Assessment of ECG frequency and morphology parameters for automatic classification of life-threatening cardiac arrhythmias
|
|
- Marcia Andrews
- 6 years ago
- Views:
Transcription
1 INSTITUTE OF PHYSICS PUBLISHING Physiol. Meas. 26 (2005) PHYSIOLOGICAL MEASUREMENT doi: / /26/5/011 Assessment of ECG frequency and morphology parameters for automatic classification of life-threatening cardiac arrhythmias Vessela Krasteva and Irena Jekova Centre of Biomedical Engineering Prof. Ivan Daskalov, Bulgarian Academy of Science, Acad. G. Bonchev str. Bl.105, 1113 Sofia, Bulgaria and Received 13 January 2005, accepted for publication 26 May 2005 Published 17 June 2005 Online at stacks.iop.org/pm/26/707 Abstract The reliable recognition and adequate electrical shock therapy of lifethreatening cardiac states depend on the electrocardiogram (ECG) descriptors which are used by the defibrillator-embedded automatic arrhythmia analysis algorithms. We propose a method for real-time ECG processing and parameter set extraction using band-pass digital filtration and ECG peak detection. Twelve parameters were derived: (i) seven parameters from the band-pass filter output six threshold parameters and one peak counter; (ii) five parameters from the ECG peak detection branch, which assess the heart rate, the periodicity and the amplitude/slope symmetry of the ECG peaks. The statistical assessment for more than 36 h of cardiac arrhythmia episodes collected from the public AHA and MIT databases showed that some of the parameters achieved high specificity and sensitivity, but there was no parameter providing 100% separation between non-shockable and shockable rhythms. In order to estimate the influence of the wide variety of cardiac arrhythmias and the different artifacts in real recording conditions, we performed a more detailed study for eight non-shockable and four shockable rhythm categories. The combination of the six top-ranked parameters provided specificity: (i) more than 99% for rhythms with narrow supraventricular complexes, premature ventricular contractions, paced beats and bradycardias; (ii) almost 95% for supraventricular tachycardias; (iii) 91.5% for bundle branch blocks; (iv) 92.2% for slow ventricular tachycardias. The attained sensitivity was above 98% for coarse and fine ventricular fibrillations and 94% for the rapid ventricular tachycardias. The accuracy for the noise contaminated nonshockable and shockable signals exceeded 93%. The proposed parameter set guarantees an accuracy that meets the AHA performance goal for each rhythm category and could be a reliable facility for AED shock-advisory algorithms /05/ $ IOP Publishing Ltd Printed in the UK 707
2 708 VKrastevaandIJekova Keywords: non-shockable and shockable arrhythmias, automatic ECG analysis, band-pass digital filtration, ECG peak detection, automatic external defibrillator 1. Introduction The reliability of defibrillator-embedded algorithms for automatic arrhythmia analysis from an external electrocardiogram (ECG) is of crucial importance in the diagnosis of life-threatening cardiac states. According to the American Heart Association (AHA) recommendations for automatic external defibrillator (AED) algorithm performance (Kerber et al 1997), decision for application of high voltage defibrillation shock is obligatory for coarse ventricular fibrillation and rapid ventricular tachycardia (more manufacturers specify frequency above 150 beats min 1 ). Patients with shockable rhythms potentially receive the greatest benefit from defibrillation at essentially no risk, and high sensitivity for AED analysis is required for this group. Moreover, the shock application is not allowed for normal sinus rhythms and other non-lethal cardiac dysfunctions such as supraventricular tachycardia, sinus bradycardia, atrial fibrillation and flutter, heart block, idioventricular rhythms, premature ventricular contractions and other rhythms accompanied by a palpable pulse and/or occurring in a conscious patient. Patients with non-shockable rhythms derive no benefit from defibrillation and are at maximum risk, therefore high specificity is required for this group. In addition, AHA specifies a third group of intermediate rhythms, which include low survival rate fine ventricular fibrillation and ventricular tachycardia that does not meet the criteria for inclusion in the shockable rhythm category. AHA did not set any performance goals for AED analysis for these rhythms, since the patients with intermediate rhythms are unlikely to derive benefit or be at risk from defibrillation. The automatic arrhythmia analysis algorithms use ECG descriptors which, to some extent, overlap for the wide variety of cardiac disfunctions and disturb the algorithm accuracy. Therefore, AHA specifies different algorithm accuracy performance goals for the specific shockable and non-shockable rhythm categories. During the last few decades, a number of detection and analysis techniques for shockable and non-shockable rhythm discrimination were proposed. Some of them aimed at recognizing normal sinus rhythms, ventricular tachycardia and ventricular fibrillation, applying parameters from the time domain (Thakor et al 1990, Chen et al 1987), time-frequency domain (Millet- Roig et al 1999) and complexity measure (Zhang et al 1999). Other investigators distinguish only shockable from non-shockable rhythms, using frequency domain parameters (Kuo and Dillman 1978, Barro et al 1989), neural networks (Clayton et al 1994, Minami et al 1999) and nonlinear prediction (Kaplan and Cohen 1990, Denton 1992, Jekova et al 2002). The wavelet transformations were advanced as an applicable technique for rhythm classification in four categories normal rhythms, atrial fibrillation, ventricular tachycardia and ventricular fibrillation (Khadra et al 1997, Al-Fahoum and Howitt 1999). Chen (2000) proposed discrimination between supraventricular tachycardia, ventricular tachycardia and ventricular fibrillation using a total least-squares-based Prony model. That study, however, did not consider normal sinus rhythms, bradycardias or premature ventricular contractions and did not distinguish shockable from non-shockable ventricular tachycardias. The present study describes an algorithm for ECG frequency and morphology analysis suitable for embedding in real-time operating AED microprocessor systems. The statistical assessment of the derived parameters proved their reliability for making an adequate shock advisory decision for different shockable and non-shockable rhythm categories. A set of
3 Assessment of ECG frequency and morphology parameters 709 the most valuable parameters for each rhythm category was extracted and involved into discriminant analysis, in order to obtain better accuracy, which has to conform to the AHA recommendations for reliable AED algorithm performance. 2. Methods 2.1. ECG signals The study was performed with the following full-length standard ECG databases, which contain a wide variety of non-shockable and shockable rhythms: AHA fibrillation database 2 channels, 30 min (10 2 files): A8001 to A8010 (d1+d2); VFDB. MIT-BIH malignant ventricular arrhythmia database 2 channels, 35 min (22 2 files): 418 to 430 (d1+d2); 602, 605, 607, 609, 610, 611, 612, 614, 615 (d1+d2); CUDB. Creighton University ventricular tachyarrhythmia database 1 channel, 8 min (35 files): Cu01 to Cu35. All signals are sampled at 250 Hz and 12 bit resolution. These databases were annotated for each 10 s time interval by an experienced cardiologist and a biomedical engineer. The following general annotation groups were accepted: nonshockable rhythm, shockable rhythm, asystole, noise, non-defined rhythm. Signals which did not exceed 150 µv from peak to peak were labeled as asystole and signals which contained extreme artifacts were labeled as noise. The non-defined group consisted of transitions from non-shockable to shockable rhythms and signals with low signal-to-noise ratio. The present study included only signals from the non-shockable rhythm and shockable rhythm general annotation groups, which were additionally classified in 12 subgroups, taking into account the AHA recommendations (Kerber et al 1997). Fine ventricular fibrillations and slow ventricular tachycardias, which belong to the AHA intermediate rhythms, were included as subgroups in the shockable rhythm and non-shockable rhythm annotation groups, respectively. In order to provide a more informative study, we also added subgroups in the shockable rhythms and non-shockable rhythms which contain ECG signals corrupted by non-extreme artifacts. A detailed description of each subgroup is presented in table 1. Examples of all rhythm categories are shown in figure 2 ( non-shockable rhythms) and in figure 3 ( shockable rhythms ) Method The proposed method for ECG analysis was applied on subsequent 10 s ECG episodes from the above described dataset. The concept of the method includes combined assessment of the frequency components and the morphology of the ECG signal. Therefore, a parameter set extraction from two independent branches was implemented. The first branch, named ECG frequency-band estimation, analyzed the ECG frequency characteristics in a defined frequency band. The second branch, called ECG morphology estimation, aimed to detect the significant peaks in the signal and to assess their morphological characteristics. Figure 1 presents the general block diagram of the method, which is described in the sections below.
