Time-frequency analysis of microvolt T-wave alternans in chronic Chagas heart disease

Abstract

Chagas disease is an endemic infectious disease caused by theprotozoa Trypanosoma cruzi , currently affecting 10–12 million subjectsintheworld [1],primarilyrestrictedtotheAmericancontinent.Infectionis generally acquired early in childhood, with mild nonspecific clinicalpresentation, evolving into a dormant asymptomatic course, defined asindeterminatephase.Inatimecourserangingfrom10to30 years,20to30% of contaminated subjects develop cardiac abnormalities character-izedbyenlargementofcardiacchambersandlifethreateningventriculararrhythmia, the so called Chagas heart disease (ChD). A remarkablecharacteristic of ChD is the occurrence of severe ventricular tachyar-rhythmia(VT)orevensuddencardiacdeath(SCD)asafirstsignofheartdisease. Thus, the development of a diagnostic and risk stratificationmethodtodetectthosesubjectsatahigherriskismandatory,includingthose who may eventually benefit of an implantable cardioverter-defibrillator [2].Microvolt T-wave alternans (MTWA) is a promising noninvasivediagnostic tool, considered capable of detecting subjects with eitherischemicornon-ischemicheartdiseaseatahigherriskforVTandSCD.The method is based on the spectral decomposition of the T-waveamplitude time series of consecutive heartbeats. The occurrence ofalternans is confirmed when the spectral peak the 0.5 cycle-per-beatfrequencyexceedsa predefinedthreshold[3]. Sinceits initialproposaland standardization on early nineties [4,5], several studies hasdemonstrated its clinical utility for risk stratification. However, ithasnotbeentestedinChD.Thepurposeofcurrentstudyistodescribea method and its application for assessing MTWA in ChD.High resolution ECG signals were acquired using orthogonalbipolar XYZ Frank leads during 3 min, in supine position in quietand comfortable environment. Medical instrumentation employed,signal acquisition protocol and pre-processing techniques have beendescribed elsewhere [6]. The routine of ECG signal processingcomprises the following phases: i) Detection of maximum absoluteQRS complex peaks; ii) Identification and delimitation of successiveT-waves; iii) Detection of maximal absolute T-wave peaks, iv)Deployment of a T-wave peaks time series; and v) Time-frequencyanalysis of the generated time series.The QRS complex detection algorithm was based on Pan and Tomp-kins method [7] associated to correlation analysis on a convenienttemplate, which appropriately selected heart beats. A beat editingfacility allowed parameters change for improving detection accuracyand ectopic beats rejection by visual inspection.T-wave detection algorithm considered in short, the followingsteps: i) The onset and the offset of selected T-wave template wasmanually defined; ii) Considering the rate-adaptation processed asdescribed by Bazett formula, the routine further determined theboundaries of the T-wave for each subsequent beat, as oriented byboth the preceding RR interval and the T-wave template interval [8](Fig. 1-a); iii) The maximum absolute value of the analyzing T-waveinterval identified the T-wave peak (Fig. 1-b); iv) A time series ofconsecutiveT-wavepeakswasthusdeployed,whereT-wavealternanswould be eventually assessed (Fig. 1-c). An eventual T-wave peakarising from an ectopic beat in the series was replaced by the averageof accepted beats to avoid spurious oscillation.The spectral analysis was carried out on 3-min T-wave peaktime series segment using a moving short-time Fourier transformalgorithm. To build a time-frequency map, the analyzing time serieswas segmented in 128 beats windows (Fig. 1-d). Each window wasdetrended,multipliedbyaHanningwindowtoavoiddiscontinuities,and then submitted to spectral decomposition via FFT. Each spectralestimate was, then, squared and appropriately transformed tocompose a power spectral density function time–frequency map(Fig. 1-e), in which the following classical indexes were analyzedat each time slice (Fig. 1-f): i) Cumulative voltage alternans (P)(defined as the squared root of the difference between the peakspectral amplitude at 0.5 cycle per beat and the average spectralnoise), where PN1.9 μV was abnormal; and ii) Alternans ration (K)(defined as the difference of the spectral amplitude peak at 0.5 cycleper beat and the average spectral noise, divided by the standarddeviation of spectral noise), where KN3 were abnormal. In thetime frequency map, the Power spectrum is continuously assessedon 128 beats window during the 3-min segment. If at any moment,either indexes overtakes the corresponding normality threshold,alternans is defined.The study protocol was approved by National Institute ofCardiology Ethics Committee (National institute of cardiology proto-col #0190/12.02.2008).A clinically stable subject with ChD, with left ventricular ejectionfraction of 35%, spontaneous episodes of nonsustained ventriculartachycardiaandanimplantedventricularpacemaker-cardioverterwasinvited to participate and provided written informed consent.Ventricular pacemaker was set to an initial ventricular rate of90 bpm, gradually and continuously increased up to 110 bpm andthen decreased back to 100 bpm.The MTWA presence was detected in the ECG signal at aheart rate of 100 bpm. Lead Y provided the best results, and thevoltage alternans (P) and alternans ratio (K) values were, res-pectively,1.2 µV and 3.4 units. Theanalyzed time series of consecutiveT-wave peaks (128 points) where MTWA was detected had noectopic beats.