Novel feature extraction method for signal analysis based on independent component analysis and wavelet transform

Feature extraction is an important part of data processing that provides a basis for more complicated tasks such as classification or clustering. Recently many approaches for signal feature extraction were created. However, plenty of proposed methods are based on convolutional neural networks. This class of models requires a high amount of computational power to train and deploy and large dataset. Our work introduces a novel feature extraction method that uses wavelet transform to provide additional information in the Independent Component Analysis mixing matrix. The goal of our work is to combine good performance with a low inference cost. We used the task of Electrocardiography (ECG) heartbeat classification to evaluate the usefulness of the proposed approach. Experiments were carried out with an MIT-BIH database with four target classes (Normal, Vestibular ectopic beats, Ventricular ectopic beats, and Fusion strikes). Several base wavelet functions with different classifiers were used in experiments. Best was selected with 5-fold cross-validation and Wilcoxon test with significance level 0.05. With the proposed method for feature extraction and multi-layer perceptron classifier, we obtained 95.81% BAC-score. Compared to other literature methods, our approach was better than most feature extraction methods except for convolutional neural networks. Further analysis indicates that our method performance is close to convolutional neural networks for classes with a limited number of learning examples. We also analyze the number of required operations at test time and argue that our method enables easy deployment in environments with limited computing power.

280 Role of the electrocardiogram in differentiating genotype positive dilated cardiomyopathy from cardiac remodelling in athletes

Aims Physiological cardiac remodelling in highly trained athletes may overlap with dilated cardiomyopathy (DCM). The aim of this study was to investigate the role of the ECG in differentiating between physiological and pathological remodelling. Methods and results The study population consisted of 30 patients with DCM who revealed a pathogenic variant at genetic testing and 30 elite athletes with significant cardiac remodelling defined by a left ventricular (LV) end-diastolic diameter > 62 mm and/or LV ejection fraction between 45% and 50%. The ECG was abnormal in 22 (73%) patients with DCM. The most common abnormalities were low voltages (n = 14, 47%), lateral TWI (n = 6, 20%), ventricular ectopic beats (n = 5, 17%) and anterior TWI (n = 4, 13). Two athletes revealed an abnormal ECG: complete left bundle branch block (LBBB) in one case and atrial flutter in the other. The sensitivity, specificity and accuracy of the ECG in differentiating DCM from physiological adaptation to exercise in athletes was 73% [confidence interval (CI: 54–88%), 93% (CI: 78–99%), and 0.83 (CI: 0.71–0.92), respectively. Conclusions While the ECG is usually normal in athletes exhibiting significant LV dilatation and/or systolic dysfunction, this test is often abnormal in patients with DCM harbouring a pathogenic variant. Low voltages in the limb leads and lateral TWI are the most common abnormalities.

P095 Heart Rate Variability during Sleep in Paroxysmal Atrial Fibrillation patients with and without Obstructive Sleep Apnoea

Introduction Physiological studies have demonstrated the importance of the autonomic nervous system in mediating acute apnoea-induced atrial fibrillation (AF). We aimed to compare Heart Rate Variability (HRV) markers of autonomic function in paroxysmal atrial fibrillation (PAF) patients with and without obstructive sleep apnoea (OSA). A secondary aim was the analysis of ectopic beats in these groups. Methods Nocturnal ECG traces from 89 PAF patients who underwent in-laboratory polysomnography were included. After identifying ectopic beats in the ECGs, periods of arrhythmia as well as sleep apnoea events were excluded. HRV time and frequency domains were reported by sleep stage (REM vs Non-REM) for patients with and without OSA. Results Frequency domain analysis of HRV during non-REM sleep in PAF patients with OSA showed increased cardiac parasympathetic modulation (HF-nu: 39.13 ± 15.74 vs 47.98 ± 14.60, p = 0.008*) and reduced cardiac sympathetic modulation (LF/HF ratio: 2.05 ± 2.02 vs 1.17 ± 0.98, p = 0.007*). Results remained significant after adjusting for age, sex and BMI (adjusted p values 0.024 and 0.018 respectively). PAF patients with severe OSA (AHI ≥ 30/hr) had more AF beats and Ventricular Ectopic Beats than those without severe OSA (22.7 ± 42.8% vs 3.7 ± 17.9%, p = 0.006*, 1.7 ± 3.8 vs 0.3 ± 0.9%, p = 0.004* respectively). Conclusions This is the first study of HRV in AF patients with and without OSA. It suggests a chronic increase in parasympathetic nervous modulation and relative reduction in sympathetic modulation in PAF patients with OSA.