4 710 VKrastevaandIJekova Table 1. Summarized criteria for annotation of the non-shockable and shockable rhythm categories. The sample size represents the total number of 10 s time intervals from AHA and MIT databases, which were annotated in the respective categories. Shock Sample recommendation Group Label size Rhythm category Criteria Non-shockable G1:1 NSVC 4169 Narrow supraventricular QRS < 120 ms, complexes (normal sinus supraventricular origin, rate rhythm, premature atrial between 45 and 120 bpm contractions, atrial flutter and fibrillation) Non-shockable G1:2 PVC 2522 Premature ventricular Existence of PVC contractions (QRS > 120 ms) in the 10 s ECG segment Non-shockable G1:3 BL 503 Complete bundle branch QRS > 120 ms block rhythm Non-shockable G1:4 PACE 280 Paced rhythm Visible paced artifacts Non-shockable G1:5 BR 601 Bradycardia and slow Rate < 45 bpm Idioventricular rhythms Non-shockable G1:6 SVT 356 Supraventricular Supraventricular origin, tachycardia QRS < 120 ms, rate > 120 bpm Non-shockable G1:7 VT < Ventricular tachycardia Ventricular origin, X below 150 bpm QRS > 120 ms, rate < 150 bpm Non-shockable G1:8 NShR Non-shockable rhythm with Presence of not extreme Noise noise artifacts the signal-to-noise ratio is more than 1 or the artifact duration is less than 1 s Shockable G2:1 VT > Ventricular tachycardia Ventricular origin, above 150 bpm QRS > 120 ms, rate > 150 bpm Shockable G2:2 CVF 1572 Coarse ventricular Amplitude > 0.25 mv fibrillation Shockable G2:3 FVF 689 Fine ventricular fibrillation Amplitude < 0.25 mv Shockable G2:4 ShR+ 200 Shockable rhythm with Presence of not extreme noise noise artifacts the same as G1:8 All signal processing procedures were performed using the software package MATLAB 6.0 (MathWorks, Inc.) Signal preprocessing. The signal preprocessing procedure was applied in order to simulate the input hardware filters, which are usually implemented in AEDs and ECG monitors. The signal processing filtration included: (i) two successive first-order high-pass filters with 1 Hz cut-off frequency to suppress residual baseline drift; (ii) a second-order 30 Hz low-pass filter to reduce muscle noise and (iii) a notch filter to eliminate power-line interference. This filtration complied with the defibrillator monitors recommendations of IEC Committee Draft (2001) ECG frequency-band estimation branch Band-pass digital filter. The band-pass filtering around 17 Hz proved to be a useful technique for effective selection of the QRS complexes in the ECG (Thakor et al 1984).
5 Assessment of ECG frequency and morphology parameters 711 ECG DATASET (10 S TIME EPOCHS) SIGNAL PREPROCESSING BAND-PASS DIGITAL FILTER PEAK DETECTION ALGORITHM EXTRACTION OF PARAMETERS FROM THE BAND-PASS FILTER OUTPUT EXTRACTION OF PARAMETERS AFTER ECG PEAK DETECTION ECG frequency-band estimation branch. Section ECG morphology estimation branch. Section STATISTICAL ANALYSIS Figure 1. General block diagram of the method. (a) - G1:1 (NSVC) A8003d1 00:10 (e) - G1:5 (BR) A8010d2 00:10 (f) - G1:6 (SVT) A8003d1 11:00 (b) - G1:2 (PVC) Cu05 03:20 (c) - G1:3 (BL) Cu03 00:00 (g) - G1:7 (VT<150) 611d1 22:10 (d) - G1:4 (PACE) Cu32 00:50 (h) - G1:8 (NShR+Noise) Cu19 00:10 Figure 2. Examples of 10 s ECG segments (first trace) and the respective absolute value of the band-pass filter output (second trace) for all defined non-shockable rhythm categories. The detected significant peaks in both traces are marked with asterisks. Applying a similar approach, we process 10 s ECG episodes with a band-pass digital filter with integer coefficients, which was described in detail in our previous study (Jekova and Krasteva 2004). Taking into account the fact that the VF signals have frequency components below
6 712 VKrastevaandIJekova (a) - G2:1 (VT>150) Cu10 06:00 (c)- G2:3 (VFF) A8007d1 29:50 (b) - G2:2 (VFC) A8003d1 06:00 (d)- G2:4 (ShR+Noise) Cu30 02:50 Figure 3. Examples of 10 s ECG segments (first trace) and the respective absolute value of the band-pass filter output (second trace) for all defined shockable rhythm categories. The detected significant peaks in both traces are marked with asterisks. 7 Hz, according to Murray et al (1985) and Clayton et al (1994), and less than 10 Hz reported by Minami et al (1999), the proposed band-pass filter was designed to provide a low-amplitude output signal for this rhythm. On the other hand, the filter was calculated to provide relatively high-amplitude peaks for the supraventricular and ventricular complexes with frequencies up to 20 Hz and 14 Hz respectively (Minami et al 1999). Therefore, considering that the frequency range between 13 and 17 Hz contains the frequency components of the non-shockable rhythm complexes but almost none of the frequency components of the ventricular fibrillation, we selected a central frequency at 15 Hz with ±2 Hz bandwidth and designed a recursive filter with floating point precision coefficients. Aiming at a simpler solution, convenient for embedding in an AED microprocessor system, we preferred to use a digital filter with integer coefficients. A recursive filter with central frequency at 14.6 Hz and bandwidth from 13 Hz to 16.5 Hz ( 3 db) was obtained by reducing the floating point coefficients to integer coefficients. The filter equation (1), valid for 250 Hz sampling frequency, was designed in consideration of its future real-time implementation: FS i = 14FS i 1 7FS i 2 + S i S i 2 2. (1) 8 Here S i is a signal sample with index i; FS i is the filtered signal sample with index i Extraction of parameters from the band-pass filter output. The absolute value of the band-pass filter output (AbsBPF) for 10 s ECG series was considered for analysis. Examples of the AbsBPF signal for all rhythm categories which were included in the present study are illustrated in the second traces of figure 2 ( non-shockable rhythms) and figure 3 ( shockable rhythms). Seven parameters were calculated from the AbsBPF, named C1,C2,C3,C4,C5,C6 and BP PEAKS. The first six parameters represent the number of signal samples with amplitude values within a certain amplitude range. The respective ranges of these threshold parameters were defined as follows: C1 Range: 0.5 max(absbpf) to max(absbpf); C2 Range: mean(absbpf) to max(absbpf); C3 Range: mean(absbpf)-md to mean(absbpf)+md, C4 Range: 2 mean(absbpf) to max(absbpf); C5 Range: 0.75 max(absbpf) to max(absbpf); C6 Range: 0.25 max(absbpf) to max(absbpf),
7 Assessment of ECG frequency and morphology parameters 713 where max(absbpf), mean(absbpf) and MD (mean deviation) were computed in segments of 1 s. The choice of short time interval for recalculation of the ranges prevents the continuous influence of peak artifacts, as well as improves the fast adaptation to rhythm changes. In our previous study (Jekova and Krasteva 2004), the parameters C1, C2 and C3 were used and their dependence on the rhythm type (shockable or non-shockable) was explained in detail. In order to obtain supplementary assessment of the band-pass filter output, we introduced C5 and C6 by analogy with C1, and C4 analogous to C2. The parameter BP PEAKS is the number of peaks in the AbsBPF signal for 10 s epoch. The detected peaks in the band-pass filter output signals are marked with asterisks inthe examples of figures 2(a) (h) and figures 3(a) (d) (second traces). It is obvious that the value of this parameter is lower for rhythms which have well-defined peaks in the band-pass filter output and is higher for ECG signals without expressed peaks after band-pass filtration. The algorithm for detection and validation of significant peaks in the band-pass filter channel follows: Step 1. Initial detection of all peaks the samples with maximal amplitude for each 200 ms time interval Peak Amplitude(i) i=1,2,...,50 ; Step 2. Time criterion for validation of peaks a peak is validated if the time interval to the next peak is greater than 100 ms. Otherwise, the peak with higher amplitude is validated. Step 3. Amplitude criterion for validation of peaks. An amplitude threshold is calculated: AmplThr = mean(peakamplitude(i) i=1,2,...,50 ); A peak (with index i) is validated if its amplitude is higher than the amplitude threshold: PeakAmplitude(i) AmplThr. The amplitude threshold is updated only when the amplitude of the validated peak satisfies the restriction (PeakAmplitude(i) < 7 AmplThr), in order to avoid the influence of extreme artifacts: AmplThr = (AmplThr +0.5 PeakAmplitude(i))/ ECG morphology estimation branch Peak detection. We applied an in-house developed algorithm for detection of significant positive and negative peaks in the preprocessed ECG signal. The detected significant peaks in the ECG signals are marked with asterisks in the examples of figures 2(a) (h) and figures 3(a) (d) (first traces). A brief description of the method follows. Step 1. Initial detection of all positive and all negative peaks in the 10 s ECG series. One can consider that ECG sample with index i (ECG(i)) is a peak if the following conditions are satisfied: abs(ecg(i)) 10 µv. ECG(i) is an extremum in 20 ms interval around the sample. The sign of the extremum is equal to the sign of ECG(i). Step 2. Selection of peaks, according to the following criteria: For successive peaks with equal signs, the highest amplitude peak (absolute value) is selected.
