Silvia Bini

Comparative analysis of methods for automatic detection and quantification of microvolt T-wave alternans

Microvolt T-wave alternans (TWA), consisting of every-other-beat changes in ECG T-wave morphology, is an index of susceptibility to malignant ventricular arrhythmias, requiring automatic techniques to be identified. Five of these, namely, fast-Fourier-transform spectral method (FFTSM), complex-demodulation method (CDM), modified-moving-average method (MMAM), Laplacian-likelihood-ratio method (LLRM) and adaptive-match-filter method (AMFM), were applied here to simulated and sample clinical data. The aim was to compare individual methods ability to properly identify stationary and time-varying TWA, avoiding false-positive detections. The MMAM provided false-positive TWA when applied to simulated ECGs affected by amplitude variability, but TWA. Stationary TWA was properly quantified by the MMAM and, occasionally, underestimated by all other methods. The AMFM properly identified time-varying TWA. By contrast, the FFTSM detected not-stationary TWA as stationary, the MMAM introduced a time-delay in the estimated TWA-amplitude signal, while the CDM and LLRM were reliable only in the presence of slow-varying TWA. Altogether, the AMFM accomplished the best compromise between the needs to avoid false-positive TWA and to detect and characterize true-positive TWA. Results of our simulation approach were useful to explain different TWA levels measured by each competing methods applied to sample Holter ECGs from healthy subjects and coronary artery disease patients.

Automatic microvolt T-wave alternans identification in relation to ECG interferences surviving preprocessing

The aim was to investigate the effect of interferences surviving preprocessing (residual noise, baseline wanderings, respiration modulation, replaced beats, missed beats and T-waves misalignment) on automatic identification of T-wave alternans (TWA), an ECG index of risk for sudden cardiac death. The procedures denominated fast-Fourier-transform spectral method (FFTSM), complex-demodulation method (CDM), modified-moving-average method (MMAM), Laplacian-likelihood-ratio method (LLRM), and adaptive-match-filter method (AMFM) were applied to interferences-corrupted synthetic ECG tracings and Holter ECG recordings from control-healthy subjects (CH-group; n=25) and acute-myocardial-infarction patients (AMI group; n=25). The presence of interferences in simulated data caused detection of false-positive TWA by all techniques but the FFTSM and AMFM. Clinical applications evidenced a discrepancy in that the FFTSM and LLRM detected no more than one TWA case in each population, whereas the CDM, MMAM, and AMFM detected TWA in all CH-subjects and AMI-patients, with significantly lower TWA amplitude in the former group. Because the AMFM is not prone to false-positive TWA detections, the latter finding suggests TWA as a phenomenon having continuously changing amplitude from physiological to pathological conditions. Only occasional detection of TWA by the FFTSM and LLRM in clinics can be ascribed to their limited ability in identifying TWA in the presence of interferences surviving preprocessing.

Repolarization alternans heterogeneity in healthy subjects and acute myocardial infarction patients

An association between heterogeneity of repolarization alternans (RA) and cardiac electrical instability has been reported. Characterization of RA in health and identification of physiological RA heterogeneity may help discrimination of abnormal RA cases more likely associated to arrhythmic events. Thus, aim of the present study was the identification of a physiological RA region in terms of mean temporal location (MRAD) with respect to the T apex, and mean amplitude (MRAA), by application of our heart-rate adaptive match filter method to clinical ECG recordings from 51 control healthy (CH) subjects and 43 acute myocardial infarction (AMI) patients. Results indicate that RA occurring within the first half of the T wave is dominant in both CH and AMI populations (74.5% and 53.5% of cases, respectively; P<0.05). Definition of physiological RA region in the MRAD vs. MRAA plane (-83 ms ≤ MRAD ≤ 23 ms, 0≤ MRAA ≤ 30 μV) provided 0% and 32.6% abnormal RA cases among the CH subjects and AMI patients, respectively. We conclude that myocardial infarction may associate with an RA occurring early (MRAD<-83 ms) or late (MRAD >23 ms) along the JT segment, in addition or in alternative to an abnormally high RA amplitude (MRAA >30 μV).