Mitral valve prolapse with ventricular arrhythmias: does it carries a worse prognosis?

The association of mitral valve prolapse (MVP) with ventricular arrhythmias has long been known and has generally been considered a benign condition. In recent years, however, a small but not negligible risk of malignant ventricular arrhythmias and sudden cardiac death has been documented in the large population of subjects with MVP. The main predictors of major arrhythmic risk identified so far include history of syncope, ventricular repolarization abnormalities in the inferior-lateral electrocardiogram leads, right bundle branch block morphology of ventricular ectopic beats, finding of areas of myocardial fibrosis on cardiac magnetic resonance, and mitral annular disjunction (MAD) on echocardiogram, as well as a possible pro-arrhythmic genetic substrate. The stratification of arrhythmic risk, with the active search for red flags and in particular of MAD, is important to identify patients with the malignant arrhythmic variant of MVP in whom to implement closer surveillance and possible therapeutic interventions.

Mutations change excitability and the probability of re-entry in a computational model of cardiac myocytes in the sleeve of the pulmonary vein

Atrial fibrillation (AF) is a common health problem with substantial individual and societal costs. The origin of AF has been debated for more than a century, and the precise, biophysical mechanisms that are responsible for the initiation and maintenance of the chaotic electrochemical waves that define AF, remains unclear. It is well accepted that the outlet of the pulmonary veins is the primary anatomical site of AF initiation, and that electrical isolation of these regions remains the most effective treatment for AF. Furthermore, it is well known that certain ion channel or transporter mutations can significantly increase the likelihood of AF. Here, we present a computational model capable of characterizing functionally important features of the microanatomical and electrophysiological substrate that represents the transition from the pulmonary veins (PV) to the left atrium (LA) of the human heart. This model is based on a finite element representation of every myocyte in a segment of this (PV/LA) region. Thus, it allows for investigation a mix of typical PV and LA myocytes. We use the model to investigate the likelihood of ectopic beats and re-entrant waves in a cylindrical geometry representing the transition from PV to LA. In particular, we investigate and illustrate how six different AF- associated mutations can alter the probability of ectopic beats and re-entry in this region.

Accuracy of Physicians Interpreting Photoplethysmography and Electrocardiography Tracings to Detect Atrial Fibrillation: INTERPRET-AF

Aims: This study aims to compare the performance of physicians to detect atrial fibrillation (AF) based on photoplethysmography (PPG), single-lead ECG and 12-lead ECG, and to explore the incremental value of PPG presentation as a tachogram and Poincaré plot, and of algorithm classification for interpretation by physicians.Methods and Results: Email invitations to participate in an online survey were distributed among physicians to analyse almost simultaneously recorded PPG, single-lead ECG and 12-lead ECG traces from 30 patients (10 in sinus rhythm (SR), 10 in SR with ectopic beats and 10 in AF). The task was to classify the readings as ‘SR', ‘ectopic/missed beats', ‘AF', ‘flutter' or ‘unreadable'. Sixty-five physicians detected or excluded AF based on the raw PPG waveforms with 88.8% sensitivity and 86.3% specificity. Additional presentation of the tachogram plus Poincaré plot significantly increased sensitivity and specificity to 95.5% (P < 0.001) and 92.5% (P < 0.001), respectively. The algorithm information did not further increase the accuracy to detect AF (sensitivity 97.5%, P = 0.556; specificity 95.0%, P = 0.182). Physicians detected AF on single-lead ECG tracings with 91.2% sensitivity and 93.9% specificity. Diagnostic accuracy was also not optimal on full 12-lead ECGs (93.9 and 98.6%, respectively). Notably, there was no significant difference between the performance of PPG waveform plus tachogram and Poincaré, compared to a single-lead ECG to detect or exclude AF (sensitivity P = 0.672; specificity P = 0.536).Conclusion: Physicians can detect AF on a PPG output with equivalent accuracy compared to single-lead ECG, if the PPG waveforms are presented together with a tachogram and Poincaré plot and the quality of the recordings is high.