8 714 VKrastevaandIJekova IndexPeak=IndexPeak+1 AmplPeak, SlopePeak MeanThrAmplPos=0.75*AmplPosPeaks MeanThrAmplNeg=0.75*AmplNegPeaks MeanThrSlopePos=0.75*SlopePosPeaks MeanThrSlopeNeg=0.75*SlopeNegPeaks PosPeaksNumber=0; NegPeaksNumber=0 IndexPeak=1 Yes AmplPeak(IndexPeak)>0 No ThrAmpl=0.75*MeanThrAmplPos ThrSlope=0.625*MeanThrSlopePos PosPeaksNumber=PosPeaksNumber+1 ThrAmpl=0.75*MeanThrAmplNeg ThrSlope=0.625*MeanThrSlopeNeg NegPeaksNumber=NegPeaksNumber+1 (abs(amplpeak(indexpeak))>thrampl) and (abs(slopepeak(indexpeak))>thrslope) Yes Peak(IndexPeak) VALIDATION No AmplPeak(IndexPeak)>0 Yes MeanThrAmplPos= ((PosPeaksNumber-1)*MeanThrAmplPos+ThrAmpl)/PosPeaksNumber MeanThrSlopePos= ((PosPeaksNumber-1)*MeanThrSlopePos+ThrSlope)/PosPeaksNumber MeanThrAmplNeg= ((NegPeaksNumber-1)*MeanThrAmplNeg+ThrAmpl)/NegPeaksNumber MeanThrSlopePos= ((NegPeaksNumber-1)*MeanThrSlopeNeg+ThrSlope)/NegPeaksNumber No (abs(amplpeak(indexpeak))<2*thrampl) and (abs(slopepeak(indexpeak))<2*thrslope) Yes K=0.625 K=0.5 No ThrAmpl=K*AmplPeak(IndexPeak) ThrSlope=K*SlopePeak(IndexPeak) Figure 4. Flow chart of the procedure for validation of peaks according to adaptive amplitude and slope criteria. For successive peaks with different signs, the current peak is selected if the peak to peak amplitude exceeds a preset threshold value (50% of the average amplitude of all peaks which were detected in Step1). If the time interval between two peaks with equal signs is less than 100 ms, then the higher amplitude peak is selected. Step 3. Validation of peaks. The procedure for validation of peaks according to adaptive amplitude and slope criteria is presented in the flow chart of figure 4. Taking into account the peaks selected in Step2, the input parameters for Step3 are calculated as follows: IndexPeak is the index of the peak, which corresponds to the ith ECG sample; Peak amplitude AmplPeak(IndexPeak) = ECG(i); Peak slope SlopePeak(IndexPeak) = 5 j= 5 abs(ecg(i) ECG(i j)); AmplPosPeaks is the average value of the amplitudes of all positive peaks; AmplNegPeaks is the average value of the amplitudes of all negative peaks; SlopePosPeaks is the average value of the slopes of all positive peaks; SlopeNegPeaks is the average value of the slopes of all negative peaks Extraction of parameters from the ECG peak detection branch. Using the ECG peak detection method described above, we calculated the amplitude and the slope for each validated peak in the ECG signal, as well as the period between every two consecutive peaks
9 Assessment of ECG frequency and morphology parameters 715 with equal signs. The mean values of the amplitudes, slopes and periods of the positive and the negative peaks, and the respective standard deviations, were used for calculation of several summary parameters for the entire 10 s epoch. The extracted parameters can be interpreted as follows: ECG PEAKS. The mean value of the number of positive and the number of negative peaks detected in 10 s ECG series. In the case of correctly detected peaks, the value of this parameter is expected to correspond to the heart rate for non-shockable rhythms and tachycardias or to the number of fibrillation waves; AMPL PEAKS. The number of positive peaks with amplitude in the range (absolute mean amplitude of the negative peaks ± standard deviation) summed with the number of negative peaks with absolute amplitude value in the range (mean amplitude of the positive peaks ± standard deviation) and normalized to the total number of detected peaks (in per cent). This parameter is related to the amplitude symmetry of the positive and the negative peaks; SLOPE PEAKS. The same meaning as AMPL PEAKS but for the slope of the peaks; AMPL SLOPE PEAKS. The number of positive and negative peaks which satisfy together the criteria set in AMPL PEAKS and SLOPE PEAKS. The value is normalized to the total number of detected peaks (in per cent). This parameter is related to both the amplitude and the slope symmetry of the positive and the negative peaks; PERIOD PEAKS. The number of positive peaks with period in the range 75% to 125% of the mean period of the positive peaks summed with the number of negative peaks with period in the range 75% to 125% of the mean period of the negative peaks. The value is normalized to the total number of detected peaks (in per cent). This parameter estimates the periodicity of the detected peak series. 3. Results The extracted parameters with band-pass digital filtration and ECG peak detection method were calculated for all 10 s time intervals from the full-length AHA and MIT databases, which were annotated by an experienced cardiologist. The results were analyzed with the software package Statistica (StatSoft, Inc.). The ability of each parameter to differentiate between shockable and non-shockable rhythms was estimated by means of standard discriminant analysis. Initially, the parameters were assessed for the two general discriminant analysis classes: Class1 (G1) all non-shockable rhythms and Class2 (G2) all shockable rhythms. The specificity (Sp) and sensitivity (Se) provided by each parameter were calculated as follows: Sp = True Class1/Class1, where True Class1 is the number of correctly recognized non-shockable rhythms from the group of Class1. Se = True Class2/Class2, where True Class2 is the number of correctly recognized shockable rhythms from the group of Class2. The Sp values for all non-shockable rhythms (NShR) and the Se values for all shockable rhythms (ShR), which were calculated for all parameters, are shown in table 2 (column all rhythms ). Second, the reliability of each parameter to classify each rhythm category in the respective shock recommendation group (non-shockable or shockable) was assessed. For each rhythm category (described in table 1), the discriminant analysis classes were formed as follows: Non-shockable rhythm category. Class1 (G1:m m = 1,2,...,8 ) all cases which belong to the respective non-shockable category and Class2 (G2) all shockable rhythms.
10 716 VKrastevaandIJekova Table 2. Discrimination ability (represented by specificity (Sp) and sensitivity (Se)) of each parameter for all non-shockable and all shockable rhythms, and their categories. For each rhythm category, the combinations of high Sp and Se values are emphasized. Shockable rhythm All rhythms Non-shockable rhythm categories categories NShR ShR NSVC PVC BL PACE BR SVT VT NShR+ VFC VFF VT ShR+ <150 Noise >150 Noise G1 G2 G1:1 G1:2 G1:3 G1:4 G1:5 G1:6 G1:7 G1:8 G2:1 G2:2 G2:3 G2:4 Sp Se Sp Sp Sp Sp Sp Sp Sp Sp Sp Sp Sp Sp Se Se Se Se Se Se Se Se Se Se Se Se Sample size Parameters C C C C C C BP PEAKS ECG PEAKS AMPL PEAKS SLOPE PEAKS AMPL SLOPE PEAKS PERIOD PEAKS Shockable rhythm category. Class1 (G1) all non-shockable rhythms and Class2 (G2:n n = 1,2,...,4 ) all cases which belong to the respective shockable category. Considering the defined discriminant classes, the Sp and Se values for each rhythm category were calculated using the above introduced equations. The results are shown in table 2 (columns non-shockable rhythm categories and shockable rhythm categories ). For each parameter, the first row of each column represents the Sp value and the second row is the Se value.
11 Assessment of ECG frequency and morphology parameters NSVC PVC BL PACE BR SVT VT<150 C2 C3 C NShR+Noise VT>150 VFC VFF ShR+Noise NSVC PVC BL PACE BR SVT VT<150 NShR+Noise VT>150 VFC VFF ShR+Noise NSVC PVC BL PACE BR SVT VT<150 (a) (b) (c) NShR+Noise VT>150 VFC VFF ShR+Noise BP_PEAKS ECG_PEAKS SLOPE_PEAKS NSVC PVC BL PACE BR SVT VT<150 NShR+Noise VT>150 VFC VFF ShR+Noise NSVC PVC BL PACE BR SVT VT<150 NShR+Noise VT>150 VFC VFF ShR+Noise NSVC PVC BL PACE BR SVT VT<150 (d) (e) (f) NShR+Noise VT>150 VFC VFF ShR+Noise Mean ±0.95 Conf. Interval ±SD Figure 5. Distributions of the mean values, ±95% confidence intervals and ±standard deviations for the parameters: C2, C3, C6, BP PEAKS, ECG PEAKS and SLOPE PEAKS, evaluated for each rhythm category. The shockable rhythm categories are enclosed in square regions. Figure 5 illustrates the distributions of the mean values, ±95% confidence intervals and ±standard deviations for the parameters C2, C3, C6, BP PEAKS, ECG PEAKS and SLOPE PEAKS, which provide the best combinations of high Sp and high Se values evaluated for each rhythm category (see table 2). These six top ranked parameters were processed together with standard discriminant analysis. The classification ability of the selected parameter set (SPS) is summarized in table 3. The achieved values for specificity and sensitivity were compared with the AHA performance goal recommendations (only for those available in AHA rhythm categories). Further, we obtained the categorized scatterplots for non-shockable rhythm and shockable rhythm general annotation groups of the parameters: (i) C6 and BP PEAKS figure 6(a); (ii) ECG PEAKS and SLOPE PEAKS figure 6(b). To achieve more detailed investigation of these relationships within different rhythm categories, we propose the respective categorized scatterplots in figures 6(c) and (d). 4. Discussion We proposed a method for real-time analysis of the band-pass digital filter output and a simple algorithm for ECG peak detection. Thus we extracted parameters which are representative for the peak characteristics, summarized for a 10 s time epoch. The statistical assessment of these parameters for more than 36 h of cardiac arrhythmia episodes collected from the standard AHA and MIT databases showed that some of the parameters achieve high specificity and