Correlation method versus enhanced modified moving average method for automatic detection of T-wave alternans

Enhanced modified moving average method (EMMAM) and correlation method (CM) for microvolt TWA identification are compared by aid of simulated ECG tracings (cases of absence of TWA and presence of stationary or time-varying TWA) and ECG recordings from healthy subjects (H-group) and patients who survived an acute myocardial infarction (AMI-group). The two competing methods were found to be equivalent when analyzing clean ECGs affected by stationary TWA. Non-stationary TWA is correctly tracked by the CM, whereas it is identified as stationary by the EMMAM. Moreover, the EMMAM suffers for its tendency to identify as TWA noise and other kinds of repolarization variability. Such limitation is most likely the cause of its false-positive TWA production. Finally, only the CM incorporates a local threshold criterion in the TWA detection algorithm which allows better discrimination between H and AMI groups, who are well known to be at increased risk to develop TWA.

Quantitative characterization of repolarization alternans in terms of amplitude and location: What information from different methods?

Abstract The qualitative definition of repolarization alternans (RA) as an every-other-beat alternation of the repolarization amplitude allows several possible quantitative characterizations of RA. In the absence of a standardization, any correct comparison among quantitative outputs by different automatic methods requires knowledge of the differences in the RA parameterization at the basis of their algorithms. Thus, aim of the present study was to investigate the kind of information provided by five methods, namely the fast Fourier spectral method (FFTSM), the complex demodulation method (CDM), the modified moving average method (MMAM), the Laplacian likelihood ratio method (LLRM) and the heart-rate adaptive match filter method (AMFM) when characterizing RA in terms of its amplitude and location. Eight synthetic ECG recordings affected by stationary RA with uniform and triangular profiles localized along the ST segment, over the T wave, at the end of the T wave and all along the JT segment, respectively, were considered. Results indicate that quantitative RA characterization is method dependent. More specifically, the FFTSM and the LLRM provide a measure that matches the root mean square of the RA profile over the JT segment. Instead, the CDM and the AMFM compute RA amplitude as the mean value of the RA profile over the JT segment. Eventually, the MMAM provides the maximum amplitude difference between consecutive beats along repolarization. RA location is homogeneously among methods, since they all provide the time instant in correspondence of which the center of mass of the alternation occurs.

Physiological interpretation of inductance and low-resistance terms in four-element windkessel models: Assessment by generalized sensitivity function analysis

Physiological relevance of parameters of three arterial models, denominated W4P, W4S and IVW, was assessed by computation of parameter-related generalized sensitivity functions (GSFs), which allow the definition of heart-cycle time intervals where the information content of experimental data, useful for estimation of each model parameter, is concentrated. The W4P and W4S are derived from the three-element windkessel by connecting an inductance, L, in parallel or in series, respectively, with aortic characteristic impedance, R(c). In the IVW, L is placed in series at the input of a viscoelastic windkessel, incorporating a Voigt cell (a resistor, R(d), in series with a capacitor, C). Pressure and flow measured in the ascending aorta of five ferrets and five dogs were used to estimate all model parameters, by fitting to pressure. For each model structure, parameter-related GSFs were generated. Focusing on controversial L, R(c) and R(d) physical meaning, our GSF analysis yielded the conclusion that, in both the W4S and the IVW, but not in the W4P, the L-term is suitable to represent the inertial properties of blood motion. Moreover, the meaning of aortic characteristic impedance ascribed to R(c) is questionable; while R(d) is likely to account for viscous losses of arterial wall motion.

Response to Dr. Selvaraj's Comments on the “Assessment of Physiological Amplitude, Duration and Magnitude of ECG T‐Wave Alternans”

“Assessment of Physiological Amplitude, Duration and Magnitude of ECG T-Wave Alternans” Laura Burattini, Ph.D.,∗ Silvia Bini, Dr. Eng.,∗ Wojciech Zareba, Ph.D., M.D.,† and Roberto Burattini, Er. Eng.∗ From the ∗Department of Biomedical, Electronics and Telecommunication Engineering, Polytechnic University of Marche, Ancona, Italy; and †Heart-Research Follow-Up Program, Cardiology Unit, Department of Medicine and Department of Biomedical Engineering, University of Rochester, Rochester, NY