Abstract
The identification of arrhythmogenic right ventricular dysplasia (ARVD) from 12-channel standard electrocardiogram (ECG) is challenging. High density ECG data may identify lead locations and criteria with a higher sensitivity.
Eighty-channel ECG recording from patients diagnosed with ARVD and controls were quantified by magnitude and integral measures of QRS and T waves and by a measure (the average silhouette width) of differences in the shapes of the normalized ECG cycles. The channels with the best separability between ARVD patients and controls were near the right ventricular wall, at the third intercostal space. These channels showed pronounced differences in P waves compared to controls as well as the expected differences in QRS and T waves.
Multichannel recordings, as in body surface mapping, add little to the reliability of diagnosing ARVD from ECGs. However, repositioning ECG electrodes to a high anterior position can improve the identification of ECG variations in ARVD. Additionally, increased P wave amplitude appears to be associated with ARVD.
Eighty-channel electrocardiogram (ECG) recordings from arrhythmogenic right ventricular dysplasia (ARVD) patient and control groups allow the selection of recording sites that are best for separating the ECG waveforms between the two groups, using methods based on the ECG waveform rather than measures of amplitudes and durations of its components.
These methods show that high anterior chest lead positions provide the most discrimination between ARVD patients and controls and demonstrate well-known differences in T wave polarity and QRS integral.
There is also an association between increased P wave amplitude and the presence of ARVD.
The additional channels of the 80-lead ECGs add little to the discriminability between ARVD patients and controls.
Introduction
The 2010 Task force criteria for the diagnosis of arrhythmogenic right ventricular dysplasia (ARVD) include multiple 12 lead electrocardiographic (ECG) criteria.1 These contribute both major [T wave inversion (TWI)] in right sided precordial, V1–V3, or beyond in the absence of right bundle branch block (RBBB), epsilon wave and minor (TWI in V1 and V2 in the absence of RBBB or in V4–V6, TWI in V1–V4 in the presence of RBBB, terminal activation of the QRS 55 ms in the absence of complete RBBB) diagnostic criteria. These ECG criteria mostly relate to the spread of repolarization. Changes in repolarization can appear earlier in the disease progression and provide a specific finding.2 Analysis of the 12 lead ECG for dispersion of depolarization (QRS) and depolarization–repolarization (QT) intervals across the precordial leads can lend added information in risk stratification of individuals already diagnosed with ARVD.3
The analysis of the ECG for ARVD diagnosis is principally based upon standard 12 lead ECG data. However, it is not known whether ECG diagnosis may be improved with high density body surface ECG mapping or by leads in non-standard locations. The 12 lead ECG has inherent limitations. The density of ECG mapping of the right ventricle is limited, and the ability to compare electrical activity between the left and the right precordium is restricted. High density body surface mapping with an 80 lead ECG has previously been shown to improve the sensitivity of diagnosis of acute myocardial infarction,4 particularly with improvements in posterior and inferior wall infarction recognition; anatomical areas accepted to be relatively sparsely represented with conventional 12 lead ECG mapping. Published data regarding high density ECG mapping in ARVD is limited, with focus on the differentiation between ARVD and normal heart right ventricular outflow tract tachycardia (RVOT) with regard to T-wave characteristics only.5
The present study was therefore planned to determine whether a commercially available 80 lead surface ECG would lead to a characterization of depolarization and repolarization across the chest leads in patients with ARVD which could then identify ECG lead locations and criteria with higher sensitivity.
Methods
Recruitment
Nine unrelated ARVD patients, who had given written informed consent to participate in this electrophysiological study, were recruited from the West Yorkshire Arrhythmia Service Outpatient Clinic during a 6 month period in 2013. An equal number of controls of similar age and gender as well as body size were recruited from hospital and university staff. All control subjects had a normal echocardiogram. The study protocol received approval from the NHS Health Research Authority NRES Committee Yorkshire & The Humber.
Diagnosis of arrhythmogenic right ventricular dysplasia
The patients had been diagnosed with ARVD on the basis of the 2010 modified criteria which update the 1994 guidelines.1,6
Evaluation and clinical testing
The patients’ medical history was obtained from review of their records and by interview, and is summarized in Tables 1and2.