12 718 VKrastevaandIJekova Table 3. Discrimination ability (represented by specificity (Sp) and sensitivity (Se)) of the selected parameter set (SPS C2, C3, C6, BP PEAKS, ECG PEAKS, SLOPE PEAKS) for all nonshockable and all shockable rhythms, and their categories. The respective AHA performance goal and the 90% one-sided low confidence limit (LCL) are also presented for each rhythm category. (NA) data not available. Shockable rhythm All rhythms Non-shockable rhythm categories categories NShR ShR NSVC PVC BL PACE BR SVT VT NShR+ VFC VFF VT ShR+ <150 Noise >150 Noise G1 G2 G1:1 G1:2 G1:3 G1:4 G1:5 G1:6 G1:7 G1:8 G2:1 G2:2 G2:3 G2:4 Sample size SPS Sp Se NA NA Sp> Sp> Sp> Sp> Sp> Sp> NA NA Se> NA Se> NA AHA Goal 99% 95% 95% 95% 95% 95% NA NA 90% 75% LCL NA NA 97% 88% 88% 88% 88% 88% 87% NA 67% NA sensitivity, but there is no parameter providing 100% separation between the non-shockable rhythm and shockable rhythm general annotation groups (see table 2, column all rhythms ). Therefore, considering only the two general groups of shockable and non-shockable rhythms, the value range of each parameter overlaps (see the examples in figures 6(a) and (b)), but there is no information about both the specific reasons for this overlapping and the additional criteria which have to be complied to avoid it. There are many factors which introduce uncertainty for example the wide variety of cardiac arrhythmias, the non-deterministic morphology of fibrillation waves, the influence of different artifacts in real recording conditions and many other individual aspects. We tried to specify to some extent the influence of these factors by supplementary assessment of each parameter for different categories of shockable and non-shockable rhythms. Taking into account the particularities of the extracted parameter set, we selected the QRS duration, the heart rate, the amplitude of the fibrillation waves and the presence of artifacts to be the main criteria for classification of the cardiac arrhythmias in categories which were described in section 2.1, table 1. The detailed assessment with standard discriminant analysis showed that the correct classification of the different rhythm categories in the respective shock recommendation group is assured by different parameters. The combinations of high specificity and high sensitivity values for each rhythm category are emphasized in table 2. The band-pass filter threshold parameters (C2,C3,C6) provide reliable performance for both the narrow complex heart rhythms (Sp > 98.8% and Se > 96.2%) and the ventricular fibrillations (Se > 99% and Sp > 92.4), but they do not cope with the wide complex rhythms with complete bundle branch block and ventricular tachycardias (Sp and Se do not exceed 90%). This disadvantage is due to the fact that the designed band-pass filter with bandwidth from 13 Hz to 16.5 Hz suppresses to a great extent both the relatively lower frequency QRS components which are predominant in the spectrum of rhythms with bundle branch block and tachycardias, and the fibrillation waves (see figures 2(c), (g) and 3). Therefore, the band-pass filter threshold parameters are not reliable for discrimination of these arrhythmias and classify part of them as shockable rhythms. The parameter BP PEAKS was found to be more useful since it provides the highest accuracy for all rhythm categories, except for the slow tachycardia. This can be explained by the existence of well-expressed peaks in the band-pass filter output for all non-shockable categories except the slow tachycardias and the absence of such peaks for the shockable rhythms (see figures 2 and 3). Thus, BP PEAKS has higher values for both
13 Assessment of ECG frequency and morphology parameters 719 (a) (b) Figure 6. Categorized scatterplots by groups: (a) the distributions of the parameters C6 and BP PEAKS for non-shockable rhythm and shockable rhythm general annotation groups; (b) the distributions of the parameters ECG PEAKS and SLOPE PEAKS for non-shockable rhythm and shockable rhythm general annotation groups; (c) the distributions of the parameters C6 andbp PEAKS for all defined non-shockable and shockable rhythm categories; (d) the distributions of the parameters ECG PEAKS and SLOPE PEAKS for all defined non-shockable and shockable rhythm categories. the shockable rhythms and the slow tachycardias, and corresponds to the heart rate for the remaining non-shockable rhythms (see figure 5(d)). The parameters which were extracted with the ECG peak detection branch presented relatively worse classification accuracy. However, the parameters ECG PEAKS, SLOPE PEAKS and AMPL SLOPE PEAKS proved to be the most informative for the correct classification of slow tachycardias, since ECG PEAKS represents the heart rate (see figure 5(e)), and SLOPE PEAKS and AMPL SLOPE PEAKS are related to the relatively
14 720 VKrastevaandIJekova (c) (d) Figure 6. (Continued.) small symmetry of the slope or of both the amplitude and slope of the positive and the negative tachycardia peaks (see figure 5(f )). The six top-ranked parameters (C2, C3, C6, BP PEAKS, ECG PEAKS, SLOPE PEAKS), which provided the best classification accuracy for the different rhythm categories were combined in a set in order to obtain higher specificity and sensitivity for non-shockable rhythm and shockable rhythm general annotation groups (see table 3, column all rhythms ). By processing the selected parameter set with standard discriminant analysis, we achieved specificity of 95.6% and sensitivity of 98%. A more detailed study proved the reliability of the proposed parameter set for the different rhythm categories as follows: (i) Specificity more than 99% for NSVC, PVC, PACE and BR. The categorized scatterplot by rhythm category of parameters C6 and BP PEAKS (figure 6(c)) proves that the parameters
15 Assessment of ECG frequency and morphology parameters 721 distributions for these rhythm categories do not overlap with the observed distributions for all shockable rhythms. (ii) Specificity of almost 95% for SVT. The accuracy reduction is due to the fact that the distributions of C6 and BP PEAKS (the top-ranked SVT rhythm parameters) for SVTs overlaps to a slight extent with the respective distributions for VFC and ShR+Noise categories from the shockable rhythm group (figure 6(c)). (iii) Specificity of 91.5% for BL. This is the only non-shockable rhythm category for which the specified AHA performance goal is not satisfied, but the range of 90% one-sided low confidence limit is exceeded. The categorized scatterplot of C6 and BP PEAKS shows slight overlapping for BL rhythm and some of the shockable rhythm categories VFC, rapid VT and ShR+Noise (see figure 6(c)). We note a similar overlapping in the scatterplot of ECG PEAKS and SLOPE PEAKS for BL rhythm and the two shockable VFC and ShR+Noise categories (see figure 6(d)). (iv) Specificity of 92.2% for the slow VTs (VT < 150). This relatively high specificity value for the traditionally difficult to classify, slow VTs is due to the efficient combination between ECG PEAKS and SLOPE PEAKS, which can be observed from the well separable distributions of these parameters for VT < 150 and all shockable rhythms (see the categorized scatterplot in figure 6(d)). (v) Sensitivity above 98% for ventricular fibrillation and 94% for the rapid VT (VT > 150). Although there is an overlapping in the scatterplots (figures 6(c) and (d)) for all shockable categories with some of the above discussed non-shockable rhythms, we assure an accuracy which exceeds the AHA performance recommendations. (vi) Accuracy for noise contaminated non-shockable and shockable categories above 93%. The existence of artifacts impedes the correct detection of significant ECG peaks and the proper interpretation of the band-pass filter output. Thus, the extracted parameters which are representative for the peak characteristics have relatively overlapping distributions (see the NShR+Noise and ShR+Noise categorized scatterplots in figures 6(c) and (d)). However, the combination of six parameters and the application of standard discriminant analysis assure the relatively high specificity and sensitivity for these rhythm categories. The strict comparison between the above presented statistical indices and the algorithm performances reported by other investigators could not be done, since different rhythm categories had been defined and different databases were used for testing. However, aiming to provide a comparative study, we present in table 4 the detection accuracies for published algorithms which evaluate different arrhythmia types (all rhythms are cited exactly as defined by the authors). It should be pointed out the special conditions for recording of ECG data containing malignant arrhythmia. Some of the authors use the publicly available databases from small surface electrodes, e.g. AHA and MIT, thus obtaining comparable and reproducible results. Others collect their own databases from ECG recordings during electrophysiological procedures, holter or monitoring systems in coronary care units. These datasets are usually from surface leads or patch and catheter electrodes, but they are not accessible and the examinations are not reproducible. Moreover, there is no publicly available malignant reference database from patch chest electrodes suitable for testing of AED shock-advisory algorithms. While reporting our results, we should note that the small surface electrodes signals are more influenced by noise and drift artifacts compared to large adhesive electrodes. We expect increased accuracy with signals from AED electrodes, especially for NShR+Noise and ShR+Noise categories. Although the results shown in table 4 concern different rhythm categories, it is obvious that the performance of our method is comparable or even better than published techniques.
