P waves in the electrocardiogram recording of tachycardia: 'You can run, but you cannot hide'

This editorial refers to ‘A new method of filtering T waves to detect hidden P waves in ECG signals’ by D. Goldwasser et al., on page 1028. The history of the evolution of recording of cardiac electrical activities by electrocardiogram (ECG) is quite fascinating.1 The initial ECG recording systems used by Waller, Einthoven, and Lewis consisted of mercury capillary electrometers and string galvanometers that were cumbersome to use and provided limited information.1–3 In Waller's demonstration of the human ECG, only ventricular activities were recorded and the P waves were not discernible.1 Interestingly, even today there are situations where in the ECG recordings obtained with modern and sophisticated ECG machines the P waves may not be easily discernible. Interpretation of ECG recordings remains fundamental to the clinical investigation of cardiac arrhythmias. To confirm the diagnosis of ventricular tachycardia from the ECG tracing of a wide complex tachycardia would require accurate detection of the P waves to determine atrioventricular (AV) dissociation. Likewise, detection of the P waves, their morphological characteristics, axis, and timing relationships with the R waves would be important to confirm and further differentiate supraventricular tachycardias (SVTs) into AV nodal reentrant tachycardia (AVNRT), accessory bypass tract-mediated AV reentrant tachycardia (AVRT), atrial tachycardia, junctional ectopic tachycardia, or sinus tachycardia. However, simple as it sounds, the task of finding P waves let alone their characterization may be extremely difficult, if not impossible. In an approach to make a correct diagnosis of SVT recorded by 12-lead ECG, general criteria such as RP intervals, QRS alternans, pseudo-r′, pseudo-S-wave, etc. although frequently used, lack robustness in their sensitivity and specificity. Because the P waves are generated by the atrial chambers that have lower muscle mass, they are of lower amplitude when compared with the QRS complexes. As such, during SVT the recorded small P waves may be difficult to differentiate if co-incidental with the QRS complexes or the T waves. Given the close proximity to the left atrium, the oesophageal approach may be used for better recording of P waves to diagnose and even treat some SVTs.4 Another recording technique, the Lewis's lead, where the right- and left-arm electrodes are applied at the two sides of the sternum at the second and fourth intercostal spaces, respectively, may be used to magnify P waves. Of course, P waves may be recorded by intracardiac recordings during invasive electrophysiological (EP) studies. To clinicians, the engineering aspect of ECG remains overwhelmingly a source of confusion. Functionally, the ECG can be divided into different components; ECG signal acquisition and processing, real-time display, automated interpretation, storage, and transmission of the reports. Standard ECG is commonly performed with a sampling rate of 1000–200 Hz to avoid missing brief signals or introduction artefacts. Using sophisticated amplifiers, the machine enhances low-level recorded signals, and converts them into digital signals.2,3 These signals are subjected to further filtering. For routine electrocardiography, the American Heart Association standards require a bandwidth of 0.05–150 Hz for adults, with an extension to 250 Hz for children. A high-pass and a low-pass filter is used to attenuate the very low-frequency and high-frequency signals, respectively. A notch filter attenuates interference from alternating power current. However, in an attempt to attenuate or remove some frequencies from the input data, the filters may also attenuate otherwise normal components of ECGs. Pertaining to the study by Goldwasser et al.,5 the authors have built on their previous work that involved a concept of signal filtering using analogue filters to detect hidden P waves in the T waves.6 In the present study, the authors used a method of wavelet transformation that utilizes a special type of Fourier transformation to decompose the signals in different forms of representation in time and frequency. Thus, the original wavelet may be further characterized under different cut-off frequencies; a strong larger signal and a small but detectable signal corresponding to the QRS complex and the P-wave, respectively. The study cohort involved 42 patients with SVTs who underwent EP studies. The intracardiac electrograms recorded from the right atrium served to locate corresponding P waves and their relationships in the R–R interval on the ECGs. T-wave filtering and P-wave detection algorithms using wavelet transformation technique were then applied in a systematic manner. Using this methodology, the T-wave filtering and P-wave detection algorithm showed