Standard 6-lead ECGs have been used clinically for over 100 years to document heart rate and diagnose cardiac arrhythmias.1,2 Recently, a relatively inexpensive, novel technology with a dedicated interface capable of recording ECGs with a personal mobile communication device (ie, smartphone or tablet) has been developed. The device” incorporates electrodes in a handheld case that can be attached directly to a smartphone or tablet (iOSb or Android platform) or connected via a Bluetooth wireless signal, allowing recording of ECGs. It was approved by the FDA for use in humans in 20123 and allows relatively easy and rapid generation of a single-lead ECG rhythm strip that users can store, print, or forward to other individuals for review. In humans, this wireless system has been used to assess response to drug treatment,4 monitor patients following ablation,5 assess arrhythmias,6 and allow for initial evaluation of myocardial infarction.7 Anecdotal evidence suggests that increasing numbers of veterinarians and clients are using this technology to monitor and evaluate cardiac arrhythmias in small and large animal patients. However, no studies have examined the diagnostic accuracy of this smartphone-based ECG device in veterinary patients. Because of the relatively small dipole created by the electrodes of the device, concerns exist about the resolution of the ECG tracing and, consequently, the ability to correctly identify arrhythmias or conduction disturbances.
Therefore, the objective of the study reported here was to evaluate the diagnostic utility of ECGs obtained with this smartphone-based device, compared with simultaneously recorded reference 6-lead ECGs, for identification of heart rate and rhythm in dogs and cats. We hypothesized that the smartphone-acquired ECGs would permit accurate identification of heart rate and rhythm in dogs and cats with normal sinus rhythm or spontaneous arrhythmias when compared with reference ECGs.
Materials and Methods
Animals
This was a prospective nonrandomized clinical method comparison study. Between April 2012 and January 2013, a cohort of consecutive canine and feline patients examined at the Cornell University Hospital for Animals cardiology service was enrolled. No specific inclusion or exclusion criteria were defined. Specific patient characteristics were not recorded. All owners provided informed consent. The study protocol was approved by the Cornell University College of Veterinary Medicine Institutional Animal Care and Use Committee.
ECG acquisition
A smartphone-based ECG device (a bipolar, single-lead recorder coupled to a smartphoneb with a dedicated smartphone applicationa) was used to acquire at least one 30-second ECG from each dog or cat included in the study. The recorder was attached to the smartphone with the positive (recording) electrode at the speaker end of the smartphone. An application on the smartphone transformed the electrical signal to an ultrasound signal that was received by the smartphone's microphone. The ECG was stored in a portable document format (PDF).
Smartphone-acquired ECGs consisted of bipolar, single-lead, high-resolution (16-bit) ECGs recorded at a sampling frequency of 300 Hz with a frequency response of 0.5 to 40 Hz and a dynamic input range of ± 10 mV. The ECG was displayed digitally at either 25 or 50 mm/s, with an amplitude of 10 or 20 mm/mV, after application of a 60-Hz hum filter.
At the same time smartphone-acquired ECGs were obtained, a reference 6-lead ECG8 recording was obtained by means of a standard digital ECG machine.c Reference ECGs were acquired at a sampling frequency of 1,000 Hz with a dynamic input range of ± 300 mV at a frequency response of 0.05 to 150 Hz. A digital notch filter (50 to 60 Hz) was applied when necessary to minimize artifact and baseline drift of the reference ECG.
Both the reference ECG and the smartphone-based ECG technology allowed users to select and alter the amplitudes and sweep speed of stored data files to optimize image resolution of the ECG tracings. All ECGs were subsequently printed out at paper speeds of 25 and 50 mm/s with a recording sensitivity of either 10 or 20 mm/mV, depending on the amplitude of the P-QRS-T complexes.
For ECG acquisition (smartphone-acquired and reference ECGs), patients were placed in right lateral recumbency while awake and resting quietly without sedation. Isopropyl alcohol and coupling gel were used to enhance conduction of the signal both for the smartphone-based ECG device and the reference ECG machine. The smartphone-acquired ECG device was positioned over the point of maximal intensity of the cardiac beat on the left side of the thorax, with the positive electrode placed caudally and the smartphone positioned at an approximately 45° angle (Figure 1). This position was similar to that for a standard lead III ECG. Gentle pressure was applied to optimize skin contact for the smartphone electrodes. The reference ECG was acquired simultaneously with 4 electrodes attached in the standard limb lead positions.