16 722 VKrastevaandIJekova Table 4. Reported algorithm performances for particular rhythm categories defined by the authors: NSR normal sinus rhythm, NR normal rhythm, AF atrial fibrillation, SVR supraventricular rhythm, SVT supraventricular tachycardia, VT ventricular tachycardia, VR ventricular rhythm, VF ventricular fibrillation. (NA) data not available. Rhythm Sp type (%) Se (%) Accuracy (%) ECG data recordings Thakor et al 1994 NSR NA NA 95.3 Endocardial leads SVT NA NA 100 VT NA NA 92.9 Khadra et al 1997 NRs NA MIT DB AF NA Holter ECG recordings VT NA (YUDB Lead II) VF NA Al-Fahoum and Howitt 1999 NR NA MIT DB AF NA Holter ECG recordings VT NA (YUDB Lead II, VF NA MMSDB) Minami et al 1999 SVR NA Own database Lead II VR NA VF NA Chen 2000 SVT NA NA 95.2 MIT-BIH database VT NA NA 97.8 VF NA NA 96 Moreover, the direct comparison of our results with the AHA recommendations for each rhythm category proved that the proposed parameter set provides accuracy which meets the AHA performance goal and could be a reliable facility for AED shock-advisory algorithms. The additional sub-classification of the non-shockable and shockable rhythms into 12 categories clarified the reasons for the observed overlapping for the non-shockable rhythm and the shockable rhythm groups, and is useful for further rhythm analysis and parameter selection. 5. Conclusion We proposed a method for real-time ECG processing using band-pass digital filtration and ECG peak detection. Twelve ECG frequency and morphology parameters for automatic classification of life-threatening cardiac arrhythmias were calculated. The statistical assessment of more than 36 h of cardiac arrhythmia episodes collected from the public AHA and MIT databases presented the ability of each parameter to classify the ECG signals in different shockable and non-shockable rhythm categories. The combination of the six top-ranked parameters provided specificity: (i) more than 99% for rhythms with narrow supraventricular complexes, premature ventricular contractions, paced beats and bradycardias; (ii) almost 95% for supraventricular tachycardias; (iii) 91.5% for bundle branch blocks; (iv) 92.2% for slow ventricular tachycardias. The attained sensitivity was above 98% for coarse and fine ventricular fibrillations and 94% for the rapid ventricular tachycardias. The accuracy for the noise contaminated non-shockable and shockable signals exceeded 93%. The proposed parameter set provides an accuracy which is comparable or even better than published techniques, and complies with the AHA recommendations for reliable AED shock-advisory algorithm performance.
17 Assessment of ECG frequency and morphology parameters 723 References Al-Fahoum A S and Howitt I 1999 Combined wavelet transformation and radial basis neural networks for classifying life-threatening cardiac arrhythmias Med. Biol. Eng. Comput American Heart Association (AHA) Ventricular Arrhythmia ECG Database. Emergency care Research Institute 5200 Butler Pike, Plymouth Meeting, PA 19462, USA, Barro S, Ruiz R, Cabello D and Mira J 1989 Algorithmic sequential decision-making in a frequency domain for life threatening ventricular arrhythmias and imitative artifacts: a diagnostic system J. Biomed. Eng Chen S 2000 A two-stage discrimination of cardiac arrhythmias using a total least squares-based prony modeling algorithm IEEE Trans. Biomed. Eng Chen S, Thakor N V and Mower M M 1987 Ventricular fibrillation detection by a regression test on the autocorrelation function Med. Biol. Eng. Comput Clayton R H and Murray A 1999 Linear and non-linear analysis of the surface ECG during human ventricular fibrillation shows evidence of order in the underlying mechanism Med. Biol. Eng. Comput Clayton R H, Murray A and Campbell R W F 1994 Changes in the surface electrocardiogram during the onset of spontaneous ventricular fibrillation in man Eur. Heart J Denton T A, Diamond G A, Khan S S and Karagueuzan H 1992 Can the techniques of nonlinear dynamics detect chaotic behavior in electrocardiographic signals J. Electrocardiol IEC Committee Draft 2001 Amendment to IEC , 2nd edn, 62D/382/DCV, 39 Jekova I, Dushanova J and Popivanov D 2002 Method for ventricular fibrillation detection in the external electrocardiogram using nonlinear prediction Physiol. Meas Jekova I and Krasteva V 2004 Real time detection of ventricular fibrillation and tachycardia Physiol. Meas Kaplan D and Cohen R 1990 Is fibrillation chaos? Circ. Res Kerber R E et al 1997 Automatic external defibrillators for public access defibrillation: recommendations for specifying and reporting arrhythmia analysis algorithm performance, incorporating new waveforms, and enhancing safety Circulation Khadra L, Al-Fahoum A S and Al-Nashash H 1997 Detection of life-threatening cardiac arrhythmias using the wavelet transformation Med. Biol. Eng. Comput Kuo S and Dillman R 1978 Computer detection of ventricular fibrillation Proc. Computers in Cardiology (Long Beach, CA: IEEE Computer Society) pp Millet-Roig J, Rieta-Ibanez J J, Vilanova E, Mocholi A and Chorro F J 1999 Time-frequency analysis of a single ECG to discriminate between ventricular tachycardia and ventricular fibrillation IEEE Comput. Cardiol Minami K, Nakajima H and Toyoshima T 1999 Real-time discrimination of ventricular tachyarrhythmia with Fouriertransform neural network IEEE Trans. Biomed. Eng MIT database URL: or/vfib Murray A, Campbel R W F and Julian D G 1985 Characteristics of the ventricular fibrillation waveform Proc. Computers in Cardiology (Washington DC: IEEE Computer Society) pp Thakor N V, Natarajan A and Tomaselli G F 1994 Multiway sequential hypothesis testing for tachyarrhythmia discrimination IEEE Trans. Biomed. Eng Thakor N V, Webster J G and Tompkins W J 1984 Estimation of QRS complex power spectra for design of a QRS filter IEEE Trans. Biomed. Eng Thakor N V, Zhu Y S and Pan K Y 1990 Ventricular tachycardia and fibrillation detection by a sequential hypothesis testing algorithm IEEE Trans. Biomed. Eng Zhang X S, Zhu Y S, Thakor N V and Wang Z Z 1999 Detecting ventricular tachycardia and fibrillation by complexity measure IEEE Trans. Biomed. Eng
SPECTRAL ANALYSIS OF LIFE-THREATENING CARDIAC ARRHYTHMIAS
SPECTRAL ANALYSIS OF LIFE-THREATENING CARDIAC ARRHYTHMIAS Vessela Tzvetanova Krasteva, Irena Ilieva Jekova Centre of Biomedical Engineering Prof. Ivan Daskalov - Bulgarian Academy of Sciences Acad.G.Bonchev
More informationDIFFERENCE-BASED PARAMETER SET FOR LOCAL HEARTBEAT CLASSIFICATION: RANKING OF THE PARAMETERS
DIFFERENCE-BASED PARAMETER SET FOR LOCAL HEARTBEAT CLASSIFICATION: RANKING OF THE PARAMETERS Irena Ilieva Jekova, Ivaylo Ivanov Christov, Lyudmila Pavlova Todorova Centre of Biomedical Engineering Prof.
More informationNOISE DETECTION ALGORITHM
ELECTRONICS 26 2-22 September, Sozopol, BULGARIA NOISE DETECTION ALGORITHM FOR AUTOMATIC EXTERNAL DEFIBRILLATORS Irena Ilieva Jekova, Vessela Tzvetanova Krasteva Centre of Biomedical Engineering Prof.
More informationMORPHOLOGICAL CHARACTERIZATION OF ECG SIGNAL ABNORMALITIES: A NEW APPROACH
MORPHOLOGICAL CHARACTERIZATION OF ECG SIGNAL ABNORMALITIES: A NEW APPROACH Mohamed O. Ahmed Omar 1,3, Nahed H. Solouma 2, Yasser M. Kadah 3 1 Misr University for Science and Technology, 6 th October City,
More informationECG Beat Recognition using Principal Components Analysis and Artificial Neural Network
International Journal of Electronics Engineering, 3 (1), 2011, pp. 55 58 ECG Beat Recognition using Principal Components Analysis and Artificial Neural Network Amitabh Sharma 1, and Tanushree Sharma 2
More informationCHAPTER 5 WAVELET BASED DETECTION OF VENTRICULAR ARRHYTHMIAS WITH NEURAL NETWORK CLASSIFIER
57 CHAPTER 5 WAVELET BASED DETECTION OF VENTRICULAR ARRHYTHMIAS WITH NEURAL NETWORK CLASSIFIER 5.1 INTRODUCTION The cardiac disorders which are life threatening are the ventricular arrhythmias such as
More informationAssessment of the Performance of the Adaptive Thresholding Algorithm for QRS Detection with the Use of AHA Database
Assessment of the Performance of the Adaptive Thresholding Algorithm for QRS Detection with the Use of AHA Database Ivaylo Christov Centre of Biomedical Engineering Prof. Ivan Daskalov Bulgarian Academy
More informationTesting the Accuracy of ECG Captured by Cronovo through Comparison of ECG Recording to a Standard 12-Lead ECG Recording Device
Testing the Accuracy of ECG Captured by through Comparison of ECG Recording to a Standard 12-Lead ECG Recording Device Data Analysis a) R-wave Comparison: The mean and standard deviation of R-wave amplitudes
More informationISSN: ISO 9001:2008 Certified International Journal of Engineering and Innovative Technology (IJEIT) Volume 2, Issue 10, April 2013
ECG Processing &Arrhythmia Detection: An Attempt M.R. Mhetre 1, Advait Vaishampayan 2, Madhav Raskar 3 Instrumentation Engineering Department 1, 2, 3, Vishwakarma Institute of Technology, Pune, India Abstract
More informationVital Responder: Real-time Health Monitoring of First- Responders
Vital Responder: Real-time Health Monitoring of First- Responders Ye Can 1,2 Advisors: Miguel Tavares Coimbra 2, Vijayakumar Bhagavatula 1 1 Department of Electrical & Computer Engineering, Carnegie Mellon
More informationAUTOMATIC ANALYSIS AND VISUALIZATION OF MULTILEAD LONG-TERM ECG RECORDINGS
AUTOMATIC ANALYSIS AND VISUALIZATION OF MULTILEAD LONG-TERM ECG RECORDINGS Vessela Tzvetanova Krasteva 1, Ivo Tsvetanov Iliev 2 1 Centre of Biomedical Engineering Prof. Ivan Daskalov - Bulgarian Academy
More informationWavelet Decomposition for Detection and Classification of Critical ECG Arrhythmias
Proceedings of the 8th WSEAS Int. Conference on Mathematics and Computers in Biology and Chemistry, Vancouver, Canada, June 19-21, 2007 80 Wavelet Decomposition for Detection and Classification of Critical
More informationAssessment of Reliability of Hamilton-Tompkins Algorithm to ECG Parameter Detection
Proceedings of the 2012 International Conference on Industrial Engineering and Operations Management Istanbul, Turkey, July 3 6, 2012 Assessment of Reliability of Hamilton-Tompkins Algorithm to ECG Parameter
More informationBIOAUTOMATION, 2009, 13 (2), 84-96
Rhythm Analysis by Heartbeat Classification in the Electrocardiogram (Review article of the research achievements of the members of the Centre of Biomedical Engineering, Bulgarian Academy of Sciences)
More informationUSING CORRELATION COEFFICIENT IN ECG WAVEFORM FOR ARRHYTHMIA DETECTION
BIOMEDICAL ENGINEERING- APPLICATIONS, BASIS & COMMUNICATIONS USING CORRELATION COEFFICIENT IN ECG WAVEFORM FOR ARRHYTHMIA DETECTION 147 CHUANG-CHIEN CHIU 1,2, TONG-HONG LIN 1 AND BEN-YI LIAU 2 1 Institute
More informationTemporal Analysis and Remote Monitoring of ECG Signal
Temporal Analysis and Remote Monitoring of ECG Signal Amruta Mhatre Assistant Professor, EXTC Dept. Fr.C.R.I.T. Vashi Amruta.pabarekar@gmail.com Sadhana Pai Associate Professor, EXTC Dept. Fr.C.R.I.T.