Baseline characteristics
| Controls (n = 10) | Patients (n = 9) | P | |
|---|---|---|---|
| Age (years) | 50 ± 15 | 54 ± 14 | 0.563 |
| Male (%) | 80 | 80 | 1.00 |
| Height (m) | 1.73 ± 0.09 | 1.79 ± 0.10 | 0.163 |
| Weight (kg) | 79 ± 15 | 86 ± 14 | 0.295 |
| BMI (kg/m2) | 26 ± 3 | 27 ± 4 | 0.753 |
| Heart rate (b/min) | 70 ± 8 | 58 ± 10 | 0.009 |
| PR (ms) | 161 ± 14 | 181 ± 67 | 0.376 |
| QRS duration (ms) | 91 ± 11 | 98 ± 19 | 0.369 |
| QT interval | 415 ± 19 | 429 ± 27 | 0.258 |
| Controls (n = 10) | Patients (n = 9) | P | |
|---|---|---|---|
| Age (years) | 50 ± 15 | 54 ± 14 | 0.563 |
| Male (%) | 80 | 80 | 1.00 |
| Height (m) | 1.73 ± 0.09 | 1.79 ± 0.10 | 0.163 |
| Weight (kg) | 79 ± 15 | 86 ± 14 | 0.295 |
| BMI (kg/m2) | 26 ± 3 | 27 ± 4 | 0.753 |
| Heart rate (b/min) | 70 ± 8 | 58 ± 10 | 0.009 |
| PR (ms) | 161 ± 14 | 181 ± 67 | 0.376 |
| QRS duration (ms) | 91 ± 11 | 98 ± 19 | 0.369 |
| QT interval | 415 ± 19 | 429 ± 27 | 0.258 |
BMI, body mass index.
Baseline characteristics
| Controls (n = 10) | Patients (n = 9) | P | |
|---|---|---|---|
| Age (years) | 50 ± 15 | 54 ± 14 | 0.563 |
| Male (%) | 80 | 80 | 1.00 |
| Height (m) | 1.73 ± 0.09 | 1.79 ± 0.10 | 0.163 |
| Weight (kg) | 79 ± 15 | 86 ± 14 | 0.295 |
| BMI (kg/m2) | 26 ± 3 | 27 ± 4 | 0.753 |
| Heart rate (b/min) | 70 ± 8 | 58 ± 10 | 0.009 |
| PR (ms) | 161 ± 14 | 181 ± 67 | 0.376 |
| QRS duration (ms) | 91 ± 11 | 98 ± 19 | 0.369 |
| QT interval | 415 ± 19 | 429 ± 27 | 0.258 |
| Controls (n = 10) | Patients (n = 9) | P | |
|---|---|---|---|
| Age (years) | 50 ± 15 | 54 ± 14 | 0.563 |
| Male (%) | 80 | 80 | 1.00 |
| Height (m) | 1.73 ± 0.09 | 1.79 ± 0.10 | 0.163 |
| Weight (kg) | 79 ± 15 | 86 ± 14 | 0.295 |
| BMI (kg/m2) | 26 ± 3 | 27 ± 4 | 0.753 |
| Heart rate (b/min) | 70 ± 8 | 58 ± 10 | 0.009 |
| PR (ms) | 161 ± 14 | 181 ± 67 | 0.376 |
| QRS duration (ms) | 91 ± 11 | 98 ± 19 | 0.369 |
| QT interval | 415 ± 19 | 429 ± 27 | 0.258 |
BMI, body mass index.
