INTEGRATING PORTABLE ECG DEVICES WITH CLOUD SYSTEMS FOR CARDIAC ARRHYTHMIA DETECTION

Aim: The majority of machine learning (ML) models in healthcare are built on retrospective data, much of which is collected without explicit patient consent for use in artificial intelligence (AI) and ML applications. The primary aim of this study was to evaluate whether clinicians and scientific researchers themselves consent to provide their own data for the training of ML models. Materials and Methods: This survey was conducted through an anonymous online survey, utilizing platforms such as Telegram, LinkedIn, and Viber. The target audience comprised specific international groups, primarily Russian, German, and English-speaking, of clinicians and scientific researchers. These participants ranged in their levels of expertise and experience, from beginners to veterans. The survey centered on a singular, pivotal question: “Do You Consent to the Use of Your Biological and Private Data for Training Machine Learning and AI Models?” Respondents had the option to choose from three responses: “Yes” and “No”. Results: The survey was conducted in January 2024. A total of 119 unique and verified individuals participated in the survey. The results revealed that only 50% of respondents (63 persons) expressed consent to provide their own data for the training of ML and AI models. Conclusion: In the development of ML and AI models, particularly open-source ones, it is crucial to ascertain whether participants are willing to provide their private data. While ML algorithms can transform the nature of data, it is important to remember that the primary owner of this data is the individual. Our findings show that in 50% of the cases, even participants from scientific research and clinical backgrounds – individuals typically accountable for ensuring data quality in AI and ML model development – do not consent to the use of their data in AI and ML settings. This highlights the need for more stringent consent processes and ethical considerations in the utilization of personal data in AI and ML research.

The purpose of this study is to obtain the normal ranges of electrocardiograms of Japanese elementary school children and to compare with each group separated by age and sex. Materials and Methods The subjects were 984 healthy children (506 boys and 478 girls) selected at random from all children (3532 boys and girls) in an elementary school in Nagoya, Japan. They were accepted as healthy children on the basis of their daily unrestricted activity and through a series of medical examinations. The subjects were sub-divided into twelve groups by sex and age for further analysis of their electrocardiograms. The 12 conventional leads and Lead V3R were recorded for all subjects. The amplitudes and intervals of electrocardiographic deflections were carefully measured in each lead. In this paper 95 percentile of frequency distribution was accepted as a normal range for each component of the record. Results and Discussion RR interval was longer with increasing age both in boys and girls. The increasing rate of RR interval was more prominent in the boys than in the girls, and the difference of mean RR intervals reached a statistical significant level in the groups of 10 and 11 years old. According to above data, it was concluded that the normal range of RR interval should be set up separately; 0.48 to 1.00sec. for 6-9 years, and 0.54 to 1.08 sec. for 10-11 years. Although PQ interval also increased along with age, no significant difference was found for boys and girls. The normal ranges of PQ interval were 0.12 to 0.18 sec. for 6-9 years and 0.13 to 0.19sec. for 10-11 years respectively. The increasing rate of QT interval was less remarkable than RR or PQ intervals, and the mean of boys was significantly longer than that of girls in elder age group. A single normal range could be accepted for QT intervals as 0.28 to 0.40sec. for the all age groups. Though PQ interval tended to be longer with increasing RR interval except in the group with tachycardia, no significant correlation was observed for PQ and RR intervals. Correlation coefficients of RR with QT intervals were very high and proved statistically significant both in boys and girls and for all age groups.