More informationSequential algorithm for life threatening cardiac pathologies detection based on mean signal strength and EMD functions
Anas et al. BioMedical Engineering OnLine 2, 9:43 RESEARCH Open Access Sequential algorithm for life threatening cardiac pathologies detection based on mean signal strength and EMD functions Emran M Abu
More informationCHAPTER IV PREPROCESSING & FEATURE EXTRACTION IN ECG SIGNALS
CHAPTER IV PREPROCESSING & FEATURE EXTRACTION IN ECG SIGNALS are The proposed ECG classification approach consists of three phases. They Preprocessing Feature Extraction and Selection Classification The
More informationECG Signal Analysis for Abnormality Detection in the Heart beat
GRD Journals- Global Research and Development Journal for Engineering Volume 1 Issue 10 September 2016 ISSN: 2455-5703 ECG Signal Analysis for Abnormality Detection in the Heart beat Vedprakash Gujiri
More informationAvailable online at ScienceDirect. Procedia Technology 24 (2016 )
Available online at www.sciencedirect.com ScienceDirect Procedia Technology 4 (016 ) 1406 1414 International Conference on Emerging Trends in Engineering, Science and Technology (ICETEST - 015) Cardiac
More informationDETECTION OF HEART ABNORMALITIES USING LABVIEW
IASET: International Journal of Electronics and Communication Engineering (IJECE) ISSN (P): 2278-9901; ISSN (E): 2278-991X Vol. 5, Issue 4, Jun Jul 2016; 15-22 IASET DETECTION OF HEART ABNORMALITIES USING
More informationPremature Ventricular Contraction Arrhythmia Detection Using Wavelet Coefficients
IOSR Journal of Electronics and Communication Engineering (IOSR-JECE) e-issn: 2278-2834,p- ISSN: 2278-8735.Volume 9, Issue 2, Ver. V (Mar - Apr. 2014), PP 24-28 Premature Ventricular Contraction Arrhythmia
More informationInternational Journal of Advanced Research in Computer Science and Software Engineering
ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: AR Modeling for Automatic Cardiac Arrhythmia Diagnosis using
More informationHST-582J/6.555J/16.456J-Biomedical Signal and Image Processing-Spring Laboratory Project 1 The Electrocardiogram
HST-582J/6.555J/16.456J-Biomedical Signal and Image Processing-Spring 2007 DUE: 3/8/07 Laboratory Project 1 The Electrocardiogram 1 Introduction The electrocardiogram (ECG) is a recording of body surface
More informationBasic Dysrhythmia Interpretation
Basic Dysrhythmia Interpretation Objectives 2 To understand the Basic ECG To understand the meaning of Dysrhythmia To describe the normal heart conduction system. To describe the normal impulse pathways.
More informationAutomatic Identification of Implantable Cardioverter-Defibrillator Lead Problems Using Intracardiac Electrograms
Automatic Identification of Implantable Cardioverter-Defibrillator Lead Problems Using Intracardiac Electrograms BD Gunderson, AS Patel, CA Bounds Medtronic, Inc., Minneapolis, USA Abstract Implantable
More informationBiomedical. Measurement and Design ELEC4623. Lectures 15 and 16 Statistical Algorithms for Automated Signal Detection and Analysis
Biomedical Instrumentation, Measurement and Design ELEC4623 Lectures 15 and 16 Statistical Algorithms for Automated Signal Detection and Analysis Fiducial points Fiducial point A point (or line) on a scale
More informationWe are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors
We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists 3,9 116, 12M Open access books available International authors and editors Downloads Our authors
More informationMULTILEAD SIGNAL PREPROCESSING BY LINEAR TRANSFORMATION
MULTILEAD SIGNAL PREPROCESSING BY LINEAR TRANSFORMATION TO DERIVE AN ECG LEAD WHERE THE ATYPICAL BEATS ARE ENHANCED Chavdar Lev Levkov Signa Cor Laboratory, Sofia, Bulgaria, info@signacor.com ECG signal
More informationPCA Enhanced Kalman Filter for ECG Denoising
IOSR Journal of Electronics & Communication Engineering (IOSR-JECE) ISSN(e) : 2278-1684 ISSN(p) : 2320-334X, PP 06-13 www.iosrjournals.org PCA Enhanced Kalman Filter for ECG Denoising Febina Ikbal 1, Prof.M.Mathurakani
More information-RHYTHM PRACTICE- By Dr.moanes Msc.cardiology Assistant Lecturer of Cardiology Al Azhar University. OBHG Education Subcommittee
-RHYTHM PRACTICE- By Dr.moanes Msc.cardiology Assistant Lecturer of Cardiology Al Azhar University The Normal Conduction System Sinus Node Normal Sinus Rhythm (NSR) Sinus Bradycardia Sinus Tachycardia
More informationA Novel Application of Wavelets to Real-Time Detection of R-waves
A Novel Application of Wavelets to Real-Time Detection of R-waves Katherine M. Davis,* Richard Ulrich and Antonio Sastre I Introduction In recent years, medical, industrial and military institutions have
More informationAssessing Arrhythmia Performance ST/AR Algorithm
Assessing Arrhythmia Performance ST/AR Algorithm Application Note This report provides the arrhythmia performance of the ST/AR (ST and Arrhythmia) algorithm. For a description of the algorithm, see the
More informationA Review on Arrhythmia Detection Using ECG Signal
A Review on Arrhythmia Detection Using ECG Signal Simranjeet Kaur 1, Navneet Kaur Panag 2 Student 1,Assistant Professor 2 Dept. of Electrical Engineering, Baba Banda Singh Bahadur Engineering College,Fatehgarh
More informationHeart Rate Calculation by Detection of R Peak
Heart Rate Calculation by Detection of R Peak Aditi Sengupta Department of Electronics & Communication Engineering, Siliguri Institute of Technology Abstract- Electrocardiogram (ECG) is one of the most
More informationUNDERSTANDING YOUR ECG: A REVIEW
UNDERSTANDING YOUR ECG: A REVIEW Health professionals use the electrocardiograph (ECG) rhythm strip to systematically analyse the cardiac rhythm. Before the systematic process of ECG analysis is described
More informationExtraction of Unwanted Noise in Electrocardiogram (ECG) Signals Using Discrete Wavelet Transformation
Extraction of Unwanted Noise in Electrocardiogram (ECG) Signals Using Discrete Wavelet Transformation Er. Manpreet Kaur 1, Er. Gagandeep Kaur 2 M.Tech (CSE), RIMT Institute of Engineering & Technology,
More informationDesign of the HRV Analysis System Based on AD8232
207 3rd International Symposium on Mechatronics and Industrial Informatics (ISMII 207) ISB: 978--60595-50-8 Design of the HRV Analysis System Based on AD8232 Xiaoqiang Ji,a, Chunyu ing,b, Chunhua Zhao
More informationWavelet based detection of ventricular arrhythmias with neural network classifier
Vol.2, No.6, 439-444 (2009) doi:10.4236/jbise.2009.26064 JBiSE Wavelet based detection of ventricular arrhythmias with neural network classifier Sankara Subramanian Arumugam 1, Gurusamy Gurusamy 2, Selvakumar
More informationAn ECG Beat Classification Using Adaptive Neuro- Fuzzy Inference System
An ECG Beat Classification Using Adaptive Neuro- Fuzzy Inference System Pramod R. Bokde Department of Electronics Engineering, Priyadarshini Bhagwati College of Engineering, Nagpur, India Abstract Electrocardiography
More informationECG Signal Characterization and Correlation To Heart Abnormalities
ECG Signal Characterization and Correlation To Heart Abnormalities Keerthi G Reddy 1, Dr. P A Vijaya 2, Suhasini S 3 1PG Student, 2 Professor and Head, Department of Electronics and Communication, BNMIT,
More informationPOWER EFFICIENT PROCESSOR FOR PREDICTING VENTRICULAR ARRHYTHMIA BASED ON ECG
POWER EFFICIENT PROCESSOR FOR PREDICTING VENTRICULAR ARRHYTHMIA BASED ON ECG Anusha P 1, Madhuvanthi K 2, Aravind A.R 3 1 Department of Electronics and Communication Engineering, Prince Shri Venkateshwara
More informationRemoval of Baseline wander and detection of QRS complex using wavelets
International Journal of Scientific & Engineering Research Volume 3, Issue 4, April-212 1 Removal of Baseline wander and detection of QRS complex using wavelets Nilesh Parihar, Dr. V. S. Chouhan Abstract
More informationA RECOGNITION OF ECG ARRHYTHMIAS USING ARTIFICIAL NEURAL NETWORKS
A RECOGNITION OF ECG ARRHYTHMIAS USING ARTIFICIAL NEURAL NETWORKS Yüksel Özbay 1 and Bekir Karlik 2 1 Selcuk University, Electrical & Electronics Eng., Konya, Turkey 2 Ege University, International Computing
More informationRate: The atrial and ventricular rates are equal; heart rate is greater than 100 bpm (usually between bpm).