Clinical characteristics of ARVD patients
| Global or regional dysfunction | Repolarization | Depolarization/ | Arrhythmias | Family history | Exact details | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| and structural alterations | abnormalities | conduction abnormalities | ||||||||||
| Study number | MRI | RV angio | ||||||||||
| Major | Minor | Major | Major | Minor | Major | Minor | Major | Minor | Major | Minor | ||
| ACP001 | 1 | 1 | 1 | Repol: (2012)—TWI in V1 + V2, Depol: (2006)—epsilon wave in V1–V3 on ECG RV angio: (2006)—RV dysplasia | ||||||||
| ACP002 | 1 | 1 | 1 | MRI: (2007)—Akinesis, EF < 40%.Repol abn: (2007)—T wave inv V1 + V2, Arrhyth:(1997) >500 VEs on 24 h tape | ||||||||
| ACP003 | 1 | 1 | 1 | MRI: (2002)—RV dyskinesia + RVEF < 40%.Depolar: (2009)—epsilon wave in V1–V3. Arrhyth: (2002)—VT with LBBB and inf axis | ||||||||
| ACP004 | 1 | 1 | 1 | 1 | MRI: (2011)—report states ‘meets minor criteria for taskforce’.Depol (2011)—epsilon wave in V1–V3. Arrhyth: (2011)—VT with LBB and inf axis, Repol: (2011)—TWI V1 + V2 | |||||||
| ACP005 | 1 | 1 | MRI (2007)—abnormal fatty deposits, RWMA, Depol (2007)—epsilon wave in V1–V3 | |||||||||
| ACP006 | 1 | 1 | RV angio (2000)—‘dilated, abnormally contracting RV'. Positive gene test for ARVD. | |||||||||
| ACP007 | 1 | 1 | 1 | 1 | MRI: (2002)—RV dysplasia + EF < 40% Arrhyth:(2002) Sustained VT with LBBB with superior axis, Repol: (2003) - TWI in V1 + V2. Family Hx (2002)—sudden death of father (no ARVD Dx) | |||||||
| ACP008 | 1 | 1 | 2 | MRI: (2000)—RV regional dyskinesia, fatty deposits, RVEDV 194.9 mL, RVESV 109.6 mL, RV stoke vol 83.2 mL, RV EF < 40%. Repol abnormalities:(2000) T wave inv V1 + V2 (no RBBB), Arrhyth: (1997 + 1999) > 500 VEs on 24 h tape, (1997) VT on EP study—LBBB with inf axis | ||||||||
| ACP009 | 1 | 1 | 1 | Arrhyth: sustained VT with LBBB with superior axis, Repol abn: TWI V1 + V2(2009), MRI: (2008) RV akinesia + RVEDV/BSA = 100 mL/m2 | ||||||||
| Global or regional dysfunction | Repolarization | Depolarization/ | Arrhythmias | Family history | Exact details | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| and structural alterations | abnormalities | conduction abnormalities | ||||||||||
| Study number | MRI | RV angio | ||||||||||
| Major | Minor | Major | Major | Minor | Major | Minor | Major | Minor | Major | Minor | ||
| ACP001 | 1 | 1 | 1 | Repol: (2012)—TWI in V1 + V2, Depol: (2006)—epsilon wave in V1–V3 on ECG RV angio: (2006)—RV dysplasia | ||||||||
| ACP002 | 1 | 1 | 1 | MRI: (2007)—Akinesis, EF < 40%.Repol abn: (2007)—T wave inv V1 + V2, Arrhyth:(1997) >500 VEs on 24 h tape | ||||||||
| ACP003 | 1 | 1 | 1 | MRI: (2002)—RV dyskinesia + RVEF < 40%.Depolar: (2009)—epsilon wave in V1–V3. Arrhyth: (2002)—VT with LBBB and inf axis | ||||||||
| ACP004 | 1 | 1 | 1 | 1 | MRI: (2011)—report states ‘meets minor criteria for taskforce’.Depol (2011)—epsilon wave in V1–V3. Arrhyth: (2011)—VT with LBB and inf axis, Repol: (2011)—TWI V1 + V2 | |||||||
| ACP005 | 1 | 1 | MRI (2007)—abnormal fatty deposits, RWMA, Depol (2007)—epsilon wave in V1–V3 | |||||||||
| ACP006 | 1 | 1 | RV angio (2000)—‘dilated, abnormally contracting RV'. Positive gene test for ARVD. | |||||||||
| ACP007 | 1 | 1 | 1 | 1 | MRI: (2002)—RV dysplasia + EF < 40% Arrhyth:(2002) Sustained VT with LBBB with superior axis, Repol: (2003) - TWI in V1 + V2. Family Hx (2002)—sudden death of father (no ARVD Dx) | |||||||
| ACP008 | 1 | 1 | 2 | MRI: (2000)—RV regional dyskinesia, fatty deposits, RVEDV 194.9 mL, RVESV 109.6 mL, RV stoke vol 83.2 mL, RV EF < 40%. Repol abnormalities:(2000) T wave inv V1 + V2 (no RBBB), Arrhyth: (1997 + 1999) > 500 VEs on 24 h tape, (1997) VT on EP study—LBBB with inf axis | ||||||||
| ACP009 | 1 | 1 | 1 | Arrhyth: sustained VT with LBBB with superior axis, Repol abn: TWI V1 + V2(2009), MRI: (2008) RV akinesia + RVEDV/BSA = 100 mL/m2 | ||||||||
Repol, repolarization; Depol, depolarization; MRI, magnetic resonance imaging; ARVD, arrhythmogenic right ventricular dysplasia; RBBB, right bundle branch block; ECG, electrocardiogram; TWI, T wave inversion; RV, right ventricle; RVEF, right ventricular ejection fraction; VT, ventricular tachycardia; LBBB, left bundle branch block; RWMA, regional wall motion abnormality; RVEDV, right ventricular end-diastolic volume; BSA, body surface area..