Representative photograph of a dog that illustrates positioning of a smartphone-based ECG devicea in a study comparing use of this device with a standard reference 6-lead ECG machine for identification of heart rate and rhythm in 51 dogs and 27 cats. The dog is positioned in right lateral recumbency.
Citation: Journal of the American Veterinary Medical Association 249, 2; 10.2460/javma.249.2.189
ECG analysis
For each patient, we examined both the smartphone-acquired ECG and reference ECG recordings to identify identical ECG complexes. Once these were identified, we obtained instantaneous heart rates from 5 QRS complexes on both recordings. When identical ECG complexes could not be accurately identified, a 15-second mean heart rate was calculated by use of artifacts or extrasystoles present on both tracings to synchronize the measured sections as closely as possible. Mean heart rate for all recordings was calculated and compared by 1 investigator (MSK).
Pairs of ECGs from each patient were printed out and labeled with a code. The ECGs were separated into smartphone-acquired ECG and reference ECG recordings. A deck of cards was numbered from 1 to 51 and shuffled. Electrocardiograms were numbered from 1 to 51, and the order of the numbers in the deck of cards was used to arrange the order of the ECGs. Copies of the ECGs were provided to 3 board-certified cardiologists for analysis. Each cardiologist examined both sets of ECGs and provided, first, a clinical rhythm diagnosis and, second, the polarity of the QRS depolarization for each ECG. Polarity of the QRS complex was compared between the smartphone-acquired ECG and lead III of the reference ECG. Polarity was defined as positive, negative, or biphasic (neutral).
Statistical analysis
Intra- and interobserver agreement was evaluated for both ECG rhythm diagnosis and QRS polarity identification. Intraobserver agreement was defined with a tripartite scoring system. For this, 2 points were assigned for complete agreement on an ECG (eg, ventricular tachycardia was identified on both recordings), 1 point was assigned for partial agreement (eg, supraventricular tachycardia was identified on the smartphone-acquired ECG, but atrial fibrillation was identified on the reference ECG), and 0 points were assigned for disagreement (eg, ventricular premature complexes were identified on the smartphone-acquired ECG, but atrial premature complexes were identified on the reference ECG). Similarly, interobserver agreement was examined for both the reference ECGs and smartphone-acquired ECGs independently, with a tripartite scoring system. Complete agreement indicated that all 3 investigators provided the same clinical diagnosis or same polarity for the QRS complexes, partial agreement indicated that 2 of 3 investigators agreed on the clinical diagnosis, and disagreement indicated that all 3 investigators provided different diagnoses or different polarities. Proportions of agreement for the 3 cardiologists between the 2 ECG modalities were compared with the Fisher exact test for > 2 proportions. A value of P < 0.05 was considered significant. Commercial statistical software was used.d
Results
We enrolled 51 dogs and 27 cats in the study. Instantaneous and mean heart rates were identical in all cases where exact matches could be made for comparison between the smartphone-acquired ECG and the reference ECG, and were within 1 beat of each other when mean heart rates were calculated. Mean heart rate recorded in the 51 dogs was 107 beats/min (range, 33 to 188 beats/min), and mean heart rate in the 27 cats was 181 beats/min (range, 107 to 250 beats/min). A representative example of 100% agreement in heart rate between the 2 ECG systems in sinus rhythm and ventricular ectopy could be illustrated (Figure 2).
Representative ECG tracings from a dog in sinus rhythm (A) and a cat in with ventricular ectopy (B). Tracings were obtained with a smartphone-based ECG devicea and a reference 6-lead ECG machine.c Notice that heart rate and rhythm are identical for both tracings in each panel. Paper speed = 25 mm/s; 1 cm = 1 mV. REF ECG = Reference ECG. SP ECG = Smartphone-acquired ECG.
Citation: Journal of the American Veterinary Medical Association 249, 2; 10.2460/javma.249.2.189
Intraobserver agreement for rhythm diagnosis (ie, how often each observer made the same diagnosis for a pair of ECG recordings) was high for both canine and feline ECGs (Tables 1 and 2). Similarly, the intraobserver agreement for QRS polarity in dogs was high. However, intraobserver agreement for QRS polarity in cats was substantially poorer.