Sinus Bradycardia Regularity: The R-R intervals are constant; the rhythm is regular. Rate: The atrial and ventricular rates are equal; heart rate is less than 60 bpm. P wave: There is a uniform P wave
More informationA Novel Approach for Different Morphological Characterization of ECG Signal
A Novel Approach for Different Morphological Characterization of ECG Signal R. Harikumar and S. N. Shivappriya Abstract The earlier detection of Cardiac arrhythmia of ECG waves is important to prevent
More informationLABVIEW based expert system for Detection of heart abnormalities
LABVIEW based expert system for Detection of heart abnormalities Saket Jain Piyush Kumar Monica Subashini.M School of Electrical Engineering VIT University, Vellore - 632014, Tamil Nadu, India Email address:
More informationPanorama. Arrhythmia Analysis Frequently Asked Questions
Panorama Arrhythmia Analysis Frequently Asked Questions What ECG vectors are used for Beat Detection? 3-wire lead set 5-wire lead set and 12 lead What ECG vectors are used for Beat Typing? 3-wire lead
More informationPerformance Identification of Different Heart Diseases Based On Neural Network Classification
Performance Identification of Different Heart Diseases Based On Neural Network Classification I. S. Siva Rao Associate Professor, Department of CSE, Raghu Engineering College, Visakhapatnam, Andhra Pradesh,
More informationReal-time Heart Monitoring and ECG Signal Processing
Real-time Heart Monitoring and ECG Signal Processing Fatima Bamarouf, Claire Crandell, and Shannon Tsuyuki Advisors: Drs. Yufeng Lu and Jose Sanchez Department of Electrical and Computer Engineering Bradley
More informationRobust system for patient specific classification of ECG signal using PCA and Neural Network
International Research Journal of Engineering and Technology (IRJET) e-issn: 395-56 Volume: 4 Issue: 9 Sep -7 www.irjet.net p-issn: 395-7 Robust system for patient specific classification of using PCA
More information2012, IJARCSSE All Rights Reserved Page 402
ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: Efficient ECG Abnormalities Recognition Using Neuro-Fuzzy Approach
More informationREVIEW ON ARRHYTHMIA DETECTION USING SIGNAL PROCESSING
REVIEW ON ARRHYTHMIA DETECTION USING SIGNAL PROCESSING Vishakha S. Naik Dessai Electronics and Telecommunication Engineering Department, Goa College of Engineering, (India) ABSTRACT An electrocardiogram
More informationQuick detection of QRS complexes and R-waves using a wavelet transform and K-means clustering
Bio-Medical Materials and Engineering 26 (2015) S1059 S1065 DOI 10.3233/BME-151402 IOS Press S1059 Quick detection of QRS complexes and R-waves using a wavelet transform and K-means clustering Yong Xia
More informationDynamic Time Warping As a Novel Tool in Pattern Recognition of ECG Changes in Heart Rhythm Disturbances
2005 IEEE International Conference on Systems, Man and Cybernetics Waikoloa, Hawaii October 10-12, 2005 Dynamic Time Warping As a Novel Tool in Pattern Recognition of ECG Changes in Heart Rhythm Disturbances
More informationQUANTIFICATION OF ELECTROCARDIOGRAM RHYTHMICITY TO DETECT LIFE THREATENING CARDIAC ARRHYTHMIAS USING SPECTRAL ENTROPY
Journal of Engineering Science and Technology Vol. 8, No. 5 (23) 588-62 School of Engineering, Taylor s University QUANTIFICATION OF ELECTROCARDIOGRAM RHYTHMICITY TO DETECT LIFE THREATENING CARDIAC ARRHYTHMIAS
More informationAutomatic Detection of Heart Disease Using Discreet Wavelet Transform and Artificial Neural Network
e-issn: 2349-9745 p-issn: 2393-8161 Scientific Journal Impact Factor (SJIF): 1.711 International Journal of Modern Trends in Engineering and Research www.ijmter.com Automatic Detection of Heart Disease
More informationNeural Network based Heart Arrhythmia Detection and Classification from ECG Signal
Neural Network based Heart Arrhythmia Detection and Classification from ECG Signal 1 M. S. Aware, 2 V. V. Shete *Dept. of Electronics and Telecommunication, *MIT College Of Engineering, Pune Email: 1 mrunal_swapnil@yahoo.com,
More information1196 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 7, JULY 2004
1196 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 7, JULY 2004 Automatic Classification of Heartbeats Using ECG Morphology and Heartbeat Interval Features Philip de Chazal*, Member, IEEE,
More informationArrhythmic Complications of MI. Teferi Mitiku, MD Assistant Clinical Professor of Medicine University of California Irvine
Arrhythmic Complications of MI Teferi Mitiku, MD Assistant Clinical Professor of Medicine University of California Irvine Objectives Brief overview -Pathophysiology of Arrhythmia ECG review of typical
More informationBiomedical Signal Processing
DSP : Biomedical Signal Processing What is it? Biomedical Signal Processing: Application of signal processing methods, such as filtering, Fourier transform, spectral estimation and wavelet transform, to
More informationD)
TFQO: Rudolph Koster Author(s): Alfredo Sierra & Kevin Nation EVREV Date 12 July 2014 QUESTION: Among adults and children who are in cardiac arrest in any setting (P), does analysis of cardiac rhythm during
More informationAutomatic Detection of Abnormalities in ECG Signals : A MATLAB Study
Automatic Detection of Abnormalities in ECG Signals : A MATLAB Study M. Hamiane, I. Y. Al-Heddi Abstract The Electrocardiogram (ECG) is a diagnostic tool that measures and records the electrical activity
More informationAutomated Diagnosis of Cardiac Health
Automated Diagnosis of Cardiac Health Suganya.V 1 M.E (Communication Systems), K. Ramakrishnan College of Engineering, Trichy, India 1 ABSTRACT Electrocardiogram (ECG) is the P, QRS, T wave representing
More informationECG Rhythm Analysis by Using Neuro-Genetic Algorithms
MASAUM Journal of Basic and Applied Sciences, Vol. 1, No. 3, October 2009 522 ECG Rhythm Analysis by Using Neuro-Genetic Algorithms Safaa S. Omran, S.M.R. Taha, and Nassr Ali Awadh Abstract The heart is
More informationDelineation of QRS-complex, P and T-wave in 12-lead ECG
IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.4, April 2008 185 Delineation of QRS-complex, P and T-wave in 12-lead ECG V.S. Chouhan, S.S. Mehta and N.S. Lingayat Department
More informationPrediction of Ventricular Tachyarrhythmia in Electrocardiograph Signal using Neuro-Wavelet Approach
iction of Ventricular Tachyarrhythmia in Electrocardiograph Signal using Neuro-Wavelet Approach Rahat Abbas, Wajid Aziz, Muhammad Arif rahat_abbas@yahoo.com, kh_wajid@yahoo.com and marif@pieas.edu.pk Department
More informationSimulation Based R-peak and QRS complex detection in ECG Signal
Simulation Based R-peak and QRS complex detection in ECG Signal Name: Bishweshwar Pratap Tasa Designation: Student, Organization: College: DBCET, Azara, Guwahati, Email ID: bish94004@gmail.com Name: Pompy
More informationECG QRS Detection. Valtino X. Afonso
12 ECG QRS Detection Valtino X. Afonso Over the past few years, there has been an increased trend toward processing of the electrocardiogram (ECG) using microcomputers. A survey of literature in this research
More informationFuzzy Based Early Detection of Myocardial Ischemia Using Wavelets
Fuzzy Based Early Detection of Myocardial Ischemia Using Wavelets Jyoti Arya 1, Bhumika Gupta 2 P.G. Student, Department of Computer Science, GB Pant Engineering College, Ghurdauri, Pauri, India 1 Assistant
More informationA MULTI-STAGE NEURAL NETWORK CLASSIFIER FOR ECG EVENTS
A MULTI-STAGE NEURAL NETWORK CLASSIFIER FOR ECG EVENTS H. Gholam Hosseini 1, K. J. Reynolds 2, D. Powers 2 1 Department of Electrotechnology, Auckland University of Technology, Auckland, New Zealand 2
More informationModule 1: Introduction to ECG & Normal ECG
Module 1: Introduction to ECG & Normal ECG Importance of Correct anatomical positions Measurements & Morphologies ONLY accurate if Precise anatomical positions adhered to Standardised techniques are used
More informationDeveloping Electrocardiogram Mathematical Model for Cardiovascular Pathological Conditions and Cardiac Arrhythmia
Indian Journal of Science and Technology, Vol 8(S10), DOI: 10.17485/ijst/015/v8iS10/84847, December 015 ISSN (Print) : 0974-6846 ISSN (Online) : 0974-5645 Developing Electrocardiogram Mathematical Model