Successful Catheter Ablation of the “R on T” Ventricular Fibrillation

In patients with idiopathic ventricular fibrillation (VF), recurrent implantable cardioverter-defibrillator (ICD) shocks might increase mortality risk and reduce patients’ quality of life. Catheter ablation of triggering ectopic beats is considered to be an effective method. We present a patient with recurrent VF, caused by the “R on T” premature ventricular complexes. In the presented case radiofrequency catheter ablation efficiently eliminated arrhythmia trigger, which was possible to detect thanks to the intracardiac electrocardiograms (ECG’s) stored in the ICD.

Interpretation of Electrocardiogram Heartbeat by CNN and GRU

The diagnosis of electrocardiogram (ECG) is extremely onerous and inefficient, so it is necessary to use a computer-aided diagnosis of ECG signals. However, it is still a challenging problem to design high-accuracy ECG algorithms suitable for the medical field. In this paper, a classification method is proposed to classify ECG signals. Firstly, wavelet transform is used to denoise the original data, and data enhancement technology is used to overcome the problem of an unbalanced dataset. Secondly, an integrated convolutional neural network (CNN) and gated recurrent unit (GRU) classifier is proposed. The proposed network consists of a convolution layer, followed by 6 local feature extraction modules (LFEM), a GRU, and a Dense layer and a Softmax layer. Finally, the processed data were input into the CNN-GRU network into five categories: nonectopic beats, supraventricular ectopic beats, ventricular ectopic beats, fusion beats, and unknown beats. The MIT-BIH arrhythmia database was used to evaluate the approach, and the average sensitivity, accuracy, and F1-score of the network for 5 types of ECG were 99.33%, 99.61%, and 99.42%. The evaluation criteria of the proposed method are superior to other state-of-the-art methods, and this model can be applied to wearable devices to achieve high-precision monitoring of ECG.

Interplay between temporal and spatial dispersion of repolarization in the initiation and perpetuation of Torsades de Pointes in the chronic AV-block dog

Ventricular arrhythmias, consisting of single ectopic beats (sEB), multiple EB (mEB), and Torsades de Pointes (TdP, defined as >5 beats with QRS vector twisting around isoelectric line) can be induced in the anesthetized chronic AV-block (CAVB) dog by dofetilide (IKr-blocker). The interplay between temporal dispersion of repolarization, quantified as short-term variability (STV), and spatial dispersion of repolarization (SDR) in the initiation and perpetuation of these arrhythmias remains unclear. Five inducible (>3 TdPs/10') CAVB dogs were observed for 10' from the start of dofetilide infusion (0.025mg/kg, 5'). An intracardiac decapolar electrogram (EGM) catheter and 30 intramural cardiac needles in the left ventricle (LV) were introduced. STVARI was derived from 31 consecutive activation recovery intervals (ARI) on the intracardiac EGM, using the formula: . The mean SDR3D in the LV was determined as the three-dimensional repolarization time differences between the intramural cardiac needles. Moments of measurement included baseline (BL) and after dofetilide infusion prior to first 1) sEB (occurrence at 100±35"), 2) mEB (224±96"), and 3) non self-terminating TdP (454±298"). STVARI increased from 2.15±0.32ms at BL to 3.73±0.99ms* prior to the first sEB and remained increased without further significant progression to mEB (4.41±0.45ms*) and TdP (5.07±0.84ms*) (*p<0.05 compared to BL). SDR3D did not change from 31±11ms at BL to 43±13ms prior to sEB, but increased significantly prior to mEB (68±7ms*) and to TdP (86±9ms*+) (+p<0.05 compared to sEB). An increase in STV contributes to the initiation of sEB whereas an increase in SDR is important for the perpetuation of non self-terminating TdPs.