed Izzeldin, Dong Gyu Lee, Gyoyong shon, Keun Ho Ryu Database/ioinformatics Laboratory, Chungbuk National University, Korea {Makki, ibrahim, Mohammed, dglee, gysohn, khryu}@ dblab.chungbuk.ac.kr Abstract his paper describes new a model for analyzing and classifying large amount of ECG data. Our new model is composed of two parts, first part is to pre-process and extract features, the signal is passed through filter part composed of cascaded high-pass and low-pass integer filters to remove the noise, and then extract -intervals from heart patterns. Second part is to classify the patterns that appear in -intervals using decision tree into many classes to discover Atrial igeminy and rigeminy that produce the abnormal patterns and lead to change in intervals. Our new model helps the cardiologist to diagnose the heart disease state based on heart rate variability. 1. Introduction he heart disease is one of the main diseases in the world. here are many ways to discover these heart diseases by using electrocardiography which is important device for recording electrocardiogram (ECG) signals and variability of bioelectric potential with respect to time as human heart beats [1]. ECG is widely used for diagnosis heart activity. Among three components known as P, QRS and waves for typical ECG as they are shown in figure 2. Also it gives us useful information about the functional aspect of the heart. he early detection of the heart diseases can prolong life and enhance the quality of living through appreciates treatment. herefore, we need many techniques and working analysing on ECG signals. he state of cardiac health is generally reflected in the shape of the ECG waveform and heart rate. It contains important pointers to nature of the disease attacking the heart. Since the biosignals are nonstationary signals, this reflection may occur randomly at the time scale, in this case, diseases symptoms may not show up all the time but would manifest at certain irregular intervals patterns during the day. he analysing and reorganization of ECG signals is the main process for diagnosing heart conditions. here are approaches have been proposed and developed for analysing ECG signals such as digital signal analysis, fuzzy logic methods, artificial neural network, hidden markov model,support vector machine and other methods, each one of these has advantages and disadvantages [1,2]. he major of these methods still need artificial assistance for improving the accuracy and trying to look for more excellent method. We proposed new model for analysing and classifying long-term ECG data, helps the cardiologists in diagnosing process by classifying the patterns that appear during the intervals using incremental decision ree. 2. Cardiac arrhythmia ECG or EKG (sometimes) provides the most accurate means of identifying cardiac arrhythmia (bad rhythm) which can be diagnosed easily when we understand electrophysiology of the heart which includes normal conduction pathways. Normally sanio-atrial (SA) node generates regular sinus rhythm that paces the heart. Each pacemaker impulse from SA node spreads through both atria as an advancing wave of depolarization. On EKG there is a consistent distance (duration) between similar waves during regular a rhythm because the SA node s automaticity precisely maintains constant cycle duration between the pacing impulse that it generates as shown in figure 1 [3].
Figure 1.Shows equal distances between identical waves during normal and regular cardiac rhythm. he depolarization of the atrial myocardium produces P wave on EKG.he atrial depolarization eventually reaches the atrial-ventricle (AV) node but the conduction of depolarization slows within AV node recording a pause on EKG. After passing slowly through the AV node, depolarization proceeds rapidly through the bundle, bundle branches and their subdivisions, and through the terminal purkinje filaments to distribute depolarization to the ventricles. Ventricular depolarization produces a QRS complex on EKG.Ventricular depolarization initiates ventricular contraction, which persists (through both phases of depolarization) to the end of the wave as shown in Figure 2. Figure 2. ECG waves that are produced by the heart activities. he arrhythmia can be divided into number categories according to the arrhythmia s mechanism of origin irregular rhythm, escape, premature and tauchy-arrhythmias [3]. In this paper we are going to discuss about premature beats. he premature beat originates in irritable automaticity focus that fires spontaneously, producing beat on EKG earlier than expected in the rhythm. We need to identify the focus (atrial, junctional, or ventricular). Premature atrial beat (PA) originates suddenly in an irritable atrial automaticity focus, and it produces a P wave earlier than expected. he ventricular conduction system is usually receptive to depolarize by a premature atrial beat, but one of the bundle branches may not have completely repolarised which causes a little refractory when the other is receptive. his aberrant ventricular conduction produces a slightly widened QRS for that premature cycle as it is shown in figure 3. Figure 3. Premature Atrial eat with aberrant ventricular conduction. When the premature atrial beat (p ) is conducted to the ventricles, the ventricles are also depolarized earlier than usual. his event produces change on -intervals. Occasionally, an irritable automaticity focus fires a premature Atria eat (P ) that couples to the end of a normal cycle, and repeats this process by coupling a PA to the each successive normal cycle. his is called atrial bigeminy as it is shown in figure 4(a). Sometimes, an irritable atrial focus may prematurely fire after two normal cycles, when this couplet repeats continuously, this is called atrial triggering as it is shown in figure 4(b). Figure 4. Shows atrial bigeminy and trigeminy he atrial bigeminy and trigeminy occur with type atrial-ventricular blocks (arrhythmias). We can classify this type of arrhythmia based on the heart rate variability because they affect the heart rate variability by occurring specific patterns on heart rate. hese patterns occur on different forms based on the heart disease state as are shown in figure 4. y detecting the R-peaks and calculating the -intervals that help us to diagnose the heart disease state using data mining techniques [4].