More informationDETECTION OF EVENTS AND WAVES 183
DETECTON OF EVENTS AND WAVES 183 4.3.1 Derivative-based methods for QRS detection Problem: Develop signal processing techniques to facilitate detection of the QRS complex, given that it is the sharpest
More informationGE Healthcare. The GE EK-Pro Arrhythmia Detection Algorithm for Patient Monitoring
GE Healthcare The GE EK-Pro Arrhythmia Detection Algorithm for Patient Monitoring Table of Contents Arrhythmia monitoring today 3 The importance of simultaneous, multi-lead arrhythmia monitoring 3 GE EK-Pro
More informationClassification of electrocardiographic ST-T segments human expert vs artificial neural network
European Heart Journal (1993) 14,464-468 Classification of electrocardiographic ST-T segments human expert vs artificial neural network L. EDENBRANDT, B. DEVINE AND P. W. MACFARLANE University Department
More informationPERFORMANCE CALCULATION OF WAVELET TRANSFORMS FOR REMOVAL OF BASELINE WANDER FROM ECG
PERFORMANCE CALCULATION OF WAVELET TRANSFORMS FOR REMOVAL OF BASELINE WANDER FROM ECG AMIT KUMAR MANOCHA * Department of Electrical and Electronics Engineering, Shivalik Institute of Engineering & Technology,
More informationRemoving ECG Artifact from the Surface EMG Signal Using Adaptive Subtraction Technique
www.jbpe.org Removing ECG Artifact from the Surface EMG Signal Using Adaptive Subtraction Technique Original 1 Department of Biomedical Engineering, Amirkabir University of technology, Tehran, Iran Abbaspour
More informationGenetic Algorithm based Feature Extraction for ECG Signal Classification using Neural Network
Genetic Algorithm based Feature Extraction for ECG Signal Classification using Neural Network 1 R. Sathya, 2 K. Akilandeswari 1,2 Research Scholar 1 Department of Computer Science 1 Govt. Arts College,
More informationdiseases is highly important. Electrocardiogram (ECG) is an effective tool for diagnosis of the
Chapter 1 Introduction 1.1 Introduction Heart disease is one of the major reasons for high mortality rate. Accurate diagnosis of heart diseases is highly important. Electrocardiogram (ECG) is an effective
More informationEHRA Accreditation Exam - Sample MCQs Cardiac Pacing and ICDs
EHRA Accreditation Exam - Sample MCQs Cardiac Pacing and ICDs Dear EHRA Member, Dear Colleague, As you know, the EHRA Accreditation Process is becoming increasingly recognised as an important step for
More informationInternational Journal of Advance Engineering and Research Development
Scientific Journal of Impact Factor (SJIF): 4.72 International Journal of Advance Engineering and Research Development Volume 4, Issue 11, November -2017 e-issn (O): 2348-4470 p-issn (P): 2348-6406 Analysis
More informationComparison of Different ECG Signals on MATLAB
International Journal of Electronics and Computer Science Engineering 733 Available Online at www.ijecse.org ISSN- 2277-1956 Comparison of Different Signals on MATLAB Rajan Chaudhary 1, Anand Prakash 2,
More informationContinuous Wavelet Transform in ECG Analysis. A Concept or Clinical Uses
1143 Continuous Wavelet Transform in ECG Analysis. A Concept or Clinical Uses Mariana Moga a, V.D. Moga b, Gh.I. Mihalas b a County Hospital Timisoara, Romania, b University of Medicine and Pharmacy Victor
More informationLearning Decision Tree for Selecting QRS Detectors for Cardiac Monitoring
Learning Decision Tree for Selecting QRS Detectors for Cardiac Monitoring François Portet 1, René Quiniou 2, Marie-Odile Cordier 2, and Guy Carrault 3 1 Department of Computing Science, University of Aberdeen,
More informationMulti Resolution Analysis of ECG for Arrhythmia Using Soft- Computing Techniques
RESEARCH ARTICLE OPEN ACCESS Multi Resolution Analysis of ECG for Arrhythmia Using Soft- Computing Techniques Mangesh Singh Tomar 1, Mr. Manoj Kumar Bandil 2, Mr. D.B.V.Singh 3 Abstract in this paper,
More informationSVT Discriminators. Definition of SVT Discrimination. Identify which patient populations might benefit from these features
Definition of SVT Discrimination Identify which patient populations might benefit from these features Understand the 4 types of SVT discriminators used by St Jude Medical Be aware of programmable parameters
More informationCLASSIFICATION OF CARDIAC SIGNALS USING TIME DOMAIN METHODS
CLASSIFICATION OF CARDIAC SIGNALS USING TIME DOMAIN METHODS B. Anuradha, K. Suresh Kumar and V. C. Veera Reddy Department of Electrical and Electronics Engineering, S.V.U. College of Engineering, Tirupati,
More informationVarious Methods To Detect Respiration Rate From ECG Using LabVIEW
Various Methods To Detect Respiration Rate From ECG Using LabVIEW 1 Poorti M. Vyas, 2 Dr. M. S. Panse 1 Student, M.Tech. Electronics 2.Professor Department of Electrical Engineering, Veermata Jijabai Technological
More informationAn electrocardiogram (ECG) is a recording of the electricity of the heart. Analysis of ECG
Introduction An electrocardiogram (ECG) is a recording of the electricity of the heart. Analysis of ECG data can give important information about the health of the heart and can help physicians to diagnose
More informationPowerline Interference Reduction in ECG Using Combination of MA Method and IIR Notch
International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 29 Powerline Interference Reduction in ECG Using Combination of MA Method and IIR Notch Manpreet Kaur, Birmohan Singh 2 Department
More informationDEVELOPMENT OF A SIMPLE SOFTWARE TOOL TO DETECT THE QRS COMPLEX FROM THE ECG SIGNAL
DEVELOPMENT OF A SIMPLE SOFTWARE TOOL TO DETECT THE QRS COMPLEX FROM THE ECG SIGNAL Michaella Ignatia Tanoeihusada 1), Wahju Sediono 2) Swiss German University, Tangerang 1), Agency for the Assessment
More informationKeywords : Neural Pattern Recognition Tool (nprtool), Electrocardiogram (ECG), MIT-BIH database,. Atrial Fibrillation, Malignant Ventricular
Volume 7, Issue 2, February 2017 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Identification
More informationKeywords: Adaptive Neuro-Fuzzy Interface System (ANFIS), Electrocardiogram (ECG), Fuzzy logic, MIT-BHI database.
Volume 3, Issue 11, November 2013 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Detection
More informationSPPS: STACHOSTIC PREDICTION PATTERN CLASSIFICATION SET BASED MINING TECHNIQUES FOR ECG SIGNAL ANALYSIS
www.iioab.org www.iioab.webs.com ISSN: 0976-3104 SPECIAL ISSUE: Emerging Technologies in Networking and Security (ETNS) ARTICLE OPEN ACCESS SPPS: STACHOSTIC PREDICTION PATTERN CLASSIFICATION SET BASED
More informationManual Defibrillators, Automatic External Defibrillators, Cardioversion, and External Pacing. D. J. McMahon cewood rev
Manual Defibrillators, Automatic External Defibrillators, Cardioversion, and External Pacing D. J. McMahon 141001 cewood rev 2017-10-04 Key Points Defibrillators: - know the definition & electrical value
More informationClinical Examples as Non-uniform Learning and Testing Sets
Clinical Examples as Non-uniform Learning and Testing Sets Piotr Augustyniak AGH University of Science and Technology, 3 Mickiewicza Ave. 3-9 Krakow, Poland august@agh.edu.pl Abstract. Clinical examples
More informationStudy of Transthoracic Impedance Cardiogram for Assessment of Cardiac Hemodynamics in Atrial Fibrillation Patents
Study of Transthoracic Impedance Cardiogram for Assessment of Cardiac Hemodynamics in Atrial Fibrillation Patents Vessela Krasteva 1*, Irena Jekova 1, Elina Trendafilova 2, Sarah Ménétré 3, Tsvetan Mudrov
More informationDetection of Qrs Complexes in Ecg Signal Using K-Means Algorithm
Detection of Qrs Complexes in Ecg Signal Using K-Means Algorithm Ms. Anaya A. Dange M Tech Student Prof. Dr. S. L. Nalbalwar Prof. & Head Department of Electronics & Telecommunication Engineering, Dr.
More informationPortable ECG Electrodes for Detection of Heart Rate and Arrhythmia Classification
Portable ECG Electrodes for Detection of Heart Rate and Arrhythmia Classification 1 K. Jeeva, 2 Dr. D. Selvaraj, 3 Dr. S. Leones Sherwin Vimal Raj 1 PG Student, 2, 3 Professor, Department of Electronics
More information