The aim of this study was to evaluate the effect of ventilation on electrocardiographic time intervals as a function of the light-dark (LD) cycle in an in vivo rat model. RR, PQ, QT and QTc intervals were measured in female Wistar rats anaesthetized with both ketamine and xylazine (100 mg/15 mg/kg, i.m., open chest experiments) after adaptation to the LD cycle (12:12h) for 4 weeks. Electrocardiograms (ECG) were recorded before surgical interventions; after tracheotomy, and thoracotomy, and 5 minutes of stabilization with artificial ventilation; 30, 60, 90 and 120 seconds after the onset of apnoea; and after 5, 10, 15, and 20 minutes of artificial reoxygenation. Time intervals in intact animals showed significant LD differences, except in the QT interval. The initial significant (p<0,001) LD differences in PQ interval and loss of dependence on LD cycle in the QT interval were preserved during short-term apnoea-induced asphyxia (30–60 sec) In contrast, long-term asphyxia (90–120 sec) eliminated LD dependence in the PQ interval, but significant LD differences were shown in the QT interval. Apnoea completely abolished LD differences in the RR interval. Reoxygenation restored the PQ and QT intervals to the pre-asphyxic LD differences, but with the RR intervals, the LD differences were eliminated. We have concluded that myocardial vulnerability is dependent on the LD cycle and on changes of pulmonary ventilation.

The aim of this study was to evaluate the effect of ventilation on electrocardiographic time intervals as a function of the light-dark (LD) cycle in an in vivo rat model. RR, PQ, QT and QTc intervals were measured in female Wistar rats anaesthetized with both ketamine and xylazine (100 mg/15 mg/kg, i.m., open chest experiments) after adaptation to the LD cycle (12:12h) for 4 weeks. Electrocardiograms (ECG) were recorded before surgical interventions; after tracheotomy, and thoracotomy, and 5 minutes of stabilization with artificial ventilation; 30, 60, 90 and 120 seconds after the onset of apnoea; and after 5, 10, 15, and 20 minutes of artificial reoxygenation. Time intervals in intact animals showed significant LD differences, except in the QT interval. The initial significant (p<0,001) LD differences in PQ interval and loss of dependence on LD cycle in the QT interval were preserved during short-term apnoea-induced asphyxia (30–60 sec) In contrast, long-term asphyxia (90–120 sec) eliminated LD dependence in the PQ interval, but significant LD differences were shown in the QT interval. Apnoea completely abolished LD differences in the RR interval. Reoxygenation restored the PQ and QT intervals to the pre-asphyxic LD differences, but with the RR intervals, the LD differences were eliminated. We have concluded that myocardial vulnerability is dependent on the LD cycle and on changes of pulmonary ventilation.

A QRS detection method based on the combined wavelet entropy (CWS) is proposed for two-channel ECG signals. The raw ECG signals from two channels are firstly transformed by the continuous wavelet transform within a selected frequency interval, which can effectively suppress the interference and only focus on the QRS component. The wavelet entropy (WS) based method is employed to analyze the ECG signal from channel 1 based on an automated threshold determination. This method may fail when the current RR interval is larger than 1.5 times the average RR interval. Here, another WS has to be calculated for ECG signal from channel 2, and the CWS is defined as the sum of WS from two channel signals. The time evolution of the CWS is calculated, which can enhance detection results by using the QRS information coming from both channels. Experimental results indicate that this proposed method achieves an average detection rate of 99.84%, a sensitivity of 99.91%, and a positive prediction of 99.93% over both two leads of ECG signals from the MIT-BIH arrhythmia database, which outperform some published results of other detection algorithms.

The objective of this study is to explore new imaging techniques with the use of the deep learning method for the identification of cardiac abnormalities present in ECG recordings with 2, 3, 4, 6 and 12-lead in the framework of the PhysioNet Challenge 2021. The training set is a public database of 88253 twelve-lead ECG recordings lasting from 6 seconds to 60 seconds. Each ECG recording has one or more diagnostic labels. The six-lead, four-lead, three-lead, and two-lead are reduced-lead version of the original twelve-lead data. The deep learning method considers images that are built from raw ECG signals. This technique considers innovative 3-D images of the entire ECG signal, observing the regional constraints of the leads, obtaining time-spatial images, where the x-axis is the temporal evolution of ECG signal, the y axis is the spatial location of the leads, and the z axis (color) the amplitude. These images are used for training Convolutional Neural Networks with GoogleNet for ECG diagnostic classification. The official results of the classification accuracy of our team named 'Gio_new_img’ received score of 0.4, 0.4, 0.39, 0.4 and 0.4 (ranked 18th, 18th; 18^th,18^th, 18^th out of 39 teams) for the 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead versions of the hidden test set with the Challenge evaluation metric.