Clinical characteristics of ARVD patients
| Global or regional dysfunction | Repolarization | Depolarization/ | Arrhythmias | Family history | Exact details | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| and structural alterations | abnormalities | conduction abnormalities | ||||||||||
| Study number | MRI | RV angio | ||||||||||
| Major | Minor | Major | Major | Minor | Major | Minor | Major | Minor | Major | Minor | ||
| ACP001 | 1 | 1 | 1 | Repol: (2012)—TWI in V1 + V2, Depol: (2006)—epsilon wave in V1–V3 on ECG RV angio: (2006)—RV dysplasia | ||||||||
| ACP002 | 1 | 1 | 1 | MRI: (2007)—Akinesis, EF < 40%.Repol abn: (2007)—T wave inv V1 + V2, Arrhyth:(1997) >500 VEs on 24 h tape | ||||||||
| ACP003 | 1 | 1 | 1 | MRI: (2002)—RV dyskinesia + RVEF < 40%.Depolar: (2009)—epsilon wave in V1–V3. Arrhyth: (2002)—VT with LBBB and inf axis | ||||||||
| ACP004 | 1 | 1 | 1 | 1 | MRI: (2011)—report states ‘meets minor criteria for taskforce’.Depol (2011)—epsilon wave in V1–V3. Arrhyth: (2011)—VT with LBB and inf axis, Repol: (2011)—TWI V1 + V2 | |||||||
| ACP005 | 1 | 1 | MRI (2007)—abnormal fatty deposits, RWMA, Depol (2007)—epsilon wave in V1–V3 | |||||||||
| ACP006 | 1 | 1 | RV angio (2000)—‘dilated, abnormally contracting RV'. Positive gene test for ARVD. | |||||||||
| ACP007 | 1 | 1 | 1 | 1 | MRI: (2002)—RV dysplasia + EF < 40% Arrhyth:(2002) Sustained VT with LBBB with superior axis, Repol: (2003) - TWI in V1 + V2. Family Hx (2002)—sudden death of father (no ARVD Dx) | |||||||
| ACP008 | 1 | 1 | 2 | MRI: (2000)—RV regional dyskinesia, fatty deposits, RVEDV 194.9 mL, RVESV 109.6 mL, RV stoke vol 83.2 mL, RV EF < 40%. Repol abnormalities:(2000) T wave inv V1 + V2 (no RBBB), Arrhyth: (1997 + 1999) > 500 VEs on 24 h tape, (1997) VT on EP study—LBBB with inf axis | ||||||||
| ACP009 | 1 | 1 | 1 | Arrhyth: sustained VT with LBBB with superior axis, Repol abn: TWI V1 + V2(2009), MRI: (2008) RV akinesia + RVEDV/BSA = 100 mL/m2 | ||||||||
| Global or regional dysfunction | Repolarization | Depolarization/ | Arrhythmias | Family history | Exact details | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| and structural alterations | abnormalities | conduction abnormalities | ||||||||||
| Study number | MRI | RV angio | ||||||||||
| Major | Minor | Major | Major | Minor | Major | Minor | Major | Minor | Major | Minor | ||
| ACP001 | 1 | 1 | 1 | Repol: (2012)—TWI in V1 + V2, Depol: (2006)—epsilon wave in V1–V3 on ECG RV angio: (2006)—RV dysplasia | ||||||||
| ACP002 | 1 | 1 | 1 | MRI: (2007)—Akinesis, EF < 40%.Repol abn: (2007)—T wave inv V1 + V2, Arrhyth:(1997) >500 VEs on 24 h tape | ||||||||
| ACP003 | 1 | 1 | 1 | MRI: (2002)—RV dyskinesia + RVEF < 40%.Depolar: (2009)—epsilon wave in V1–V3. Arrhyth: (2002)—VT with LBBB and inf axis | ||||||||
| ACP004 | 1 | 1 | 1 | 1 | MRI: (2011)—report states ‘meets minor criteria for taskforce’.Depol (2011)—epsilon wave in V1–V3. Arrhyth: (2011)—VT with LBB and inf axis, Repol: (2011)—TWI V1 + V2 | |||||||
