ECG signal analysis for detection of Cardiovascular abnormalities and Ischemic episodes

Goutam Kumar sahoo
ECE Department NIT Rourkela, Odisha India-769008 Email:

Samit Ari
ECE Department NIT Rourkela, Odisha India-769008 Email:

Sarat Kumar Patra
ECE Department NIT Rourkela, Odisha India-769008 Email:

Abstract —Electrocardiogram (ECG) is generally used for diagnosis of cardiovascular abnormalities and heart disorders. An efficient method for analyzing the ECG signal towards the detection of cardiovascular abnormalities and ischemic episodes follows mainly five stages: pre-processing, feature extraction,cardiac abnormality detection, beat classification and ischemic episode recognition.The detection of cardiovascular abnormalities like bradycardia and tachycardia is based on the calculation of heart rate(HR) from the extracted ECG features.The extracted ST-segment and T-wave features are used for detection of ischemic episodes.The ability of the method was tested on European ST-T database. The performance of ischemic episode detection shows 88.08% sensitivity (Se) and 92.42% positive predictive accuracy (PPA). Index Terms—ECG, Pre-processing, QRS-complex, Heart rate(HR), ST-segment deviation, Ischemic episode detection.

I. Introduction The electrocardiogram (ECG) is a graphical recording of the electrical signals generated by the heart [1].A typical ECG recording consists of P-wave, QRS-complex, T-wave and U-wave. The amplitude and interval of different waves provides a measure for detection of various cardiac abnormalities and disorders. The most important interval used for determination of cardiovascular abnormality condition is RR-interval(distance between two R-peaks).The normal RR-interval is 600-1200ms.The key to Ischemic episodes detection is ST-segment deviation and T-wave.The normal ST-segment and T-wave duration is 80-120ms and 120160ms respectively [2]. Bradycardia, Tachycardia, Bundle Branch Block, Premature Ventricular Contraction, Wolff-Parkinson-White syndrome (WPW), etc., are the most common cardiovascular abnormalities [3].The cardiovascular abnormalities like bradycardia and tachycardia can be identified by calculating the heart rate (HR). The HR is calculated from the extracted features of ECG signal. The HR value is compared with the normal range of HR i.e.60-120BPM [3]to detect cardiovascular abnormalities. Ischemia (heart stroke) is a cardiovascular disorder which affects the heart and the blood vessels. The coronary

arteries become narrowed by atherosclerosis which restricts the flow of blood and oxygen to the heart and brain. This makes brain cells to die which creates cardiac disorder known as ischemia[4]. The detection process takes more time if analyzed by doctor using long duration ECG data. So an automatic technique is necessary for early detection of ischemia. The ECG beat classification is essential for automatic detection and diagnosis of ischemic episodes in a long duration electrocardiogram. The key to ischemic episodes detection is the ST-segment deviation and Twave amplitude changes [5]. Most importantly the parameter, ST-segment deviation is expressed as polarity change relative to isoelectric line. The isoelectric line is the baseline, typically measured between the T-wave offset and the preceding P-wave onset of electrocardiogram. Isoelectric line is used as a reference for measurement of ST-segment deviation [6]. Abnormal Twave, usually very tall or inverted, appears corresponding to the elevation or depression in ST-segment. The STsegment is considered as elevated if the segment is 0.08mV or more above the isoelectric line. The T-wave inversion or flattening is measured using first 30s of the ECG recording [7]. The methodology followed to detect cardiovascular abnormalities and ischemic episodes basically consists of five stages. In first stage the ECG recording is pre-processed to achieve noise removal.In the next stage, ECG feature extraction is carried out to locate the changes in QRScomplex, ST-segment and T-wave. Thereafter heart rate is calculated to identify various cardiovascular abnormalities.In the next stage beat is classified as normal or ischemic using some rule based on medical knowledge and the final stage provides the identification of ischemic episode which is based on the detection of two or more consecutive ischemic windows using first 30s of each ECG recording. The performance of the ischemic episode detection technique is evaluated in terms of sensitivity and positive predictive accuracy. Rest of the paper is organized as follows. Section II explains proposed methodology. Section III provides the experimental result using the European Society of Cardiology (ESC) ST-T database, Section IV provides conclusion.