In this paper two knowledge-based methods for arrhythmia detection and classification using ECG recordings are described, which utilize different information of the ECG signal. The first uses features of the ECG signal (R wave, QRS duration, P wave, RR interval, PR interval, PP interval, QRS similarity and P wave similarity), which are fed into a decision-tree like knowledge-based system. The system can classify all types of arrhythmias. The second is based on the utilization of the RR-duration signal only. Initially, rules based on medical knowledge are used for arrhythmic beat classification and the results are fed into a deterministic automato for arrhythmic episode detection and classification. The system can be used for the classification of limited types of arrhythmia due to the fact that only limited information is carried by the RR-duration signal.

Background Wearable devices capable of single-lead electrocardiograms (ECGs) are becoming increasingly prevalent. Exploring their potential for screening conditions beyond atrial fibrillation presents a compelling research opportunity. Objective This study evaluates the agreement between the Apple Watch Series 6 (APW) and simultaneous 12-lead ECG (12L) recordings in measuring PR, QRS, RR, QT, and QTc intervals in a cohort of National Health Service (NHS) cardiology patients. Methods We analysed data from the WEAR-TECH study, which included 400 simultaneous recordings from the APW and 12L ECGs. We included patients with ECGs in sinus rhythm and excluded patients with poor-quality ECGs. No exclusions were made based on underlying cardiac conditions. Measurements of the best three consecutive waveforms in leads I, II, and V5 were performed using EP Calipers (EP Studio) by a single observer. Leads I and II were compared for PR, QRS, and RR intervals, while leads I, II, and V5 were compared for QT and QTc intervals. QTc was calculated using Bazett’s formula. Differences between APW and 12L ECGs were assessed using linear mixed effects regression models with random intercepts and type II Wald chi-square tests. Clinical agreement was measured by averaging the waveforms from 12L leads and the APW and calculating the difference between the means. Results Out of 324 paired ECGs from 162 patients, 8 were excluded due to missing or uninterpretable data. The median age was 64 (IQR: 54 - 74) years, with 117 (75%) being male. PR, QRS, and RR intervals from the APW agreed best with 12L lead II, and QT and QTc intervals from APW agreed best with 12L lead V5, with no significant difference found in any of these comparisons. PR and RR intervals significantly differed between APW and 12L lead I, with the APW tending to overestimate relative to the 12L. Additionally, significant differences were found in 12L leads I and II for QT and QTc, with APW also tending to overestimate these intervals (see Table 1). Clinical agreement between the APW and all 12L leads was a minimum of 21% for a <10ms difference, 45% for a <20ms difference, and 71% for a <40ms difference (see Figure 1). Conclusions The Apple Watch Series 6 demonstrates promising potential for initial outpatient monitoring of PR, QRS, RR, QT, and QTc intervals with reasonable agreement to 12L ECG. This is especially true in lead II for PR, QRS, and RR intervals and lead V5 for QT and QTc intervals. Further reductions in variance between APW and 12L ECG measurements could enhance clinical decision-making accuracy.Table 1:The p-values of APW vs 12L leadFigure 1:Agreement of APW to 12L leads

We compared simultaneous fetal echo AV intervals (E-AV) and magentocardiogram (fMCG) PR intervals (M-PR) in SSA+ and control pregnancies to determine if an AV interval of + 2 z-scores correlated with a prolonged M-PR interval and 1° AV block. We searched the database of the University of Wisconsin Biomagnetism laboratory for pregnant women referred for fMCG with SSA antibodies. Maternal and pregnancy data were obtained from the mother's medical record. Fetal MCGs were performed using a 37-channel biomagnetometer (Magnes, 4D Neuroimaging, Inc). During the same session, we measured and averaged PR and cardiac RR intervals from 5 rhythm tracings and compared results to Echo Doppler AV interval (mitral inflow/aortic outflow) and RR intervals. Results were normalised for RR interval and compared by paired t-test. Regression analysis was used to test associations between gestational age (GA) and cardiac intervals. The mean M-PR was 23.3 ms less than the mean E-AV (95% CI: 20.1,26.5, p<0.001); this statistically significant difference persisted when normalised for RR interval. There were no significant differences between SSA and controls for RR, PR, or AV intervals. Neither E-RR interval or GA were associated with predicting E-AV; nor were M-RR or GA associated with predicting M-PR. However, the interaction between E-AV interval and GA were significantly associated with the AV-PR difference of 23.3 ms. Our results suggests that E-AV overestimates M-PR in both SSA and controls. Using the previous definition of 1° AVB (an AV interval of 150 ms) may result in unnecessary treatment of SSA+ normal fetuses. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