he normal heart rate variability must be changed in specific range [3, 7]. We suppose the normal heart variability change between α, β. α < normal Rate > β. he Atrial bigeminy beat is produced according to some failure occur in SA node, then atrial automaticity foci fire spontaneously and earlier more than expected and produce changing in heart rate variability in that time. Enhance, atrial bigeminy produces patterns like this: <Normal beat, abnormal beat, normal beat >or <Abnormal beat, normal beat, abnormal beat > Abnormal beat = interval is less than α, as it shown in figure 4 (a). If ((1 is normal) && (2 is abnormal) && (3 is normal)) if ((1 is abnormal )&&(2 is normal )&&(3 is abnormal )) Atrial bigeminy. If ((1 is normal) && (2 is normal) && (3 is abnormal)) If ((1 is abnormal) &&(2 is abnormal )&&(3 is normal )) Atrial trigeminy. When the patterns occurrences satisfy the above condition, the patterns should be classify as atrial bigeminy or trigeminy as it is shown in table 1. able 1. Shows how the bigeminy and trigeminy appear during the - intervals. 3. Proposed Method Filtering and Features Extractions - interval patterns - intervals Patterns Classification Using Decision ree Result Output Diagnosing
Figure 5. Shows the proposed method 3.1 Data Collection he first step in our proposed method begins with collecting ECG signals from human bodies using leads. he standard ECG is composed 12 leads, six limb leads are recorded by using arm and leg electrodes, and other six chest leads are recorded using electrodes at six different positions on chest [3].he heart sound using electrocardiogram will record simultaneously from patients and store in ECG database which will be used as input signal for features extraction part processing. We will use the storage technique that is similar to techniques used in physionet databases[8]. 3.2 Filtering and Features extractions It s considered as preprocessing the ECG signal which is contaminated with high frequency noise. he unwanted noise of the heart biopotential signal must be remove. ECG must also be filtered using bandpass filtered to eliminate the motion artifact, baseline wander and eliminate power line noise which can affect the QRS detection algorithm [2].In the features extractions and - intervals from ECG, our proposed method is going to extract only R wave. here are many algorithms and methods that are used for detection R wave from ECG signals [5].he -interval signal will be constructed by measuring the time interval between successive R waves. he proposed method stores the - intervals in file or dataset that could be used in classification step. his model is going to classify the igeminy and rigeminy that produce the abnormal patterns and lead to change in - intervals. 3.3 Classification Using Decision ree A decision tree is a representation of a decision procedure for determining a class label. G G 1 Pattern window Yes no Yes no yes no Yes no yes no yes no yes no Diagnosis Figure 6. Shows decision tree structure that is use to classify the ECG patterns. Where is interval time. At each internal node of the tree, there is a test (question, is -interval normal?), and a
branch corresponding to each of the possible outcomes of the test. At each leaf node, there is a class label (answer) where G is normal patterns, is trigeminy and is bigeminy as it is shown on leaf nodes in fig 6.Our proposed method is using decision tree to classify the -interval patterns into many classes to discover atrial bigeminy and trigeminy which generate abnormality patterns [6]. Each one of these patterns consists of three intervals sliding window. he classification in our proposed method on intervals is going to perform the classification using set of rules that will be provided by medical experts [7]. 4. Conclusion & Future Work In this paper we proposed new method to detect atrial bigeminy and trigeminy which may occur with type atrialventricular blocks (arrhythmias) and lead to generate abnormal patterns in heart activity. In the future, we are going to complete implementation of this proposed method and developing to help the cardiologist to diagnose the heart disease state based on heart rate variability. 5. Acknowledgement his work was supported by the grant of the Korean Ministry of Education, Science and echnology"(he Regional Core Research Program/Chungbuk I Research-Oriented University Consortium) and the Korea Science and Engineering Foundation (KOSEF) grant funded by the Korea government (MES) (No. R11-2008-014-02002- 0). 6. References [1] Fei Zhang. Jun an, Yong Lain. Jones, An Effective QRS Detection algorithm for Wearable ECG in ody Area Network, IEEE, 2007. [2] Emran M, N.HK,R.salleh, M.Yamani Idna Idris. N.Noor, A Mohd amil. Heartbeat Electrocardiogram (ECG) Signal Feature Extraction Using Discrete Wavelet ransforms (DW), Malaysia, 2008. [3] Dale Dubin, MD, Rapid Interpretation of EKG s, USA, 2001. [4]PANG-NING AN MiCHAELSV KUMAR, Introduction to Data mining, USA, 2006. [5] Jing-ain ang, Xiao-li Yang, Jun-chao Xu,Yan ang,qing Zou, he Algorithm of R peak Detection in ECG ased on Emprical Mode Decomposition, IEEE, 2008. [6] Paul E Utgo, an improved algorithm for incremental induction of decision trees eleventh international conference on machine learning, February, February 1994. [7] M.G. sipourasa, c, D.I. Fotiadisa, c, d,*, D. Siderisb,d An arrhythmia classification system based on the -interval signal, Artificial Intelligence in Medicine, Germany, 2005. [8]http://www.physionet.org/physiobank/database/mitdb/.