| ACP005 | 1 | 1 | MRI (2007)—abnormal fatty deposits, RWMA, Depol (2007)—epsilon wave in V1–V3 | |||||||||
| ACP006 | 1 | 1 | RV angio (2000)—‘dilated, abnormally contracting RV'. Positive gene test for ARVD. | |||||||||
| ACP007 | 1 | 1 | 1 | 1 | MRI: (2002)—RV dysplasia + EF < 40% Arrhyth:(2002) Sustained VT with LBBB with superior axis, Repol: (2003) - TWI in V1 + V2. Family Hx (2002)—sudden death of father (no ARVD Dx) | |||||||
| ACP008 | 1 | 1 | 2 | MRI: (2000)—RV regional dyskinesia, fatty deposits, RVEDV 194.9 mL, RVESV 109.6 mL, RV stoke vol 83.2 mL, RV EF < 40%. Repol abnormalities:(2000) T wave inv V1 + V2 (no RBBB), Arrhyth: (1997 + 1999) > 500 VEs on 24 h tape, (1997) VT on EP study—LBBB with inf axis | ||||||||
| ACP009 | 1 | 1 | 1 | Arrhyth: sustained VT with LBBB with superior axis, Repol abn: TWI V1 + V2(2009), MRI: (2008) RV akinesia + RVEDV/BSA = 100 mL/m2 | ||||||||
Repol, repolarization; Depol, depolarization; MRI, magnetic resonance imaging; ARVD, arrhythmogenic right ventricular dysplasia; RBBB, right bundle branch block; ECG, electrocardiogram; TWI, T wave inversion; RV, right ventricle; RVEF, right ventricular ejection fraction; VT, ventricular tachycardia; LBBB, left bundle branch block; RWMA, regional wall motion abnormality; RVEDV, right ventricular end-diastolic volume; BSA, body surface area..
Body surface ECGs, with the subject resting and supine, were recorded by an 80-lead system from Verathon (Bothell, Washington) Heartscape® with 58 anterior, 12 lateral, and 10 posterior leads—see Figure 1A. Electrode contact was automatically checked when the electrode vest was applied, and heart rate, QRS width, QT, QTc, and QRS axis/T axis, sample ECGs for each channel, and a standard ‘12 lead ECG’ reconstructed from the 80 channels printed after a 10 s recording. A 10 s 80 channel ECG recording at a sampling rate of 1 ms was stored. Channels with poor electrode contact identified by excessive noise on the lead were excluded from the analysis
Data analysis protocol. (A) 80 channels, 10 s, recorded at 1 kHz. (B) Each channel is zero-meaned and filtered, and the (C) amplitudes and integrals of the QRS (blue) and T wave (red) components obtained algorithmically for each channel; (D) each channel is denoised, downsampled and normalized, (E) for each channel the dynamic time warping (DTW) distance to all the other same channels in both patients and controls, for the average cycle, is estimated. (F) The separability of each channel (i.e. how well it can separate ARVD patient and control groups) is measured and mapped by the ASW. Further details can be found in the Supplementary material online.
Independent data analysis
The stored data were exported, converted from a proprietary .hs format to a .csv format by a MatLab program, for plotting and processing. For each recording, the activity of each channel was quantified by the following measures: magnitude (peak-to-peak), and sum (positive peak value + negative peak), and the integrals of the depolarization (QRS) and repolarization (T) wave components.