To evaluate the usability of electrocardiography (ECG) intervals in the diagnosis and treatment monitoring of acute carbon monoxide (CO) poisoning. An observational study. Department of Emergency Medicine, Kayseri City Hospital, Turkey, from November 2018 to May 2019. Each of 80 patients for study and control groups were prospectively included. For study group, pre- and post-treatment ECG intervals (P-wave and QRS complex periods, PP, PR, RR, QT, and QTc intervals) and carboxyhemoglobin (COHb) levels of the patients were evaluated. For control group, ECG intervals and COHb levels of the patients during admission to the Emergency Department were evaluated. As a result of the statistical analysis in which measurements of the study group and control group were compared, a statistically significant difference was found in the following values: pre-treatment group COHb level (p<0.001), PR interval (p=0.046), PP interval (p<0.001), QT interval (p<0.001), QTC interval (p=0.016), RR interval (p<0.001), and post-treatment group COHb level (p<0.001), PR interval (p=0.009), PP interval (p=0.041), QTC interval (p=0.010), and RR interval (p=0.036). QT interval values in the post-treatment group were similar to those of the control group (p=0.342). In the ROC analysis where the diagnostic performance of ECG intervals was evaluated, the area under the curve (AUC) scale was between 0.29 and 0.62. ECG intervals do not provide as much benefit as COHb measurement in the diagnosis of acute CO poisoning. However, the QT interval is a useful ECG interval in the treatment monitoring of acute CO poisoning. Key Words: Acute carbonmonoxide poisoning, ECG interval, QT interval, Emergency Department.

RR and PR intervals were extracted from single channel electrocardiograms in normal human subjects using a wavelet based technique. RR and PR intervals were shown to be correlated with respiration due to common autonomic nervous activation of the sino-atrial and atrio-ventricular nodes. However, evidence for independent autonomic nervous activation of these nodes is also shown based on RR and PR interval spectra, and on RR vs. PR plots.

The aims of this study were (i) to evaluate the effect of talazoparib (1 mg once daily) on cardiac repolarization in patients with advanced solid tumors by assessing corrected QT interval (QTc) and (ii) to examine the relationship between plasma talazoparib concentration and QTc. In this open-label phase 1 study, patients had continuous 12-lead ECG recordings at baseline followed by time-matched continuous ECG recordings and collection of talazoparib plasma pharmacokinetic samples predose and at 1, 2, 4, and 6 h postdose on treatment days 1 and 22 and before talazoparib administration on day 2. ECG recordings were submitted for independent central review where triplicate 10-s ECGs, extracted up to 15 min before pharmacokinetic samples, were assessed for RR, PR, QRS, and QT intervals and ECG morphology. QT interval was corrected for heart rate using Fridericia's (QTcF) and Bazett's (QTcB) formulae. Linear mixed-effects modeling was used to examine the relationship between QTc and RR interval change from baseline and plasma talazoparib concentration. Thirty-seven patients received talazoparib. Mean change in QTcF from time-matched baseline ranged from -3.5 to 6.9 ms, with the greatest change 1 h postdose on day 22. No clinically relevant changes in PR, QRS, QTcB, QTcF, or RR intervals, heart rate, or ECG morphology were observed. No concentration-dependent effect on heart rate or QTc was observed. No deaths, permanent treatment discontinuations due to adverse events were reported. Talazoparib (1 mg once daily) had no clinically relevant effects on cardiac repolarization.