Intraobserver agreement for heart rhythm and QRS complex polarity determined by 3 board-certified cardiologists evaluating paired smartphone-acquired and reference 6-lead ECGs from 51 dogs.
| Observer 1 | Observer 2 | Observer 3 | ||||
|---|---|---|---|---|---|---|
| Extent of agreement | Rhythm | Polarity | Rhythm | Polarity | Rhythm | Polarity |
| Complete agreement | 49 | 47 | 47 | 44 | 45 | 45 |
| Partial agreement | 1 | 2 | 2 | 4 | 4 | 2 |
| Disagreement | 1 | 2 | 2 | 3 | 2 | 4 |
Intraobserver agreement for heart rhythm and QRS complex polarity determined by 3 board-certified cardiologists evaluating paired smartphone-acquired and reference 6-lead ECGs from 27 cats.
| Observer 1 | Observer 2 | Observer 3 | ||||
|---|---|---|---|---|---|---|
| Extent of agreement | Rhythm | Polarity | Rhythm | Polarity | Rhythm | Polarity |
| Complete agreement | 25 | 18 | 23 | 15 | 24 | 19 |
| Partial agreement | 1 | 3 | 0 | 2 | 1 | 0 |
| Disagreement | 1 | 6 | 4 | 10 | 2 | 8 |
See Table 1 for key.
Despite a wide spectrum of arrhythmias recorded in both dogs and cats (Table 3), interobserver agreement for arrhythmia diagnosis was similar for the 2 ECG systems. The proportions of complete agreement, partial agreement, and disagreement between cardiologists were not significantly different for the 2 ECG systems for dogs (P = 0.18) or cats (P = 0.99; Table 4). Likewise, interobserver agreement for polarity was similar for the 2 systems. Proportions of complete agreement, partial agreement, and disagreement between cardiologists did not significantly differ for the 2 ECG systems for dogs (P = 0.62) or cats (P = 0.99; Table 5).
Heart rhythms recorded with a reference 6-lead ECG machine for the patients in Tables 1 and 2.
| Heart rhythm | Dogs (n = 51) | Cats (n = 27) |
|---|---|---|
| Sinus rhythm, sinus arrhythmia | 39 | 24 |
| Atrial fibrillation, atrial flutter | 7 | 1 |
| Ventricular tachyarrhythmia | 12 | 6 |
| Supraventricular tachyarrhythmia | 6 | 1 |
| Atrioventricular block | 7 | 0 |
| Paced ventricular rhythm | 1 | 0 |
| Electrical alternans | 12 | 0 |
| Wolff-Parkinson-White syndrome | 0 | 1 |
Nineteen dogs and 7 cats had > 1 rhythm identified.
Interobserver agreement for heart rhythm determined by 3 board-certified cardiologists evaluating smartphone-acquired and reference 6-lead ECGs from the patients in Tables 1 and 2.
| Dogs (n = 51) | Cats (n = 27) | |||
|---|---|---|---|---|
| Extent of agreement | Reference ECG | Smartphone ECG | Reference ECG | Smartphone ECG |
| Complete agreement | 46 | 41 | 25 | 24 |
| Partial agreement | 5 | 10 | 2 | 3 |
| Disagreement | 0 | 0 | 0 | 0 |
Complete agreement indicated that all 3 cardiologists provided the same clinical diagnosis, partial agreement indicated that 2 of 3 cardiologists agreed on the clinical diagnosis, and disagreement indicated that all 3 cardiologists provided different diagnoses. Proportions for complete agreement, partial agreement, and disagreement did not differ significantly between the 2 ECG systems for dogs (P = 0.18) or cats (P = 0.99).
See Table 1 for remainder of key.
Interobserver agreement for polarity of QRS complexes determined by 3 board-certified cardiologists evaluating smartphone-acquired and reference 6-lead ECGs from the patients in Tables 1 and 2.
| Dogs (n = 51) | Cats (n = 27) | |||
|---|---|---|---|---|
| Extent of agreement | Reference ECG | Smartphone ECG | Reference ECG | Smartphone ECG |
| Complete agreement | 50 | 49 | 20 | 20 |
| Partial agreement | 1 | 2 | 7 | 7 |
| Disagreement | 0 | 0 | 0 | 0 |
Complete agreement indicated that all 3 cardiologists provided the same polarity for the QRS complexes (positive, negative, or biphasic), partial agreement indicated that 2 of 3 cardiologists agreed on polarity, and disagreement indicated that all 3 cardiologists provided different polarities. Proportions of complete agreement, partial agreement, and disagreement did not differ significantly between the 2 ECG systems for dogs (P = 0.62) or cats (P = 0.99).