In order to quantify the similarity of signals in patients and controls for each channel the duration of each cycle of the ECG was normalized to eliminate the effects of heart rate. A dynamic time warping distance measurement was used to quantify the dis-similarity between any two signals. Technical details are given in the Appendix A. The compactness (see Supplementary material online, Material S1) of the patient and control clusters were measured, and the separation (see Supplementary material online, Material S2) between them, to determine which channels i.e. electrode positions gave the clearer separation between ARVD patients and controls. The separability was quantified by the average silhouette width (ASW) as defined in see Supplementary material online, Material S3.
The data analysis protocol is illustrated schematically in Figure 1.
Results
There were no differences in baseline characteristics between ARVD patients and controls (Table 1) apart from a lower heart rate in ARVD patients due to beta-blocker use. All patients had clear task force criteria for ARVD (Table 2).
Scatter plots for all 80 channels of integrals magnitudes and signed amplitudes for QRS and T wave showed no clear visual separation between ARVD and control data apart from the polarity of the T waves in the anterior chest leads. Figure 2A is an example of this illustrated for the signed T wave amplitude and QRS integral for leads placed in a similar anatomical location as leads V1–V6 on a standard 12 lead ECG. Figure 2B shows the same for leads that were placed to cover the anterior aspect of the right ventricle.
Overlap of ECG measures between patients and controls. Signed amplitude of T wave and integral of QRS for all ARVD patients (red) and controls (green). (A) shows electrode positions corresponding to V1–V6 on a standard 12 lead ECG and (B) shows electrodes positioned over an area covering the anterior aspect of the right ventricle.
Average silhouette width is a measure of the separability between ARVD and controls for an individual channel, and indicates the diagnostic usefulness of the channel. The ASW was calculated for every ECG channel 1–77, and is shown in Figure 3. Their value was colour coded and their location indicated by superposition over an image of the torso. The channels with the largest ASW i.e. the most compact and the most clearly separable clusters for the ECG time series of ARVD patients and normal subjects was channel 21, which was 3 cms above V1 (3rd intercostal space to the right of the sternum) . Most of the channels with greater ASWs were located close to and around channel 21 (30, 31, 32, 33, and 34) i.e. near the anterior right ventricular wall. Channel positions with less separable clusters of time series were also topographically close to each other.
Electrode positions on (A) anterior and (B) posterior torso. Colour map (hot metal scale) codes electrodes which discriminate between ARVD and control recordings. White represents best separation, dark red is least useful in discriminating between ARVD and controls. The numbers refer to the recording channel, and the colour codes the ASW calculated from equation A.3 (see Supplementary material online).
In order to visualize the characteristics of the waveform that were different between patients with ARVD and controls, the 10 s recording was normalized for amplitude (Figure 4). The channels with high ASW (with greater separability measure values) demonstrated clear differences between ARVD patients and controls. In the ARVD patients, the T wave polarity was more often negative and the S wave of the QRS was more delayed and pronounced as expected. Interestingly, however, the visualized P wave was larger in amplitude. This can be quantified by the half integral of the P wave (time from peak to end of wave), which for the channels A, B, and C of Figure 4 was 0.019 ± 0.015 for patients, 0.007 ± 0.005, P = 0.0007, mean ± standard deviation.
Ten seconds of superimposed normalized cycles of ECGs for single recording channels for ARVD patients (red) and controls (green) centred at the peak of the R wave. The amplitude for each channel of the 10 s was zero-meaned, denoised and normalized by its standard deviation. A, B, and C are channels 21, 31, and 32, the three channels with the highest value of ASW (ASW > 0.3). D, E, and F are channels 14, 29, and 26 with ASW < 0.15.
To ensure that the evaluation of channels which were based upon the ASW measure were plausible as a diagnostic index, subjects were classified using multiple channels.7 We anticipated that recordings from channels with greater ASWs would be more effective in the classification of subjects. Using a simulated annealing algorithm, a set of channels, as well as weights for them, were tuned for the 1NN classifier, only allowing channels with greater ASW values to be selected. 1NN algorithm based using the channels C = (20, 21, 23, 31, 41) and with weights W = (0.700, 0.630, 0.015, 0.091, 0.523) achieved an accuracy of 81.3%.