TON and TOF F locations. In the normalization process the amplitude of the signal is normalized first and then it is passed through band pass filter. The base line is chosen as the flat line between P-wave and Q-wave. J-point location. ST-segment is located at 80ms after J-point when cardiac rhythm is less than 120BPM and 60ms after J-point when the cardiac rhythm is more than 120BPM [10]. The suitable method used for finding QRS-complex is based on J.2.5 2 second 2. 2. 1. First feature towards the ischemic episode detection process is the determination of QRS-complex.5 4 Fig. 2) J-point Location: The J-point is the junction between the QRS-complex and the ST-segment of ECG signals [6]. Preprocessing stage involves normalization and filtering. TON and TOF F is then estimated by considering 35ms duration from left and right of the Twave peak respectively[1].5 4 1 Amp 0 −1 0 0. Raw ECG Signal Cardiovascular abnormality Detection Ischemic episode Detection Fig. 4) Isoelectric line and ST segment Location: Isoelectric line is the baseline or almost zero amplitude level. isoelectric line and ST-segment location were detected using the previously located Q-. Also it is the first point where the waveform flattens out to the right after QRS-complex. Also P-wave is the first positive peak on the left of tall R-peak location [1]. ECG signal after normalization and filtering B. The function of band pass filter is noise rejection. T-wave. Two points are identified by considering start and end location of zero amplitude points from peak location of P-wave respectively.5 second Normalized ECG Signal 3 3. 60Hz interference.5 1 1.5 second Band pass filtered ECG Signal 1. ECG signal extracted wave peaks and points with baseline .[8].The block diagram for the proposed method is shown in Fig. TON and TOF F Location: After knowing the location of R-peak and J-point.e. 1) P-Wave Location: A threshold level is used with reference to the normal range of ECG segments to locate P-wave. The location for isoelectric line was estimated by finding the start and end point of all zero amplitude ECG level.5 2 2. The desirable pass-band to maximize the QRS energy is approximately 5-15 Hz[9].First is the detection of cardiovascular abnormalities and then the detection of ischemic episodes. 3.II. the peak of T-wave is estimated as maximum elevation between R-peak+400ms and J-point+80ms. Fig. Block diagram of ischemic episode detection A. The normal PR-interval ranges from 120-200ms whereas the amplitude and duration of P-wave are 0. Then the lowest peak ’S’ is determined followed by Q-peak. A flat line is drawn between these two points called as isoelectric line. baseline wander and T-wave interference. The raw ECG signal. Input ECG Signal −500 Amp −1000 −1500 0 0.5 3 3. Pre-processing The pre-processing of raw ECG signal is required for removal of noises such as muscle noise. Other features like P-wave location. the signal after normalization and the normalized signal after passing through band pass filtering are shown in Fig. Tompkins algorithm [9]. The J-point location is normally at the end of QRS-complex which has the normal range of 80-120ms [2]. R. All the extracted ECG features are as shown in Fig.1. Feature Extraction The variations in ECG features like QRS-complex morphology and ST-segment deviation and T-wave alternation can be used for the detection of ischemic episodes.and Swave peaks.5 4 1 Amp 0 −1 0 0.5 1 1. etc.5 3 3. PROPOSED METHODOLOGY The methodology basically describes two different processes.3.25mV and 80ms respectively. The tallest peak i. Pre-processing Feature Extraction Beat Classification R-peak is first determined in QRS-complex. 3) T-Wave.5 1 2 2. Pan and J.