the sensitivity, specificity, and global accuracy of 85.8, 89.4, and 88.5%, respectively, for identification of P-waves. The expert's accuracy in distinguishing between AVNRT and AVRT was 75 and 100% with surface ECG and the ECG with the T-wave filtering, respectively. Although these results are modestly impressive, the study cohort is quite small. The majority (75%) of atrial signals detected in the intra-atrial ECG could not be detected by T-wave filtration as they fell within the QRS complexes. In one patient, strong noise artefacts precluded correct diagnosis. Over the years, many techniques and algorithms, none perfect and without limitations, have been described to better delineate obscured P-waves on ECG. Kuchar et al.7 used 50 Hz high-pass filtered and magnified unfiltered signal-averaged ECGs that were subjected to high-frequency analysis to temporally discriminate P-waves and assess their polarity in the X, Y, and Z leads. This method, however, being dependent on a fixed fiducial or reference timing for signal alignment could not be applied for arrhythmias with cycle length variability. Slocum et al.8 used the method of QRST subtraction to unmask P-waves during wide complex tachycardia. In their study, they applied an algorithm consisting of subtraction of a mean beat from each individual beat in leads II and V1 to generate remainder ECGs that were visually inspected for the presence of P-waves and autocorrected to automatically detect AV association or dissociation. Although very specific (specificity 100%), this method provided poor sensitivity of only 65%. Sippensgroenewegen et al.9 used body surface mapping involving an automatic 62-lead QRST subtraction algorithm to isolate ectopic P-waves from the overlying T–U wave on the surface ECG. Applications of such algorithms have limited accuracy in the presence of paced P-waves and rate-related aberrant bundle branch blocks. The issues related to changes occurring on a beat-to-beat basis such as the aberrant beats can to some extent be obviated by using the immediately preceding QRST rather than averaged QRST as subtraction template. Mehta and Lingayat10 proposed application of support vector machine method for the detection of P-waves recorded on 12-lead ECG. From the recorded ECGs, the QRS complexes and T-waves are removed and the slopes at every sampling instant for all 12 leads are obtained. The slopes are used as important criteria for wave discrimination. For P-waves, for example, the slope of the P-wave signal is much more in the P-wave region than in the region other than P-waves. The validation of the algorithm is done using 125 simultaneously recorded 12-lead ECG records of dataset of multi-lead measurement library which contains original 12-lead simultaneous ECG recordings covering a wide variety of pathological cases. The algorithm gives detection rate of 95.43% for P-waves with false-positive and -negative detections rates of 4.34 and 3.57%, respectively. This method, however, needs further validation to discern hidden P-waves during tachycardia. When the P-waves are recognizable and not linked to the R-waves, the technique described by Goldwasser et al.5 may not necessarily be of additional use unless it provides more crucial information such as the P-wave's axis and morphology. This technique, if it can be applied not just during tachycardia but also during sinus rhythm with ectopic beats, may help to determine the site of origin of the arrhythmia, and the triggers of paroxysmal atrial fibrillation recorded as distortion of T-waves during sinus rhythm. In the era of catheter-based ablation therapy, one may also argue whether it is necessary to differentiate SVTs by surface ECGs as confirmative diagnosis would be made during invasive EP study at the time of ablation procedure anyway. Certainly, prior knowledge of the arrhythmia mechanisms would be helpful in counselling the patients for the benefits and risks of the procedure, and its details such as transseptal catheterization for the left-sided arrhythmias, the use of radiofrequency vs. cryo energy, etc. In summary, for many years ECGs have been used routinely to record the electrical potentials generated by the heart. Analyses of these potentials play pivotal roles in our understanding, diagnosing, and treatment of various diseases. Although accurate recording of every potential of the ECG is important, the P-waves notoriously play a ‘hide and seek’ game. Would it not be desirable if good-quality recordings of the P-waves, hidden or not hidden, during any rhythm recordings are obtained by the surface ECG? With the resurgent interest in recording surface ECG as evident by the paper such as the one published by Goldwasser et al.5 in the journal, I am quite confident that 1 day soon the P-waves although ‘may run, may not be able to hide’. Conflict of interest: none declared.