In order to provide statistical evidence that the performance of this algorithm was not a coincidence, the accuracies of 1000 1NN classifiers based on sets of randomly chosen channels with randomly assigned weights were measured, using leave-one-out procedure. The nearest neighbour algorithm using the ECG channels that had been selected on the basis of their ASWs provided the greatest accuracy, confirming the suitability of the ASW measure to select channels for the ECG classification
Discussion
Arrhythmogenic right ventricular dysplasi is an inherited condition in which the myocardium gradually undergoes fibrofatty replacement over time.8 As its name suggests, it was initially felt to be a disease of the right ventricle but there is now evidence of left ventricular and bi-ventricular forms.9–11 Those affected are often asymptomatic and the first manifestation can be with a sudden, unexplained death. Although standard 12-lead ECGs, signal-averaged ECGs and cardiac magnetic resonance imaging (MRI) may show abnormalities, these changes generally occur once the disease is well established and so diagnosis of the condition remains a challenge. This is the first report in which multi-electrode surface ECG mapping was used to determine whether the addition of multiple leads increases the sensitivity of diagnosis.
The multichannel ECG leads are analysed in terms of their complete waveform shapes, rather than simple measures of amplitudes, intervals, and durations. A dynamic time-warping distance measure is used to quantify the differences between the ECG waveform shapes between the ARVD patients and controls, and the separability measured to determine which recording positions give the clearest separation between the ARVD patients and control groups.
Our results show that the high anterior chest leads provide the most discrimination between ARVD patients and controls with differences in T wave polarity and QRS integral. It is not surprising that the channels adjacent to the anterior right ventricle are more discriminatory than the other ECG leads since the condition generally manifests with right ventricular pathology. This lead position is also in a similar location to the previously described Fontaine lead position.11,12
A more unexpected finding is the enhanced P wave amplitude in ARVD patients in these electrode positions. The morphology of the P wave is more consistent with right atrial pathology due to the increased P wave amplitude and suggests either a right atrial cardiomyopathic process or right atrial remodelling. This would be in keeping with previous studies which have demonstrated an increased prevalence of atrial arrhythmias and structural variations in patients with ARVD.11,13–17 However, further evaluation of this finding is required to determine its use in the diagnosis of ARVD.
The majority of the channels did not provide additional discriminatory information and so high-density ECG recording does not appear to be necessary. Instead, simply repositioning standard ECG electrodes to a higher anterior position may be all that is required to support the clinical diagnosis of ARVD.
Limitations
This is a small single centre study, however, ARVD is a rare condition and our results would need to be tested in a larger multicentre population. Further, we chose patients with a definitive diagnosis, and testing these criteria in patients with phenotypically less severe disease would be useful. Our Controls were identified on the basis of a normal ECG and echocardiogram. Ideally, controls would also have a cardiac MRI to exclude quiescent forms of ARVD but this pilot study was not resourced to enable this.
Conclusion
In patients with confirmed ARVD, the high anterior chest leads appear to be better at discriminating the changes in QRS integral and T wave polarity between patients and controls. Furthermore, there appears to be an association between increased P wave amplitude and the presence of ARVD. These findings therefore suggest that the addition of high anterior leads and P wave amplitude as part of the diagnostic work up of ARVD may improve the diagnostic yield. The additional channels that can be measured with 80-lead ECGs are not useful and clinically it may be more practical and economical to place electrodes in these areas.
Supplementary material
Supplementary material is available at Europace online.
Acknowledgements
We would like to thank Fuxing Yang, Jong Tae Yuk, Matthew Caprio, Keith Coomes and Julie Lyon from Verathon for the support they provided during the study, their loan of equipment and assistance with data decoding. We would also like to thank Christopher Whitaker for his assistance with data processing.
Conflict of interest: J.O. has received a fellowship from St. Jude Medical. A.H. has received travel grants from St. Jude Medical. M.T. has received research/travel grants from St. Jude Medical, Biosense Webster, Medtronic and Boehringer Ingelheim.
Funding
Verathon provided equipment and assisted in the initial processing of raw data.