The ischemic window is detected if the 30s window contains more or equal to 75% of ischemic beats. The cardiovascular abnormalities i. HR is expressed in beats per minute (BPM)and the normal range of HR is 60-120BPM [3].The calculated HR value is compared with normal range to detect the cardiovascular abnormalities like bradycardia and tachycardia. The ischemic beat classification method in [7] is used without considering the angle of deviation (in degree) as follows.The ischemic episode detection performance is measured in terms of sensitivity (Se) and positive predictive accuracy (PPA). of consecutive ischemic window ≥ 2) then Ischemic episode is identified end if The complete flowchart of Cardiovascular abnormality and Ischemic episode detection process is given in Fig. A sliding . Cardiovascular abnormalities detection The cardiovascular abnormalities detection is based on heart rate (HR) calculation. The formula used to calculate heart rate (HR) is Heart Rate = [360/ (RR-interval in samples)] * 60 beats/min [1].08mV) (or) (ST-segment ≥ 0. The ischemic window identification and ischemic episode recognition process is summarized below. Algorithm 3 Ischemic episode detection if ([(No. Here the ischemic episode is calculated for the presence of at least two or more number of consecutive ischemic windows. Algorithm 2 Ischemic beat classification if (ST-segment ≤ 0.The cardiovascular abnormality detection algorithm is as follows.08mV) (or) (T inverted or T→ 0mV) then The beat is ischemic else The beat is normal end if E. Algorithm 1 Cardiovascular abnormality detection if (HR ≤ 60 BPM) then It is Bradycardia end if if (HR ≥120 BPM) then It is Tachycardia end if window technique is used which searches the sequences of ischemic beats exist for 30s or more.C.42 % respectively by randomly choosing ten records from European ST-T data base. Experimental Result The effectiveness of this technique is validated by choosing 10 random ECG records from European ST-T database [15]. The rules considered as the beat is ischemic when ST-deviation is more than 0. The rules defined and followed for locating T-wave inversion and ST-episode deviation uses first 30s duration ECG [13]. The HR is calculated from the extracted features of ECG signal by finding the inverse of RR-interval which is the time difference between two consecutive R-peaks present in every QRS complex [11]. The HR value and the corresponding cardiovascular abnormalities are also detected using above specified data base as given in Table 1.e. 4. Episode Recognition As per the recommendation of ESC (European Society of cardiology) the ischemic episode detection procedure considers minimum 30s duration of signal [14].08 % and 92. The beat classification is based on some clinical rules.75) then The window is ischemic else The window is normal end if if (No. bradycardia and tachycardia are detected by comparing the calculated HR value with the normal range of heart rate.08mV above or below the isoelectric line and the beat is ischemic when T-wave is inverted or flattened [7]. The fast and slow heart rate value determines these abnormalities. Beat Classification The ST-segment and T-wave are the two features generally used by cardiologist for ischemic beat classification [12]. III. The HR values less than 60BPM is termed as bradycardia (slow HR) and greater than 120BPM is called as tachycardia (fast HR)[3]. TP × 100 (T P + F N ) TP P P A(%)= × 100 (T P + F P ) Se(%)= Where TP = True Positive (correctly detected event) FP = False Positive (erroneously detected non event) FN = False Negative (erroneously missed event) The average sensitivity and average positive predictive accuracy for ischemic episodes detection were found to be 88. of Ischemic beats) /(All beats)] ≥ 0. The recognition of ischemic episode determines the existence of a series of consecutive windows which satisfies both 30s duration and 75% of ischemic beats [7]. The sensitivity parameter measures the ability to detect ischemic episode whereas positive predictive accuracy provides the estimation likelihood that the detected episode is a true Ischemic episode [16]. The first sliding window includes first 30s of the signal and the technique proceeds moving the window one beat at a time keeping window duration of 30s. T-wave inversion is measured considering T-wave amplitude variation (positive or negative) and ST-segment deviation is measured relative to the reference isoelectric line for. D.The database consists of the recordings of 90 double channel 2-hr ECG signal with sampling rate 250Hz.