See Table 1 for remainder of key.
Discussion
Results of the present study of a group of 78 small animal patients (51 dogs and 27 cats) consecutively examined by a referral cardiology service found that ECG acquisition with the smartphone-based ECG devicea was feasible and accurately detected the heart rate in dogs and cats, when compared with heart rate determined from a simultaneously acquired standard 6-lead ECG.c Moreover, our findings also suggested that heart rhythms, varying from sinus arrhythmia to complex atrial and ventricular arrhythmias, could also be readily identified in dogs and cats with the smartphone-based ECG device. The accuracy in diagnosis of arrhythmias was demonstrated by high intraobserver as well as interobserver agreement between 3 cardiologists in the present study. This study was the first that confirmed the utility of this novel device for heart rate and rhythm assessment in dogs and cats. In human patients, equally high accuracy has been demonstrated, and the device is now being used extensively in a variety of clinical settings, including standard cardiology clinics6 and emergency medical services,7 as well as for self-monitoring for arrhythmias such as atrial fibrillation by patients in the home environment.6
In the present study, the finding that sometimes the 3 observers did not make the same rhythm diagnosis for a pair of ECG recordings, particularly in dogs, was not surprising. The ECGs acquired from dogs included a wider range of more complex arrhythmias than the ECGs acquired from cats, including supraventricular arrhythmias with pre-excitation, ventricular pacing, and rhythms with intraventricular conduction disturbances. We suggest that disagreement likely depended on the complexity of the arrhythmia as well as the ability of the ECG acquisition system to record the arrhythmia with high fidelity. Thus, for simple ventricular ectopy or sinus arrhythmias, for example, the disagreement would be anticipated to be relatively small. However, with wide, complex, variable tachyarrhythmias that may have components of rate dependency, the disagreement would be anticipated to be higher with both systems. Whereas the smartphone-based ECG device consistently produced ECG traces of adequate quality to make a correct rhythm diagnosis in this study, this modality is not intended to replace or supplant the reference ECG for detailed diagnostic evaluation or ongoing patient care. The reference standard for surface ECG recording is the 6-lead or 12-lead ECG, allowing inspection of electrical depolarization patterns in multiple simultaneous leads. The smartphone-based ECG device provides a single lead recording (rhythm strip) and is thus intended for relatively simple rhythm diagnosis because simultaneous comparison of multiple leads is not possible. Furthermore, the smartphone-based ECG device cannot provide information about mean electrical axis. By providing only 1 lead for examination, the device has limited ability to identify minor variations in QRS duration or amplitude or P wave morphology. For example, a supraventricular tachycardia with pre-excitation is more likely to be recognized on a 6-lead (or 12-lead) ECG than on a single-lead ECG. Additionally, identification and evaluation of P waves present a challenge for the smartphone-based ECG device, because this system records relatively small electrical activity from only a short-distance dipole, thus producing a small waveform, particularly in cats. Some of the intraobserver variability in the present study resulted from misidentification of P waves on smartphone-acquired ECGs. As such, for 1 dog with atrial standstill with VPCs, 1 observer made a diagnosis of sinus rhythm with VPCs when examining the smartphone-acquired ECG, because the absence of distinct P waves was attributed to a low voltage recording by the smartphone-based ECG device. Similarly, an instance of complete AV block in a dog with a ventricular endocardial-paced rhythm was identified by 2 of 3 observers as complete AV block with an accelerated ventricular escape rhythm on the smartphone-acquired ECG because the small pacemaker spikes were not appreciated on the smartphone-acquired ECG. Additionally, some instances of VPCs were misidentified as atrial premature beats or sinus beats on the single-lead smartphone-acquired ECGs, whereas on the reference ECGs, the 6 leads could be scrutinized for slightly altered QRS morphologies. In addition, many patients had multiple rhythms present in the same ECG (eg, sinus rhythm with second-degree AV block or sinus rhythm with VPCs or atrial premature contractions). This yielded slightly different diagnoses by the 3 observers in the present study, contributing to the interobserver variability. Finally, a definitive diagnosis of an arrhythmia is not free from subjectivity. For example, 1 cardiologist diagnosed a rhythm as atrial fibrillation with VPCs, whereas another diagnosed the same rhythm as atrial flutter with VPCs.