Acharya. 2006. Advanced methods and tools for ECG data analysis. Egorouchkina. vol. Premature Ventricular Contraction. 537–552. 24. Kellermann. Henriques. A. [2] G. IEEE. vol. 1985. Tamil. 4. the proposed methodology detects 3693 number of 30s windows. Yes Is Episode condition satisfied ? No Episode not found Yes Ischemic Episode Detected Fig. “Automated methods for ischemia detection in long duration ecgs. Ng. pp. and M. 2011 IEEE/NIH. The performance of the proposed technique was validated on European ST-T data base. Artech House. Some heart rate values are more than 60BPM but less than 120BPM which comes under normal HR range. [8] M. 230– 236. 271–281. International Joint Conference. P. etc. and L.43 0 0 6 0 0 100 100 64. [7] C. 4. [9] J. Fang. 1. As an example.6 100 100 87. Faezipour.08 92. pp. Q. 5873–5877. I.” Biomedical Engineering. Fotiadis. vol. “A knowledge-based technique for automated detection of ischaemic episodes in long duration electrocardiograms.5 % Cardiovascular HR PPA abnormality 100 56 Bradycardia Bradycardia Normal HR Normal HR Bradycardia Bradycardia Normal HR Normal HR Normal HR Bradycardia Normal HR 83. References [1] R.” in Life Science Systems and Applications Workshop.” in SICE-ICASE. Harris. [3] F. “Real time heart ischemia detection in the smart home care system. 3. K. Liavas.67 67 Average 88. no.” Medical and Biological Engineering and Computing. pp. Faezipour. Experimental results show that average sensitivity and positive predictive accuracy are 88. Michalis. Nourani. [6] S. 39. Spaan. 2003. no.42 . A. Pan and W. 2006. Paredes. Azuaje. V. “Transient st-segment episode detection for ecg beat classification. 2007. and J.” Biomedical Signal Processing and Control. The detection of tachycardia (HR>120BPM) is not found for the chosen ECG records. 313–320. Clifford. 4. Tompkins. Antumes. Khalil. IEEE Transactions on. Nourani. Michalis. IEEE. 2010. Flowchart for cardiovascular abnormality and Ischemic episode detection TABLE I No. This technique can also be extended to detect other cardiovascular abnormalities like Bundle Branch Block. 5. Saeed. 3703–3706. Likas. M. vol.67 2 71. Bulusu. Morais. M. P. Bolz. McSharry et al.7 100 100 66. no. 2009. The performances of the proposed method are measured sensitivity (Se) and positive predictive accuracy (PPA) parameters. Mahmoud. L. IEEE-EMBS 2005. A. “Design of ambulatory ecg monitoring system to detect st pattern change. 2009. Tchoudovski.” Selected Areas in Communications. Banerjee. 6. Sufi. 121– 124. Suri. Pang. IEEE.Raw ECG Signal Pre-Processing Feature Extraction HR calculation Cardiac abnormality Detection Is Ischemic Beat Classification No Beat is Normal Rule satisfied ? Yes Is Ischemic window condition satisfied ? No Window is Normal IV. of Ischemic ST Episodes 8 6 3 6 3 3 3 3 5 3 3 ECG Data e0105 e0112 e0122 e0127 e0207 e0404 e0417 e0606 e0613 e0615 e0704 TP FP FN 11 5 2 7 6 5 3 2 11 3 3 0 1 1 0 0 0 0 1 0 0 0 3 0 0 1 % Se 78.e. of 30s Ischemic Windows Windows 3693 3672 6245 5053 3815 3884 5188 4415 4097 3832 5376 694 36 2077 316 14 30 31 4381 465 3811 5368 No. 2006.08 % and 92. 2011. 105–112. M. no. Papaloukas. pp. 2005. D. “Novel methods of faster cardiovascular diagnosis in wireless telecardiology.33 55 66. Rocha. J. The number of available ischemic ST episodes is 11 as per the European ST-T data base. and L. D. IEEE/NIH. [4] T.. Papaloukas. 2001. Likas. and A. Yu. M. 27. J. Braecklein. I. Fotiadis. Wolff-Parkinson-White syndrome. IEEE. M. Jeong and K. J. W. no. and S. an efficient method for the detection of cardiovascular abnormalities and ischemic episodes using ECG signal are proposed. 27th Annual International Conference of the.42% respectively. [11] L.” Cardiovasc Rev Rep. “A lead dependent ischemic episodes detection strategy using hermite functions. 2006. pp. bradycardia is detected using calculated HR value. 100–103.” in Engineering in Medicine and Biology Society. Advances in cardiac signal processing. [10] G. Springer Verlag. pp. pp. and S. 694 number of ischemic windows and 08 numbers of ST episodes for the ECG recording e0105. S. of No. F.67 94 100 100 100 100 100 100 100 76 58 59 78 62 58 81 3 66. IEEE Journal on. pp. and L. pp.” in Life Science Systems and Applications Workshop (LiSSA). LiSSA 2009. Tamil. “Wavelet-based denoising and beat detection of ecg signal. 2009. T. CONCLUSION In this paper. Tiwari. A. [5] C. “A real-time qrs detection algorithm. The cardiovascular abnormality i. Carvalho.

e215–e220. Stanley. pp.” in Computers in Cardiology 1994. pp. and physionet: Components of a new research resource for complex physiologic signals. Silipo. J. Mietus. Moody. Emdin. Moody. vol. pp.” European Heart Journal. Amaral. M. [16] P. P. G. 2000. Ranjith. and C. Marchesi. 1. L. pp. [14] “The European ST-T Database. Taddei. Marchesi. Glass. [15] A. no. vol. “Physiobank. R. A. A.physionet. E. L. and H. Joseph.” Circulation. vol. L. Distante. P. 1994. 24. 9. C. and C. Taddei. Zeelenberg. 23. and P. 2003. 1992. Hausdorff. no. . J. B.” ITBM-RBM. “Continuous monitoring and detection of st-t changes in ischemic patients. IEEE. E. Pisani. C. Baby. 13. 101. 1164–1172. 44–47. Peng. Ivanov. “Ecg analysis using wavelet transform: application to myocardial ischemia detection. Goldberger. C. G. “The european st-t database: standard for evaluating systems for the analysis of st-t changes in ambulatory electrocardiography.-K. physiotoolkit. P.[12] R. G.” http://www. M. 225–228. [13] A. no. Mark.