The present study was limited by the fact that we did not predetermine a list of possible diagnoses to be used by the 3 observers, which could have simplified the analysis and possibly reduced the intraobserver variability reported. However, we did not want any of the cardiologists to be restricted in their ability to describe the ECGs and thought this would be most clinically applicable. A different polarity of the QRS complexes for the smartphone-acquired ECG, compared with the reference 6-lead ECG, was found occasionally in dogs and more frequently in cats in this study, and all 3 observers identified these polarity variations equally well. It is possible that differences in QRS polarity may be a result of the dissimilar placement of the electrodes for the standard 6-lead ECGs versus the smartphone-acquired ECGs. For the smartphone-acquired ECGs, the 2 poles of the electrograph are in close proximity. Because of this, particularly in cats, we suggest that small changes in the position of the smartphone-based ECG device may result in large changes in QRS voltage and lead axis and, thus, changes in QRS polarity. Furthermore, in this study, we did not account for body condition score; a higher body condition score (eg, obese patients) could increase lead impedance, resulting in decreased signal amplitude for the smartphone-acquired ECGs. Similarly, patients were not evaluated for the presence of conditions such as pleural effusion or pericardial effusion, and these conditions could also increase lead impedance and decrease signal amplitude for the smartphone-acquired ECGs. Furthermore, anemia, sepsis, neoplasia, other diseases causing fever, and acid-base disturbances can all alter the ECG. However, these conditions should affect both ECG acquisition systems. Finally, movement artifact such as breathing or moving of the limbs during the recordings cannot be excluded because all animals were awake and without any sedation during ECG acquisition. However, movement artifacts would be expected to affect both ECG acquisition systems. In summary, results of the present study suggested that the smartphone-based ECG device is a novel tool for accurate assessment of heart rate and rhythm in dogs and cats. This new modality may allow veterinarians to identify and manage patients at risk for cardiac arrhythmias relatively inexpensively at the cage side and to share results via email with specialists if necessary.
Acknowledgments
Supported by Alivecor Inc.
ABBREVIATIONS
| AV | Atrioventricular |
| VPC | Ventricular premature contraction |
Footnotes
AliveCor, AliveCor Inc, San Francisco, Calif.
Apple iPhone 4 (iOS 6), Apple Inc, Cupertino, Calif.
Eickemeyer Vet PC-based ECG system, Medizintechnik für Tierärzte KG, Tuttlingen, Germany
Excel 2013, Microsoft Corp, Redmond, Wash.
References
1. Waller AD. A demonstration of man of electromotive changes accompanying the heart's beat. Ann Noninvasive Electrocardiol 2004; 9: 189–191.
2. Einthoven W. Un nouveau galvanometre. Arch Neerl Sci Exactes Nat 1901; 6: 625–633.
3. Saxon LA. Ubiquitous wireless ECG recording: a powerful tool physicians should embrace. J Cardiovasc Electrophysiol 2013; 24: 480–483.
4. Chung EH, Guise KD. QTC intervals can be assessed with the AliveCor heart monitor in patients on dofetilide for atrial fibrillation. J Electrocardiol 2015; 48: 8–9.
5. Tarakji KG, Wazni OM, Callahan T, et al. Using a novel wireless system in monitoring patients after the atrial fibrillation ablation procedure: the iTransmit study. Heart Rhythm 2015; 12: 554–559.
6. Haberman ZC, Jahn RT, Bose R, et al. Wireless smartphone ECG enables large scale screening in diverse populations. J Cardiovasc Electrophysiol 2015; 26: 520–526.
7. Muhlestein JB, Le V, Albert D, et al. Smartphone ECG for evaluation of STEMI: results of the ST LEUIS Pilot Study. J Electrocardiol 2015; 48: 249–259.
8. Tilley LP. Essentials of canine and feline electrocardiography. 3rd ed. Philadelphia: Lea & Febiger, 1995.