Objective The Objective of the study is to compare the ECG recordings using a novel mobile ECG recorder (EPI Mini) with a clinically validated mobile phone with ECG recording function (EPI Life). Methods The EPI Mini and EPI Life are ECG recorders that use the same type of ECG chip modules, same surface electrodes and are able to record ECG following direct contact of the inbuilt contact electrodes with skin surfaces. The ECG recordings from EPI Life have been clinically validated with standard 12 lead ECG recordings. While the EPI Life is a mobile phone which allows the recording of ECGs and its transmission to a 24 hour centre, the EPI Mini is a mobile device that is a scaled down version of the EPI Life which records the ECG and transmits it wirelessly through smartphones. 30 consecutive healthy individuals aged 18 to 48 years consented to this prospective study. ECG recording using the EPI Mini device was performed, followed immediately by the recording of ECG using the EPI Life device in 3 different lead positions. ECG recordings were performed using modified limb leads I and II (mL I and mL II) and modified V5 precordial lead (mV5), and the optimal ECG recordings were compared. ECG parameters that were compared include ECG morphology (pattern and orientation of p wave, QRS complex, ST segment and T wave), QRS amplitude (R wave, S wave and R + S waves) and ECG parameters (PR interval and QT interval). The data were statistically assessed using linear regression analysis, as the Objective was to assess the correlation between the two modalities of ECG recording. Analysis was made between two variables, with X representing data from EPI Life ECG recordings and Y representing data from EPI Mini ECG recordings. From the data, the best fit value for the slope and P values were calculated. Results A total of 180 ECG recordings were obtained for comparison; 90 ECG recordings using EPI Life and 90 recordings using EPI Mini. The linear regression graphs show that there was good correlation, as demonstrated by the linearity of the units, X and Y plotted on the graph for QRS amplitude (R wave, S wave and R + S waves) and ECG parameters (PR interval and QT interval) for all 3 leads. The R wave, S wave, R + S waves, PR interval and QT interval had best fit value for the slope of 1.0623, 1.392, 1.120, 0.855 and 1.141 respectively for mL I. The R wave, S wave, R + S waves, PR interval and QT interval had best fit value for the slope of 1.021, 1.031, 1.025, 0.991 and 1.011 respectively for mL II. The R wave, S wave, R + S waves, PR interval and QT interval had best fit value for the slope of 0.951, 1.148, 1.034, 0.856 and 1.073 respectively for mV5. The P values were found to be less than 0.0001 for all the quantitative parameters that were analysed. There was 100% correlation of the ECG morphology for ECG recordings from both devices for all the subjects when comparing the mL I, mL II and mV5 leads. Conclusions There was statistically significant correlation between the quantitative ECG parameters recorded using the novel EPI Mini and the clinically validated EPI Life. In addition, the data showed that there was good qualitative correlation when comparing between the morphology of the ECGs recorded using both devices.

A complex balance between extrinsic neural and intrinsic mechanisms is responsible for regulating atrioventricular (AV) conduction. We hypothesized that atrial excitation interval is shortened during dynamic exercise by extrinsic cardiac autonomic activity and that if AV conduction time responds inversely to fluctuation in atrial rhythm, ventricular excitation interval will be maintained at the predetermined cardiac cycle length. To examine such inverse relationship between PP interval and the subsequent change in PR interval (DeltaPR), we analyzed the beat-to-beat changes in PP, PR, and RR intervals during stair-stepping exercise for 10 min in 11 sedentary and 9 trained subjects. In the sedentary group, the average PR interval significantly shortened during exercise, in parallel with the reduction in the average PP and RR intervals. The variance of PP and RR intervals was also significantly decreased during exercise. The reduction in the variance of RR interval was, however, much greater than that of PP interval, implying that AV conduction time changes inversely to fluctuation in atrial excitation rhythm. Indeed, the variance of PR interval was augmented during exercise and there was a clear inverse relationship between PP and DeltaPR intervals. Although trained subjects were characterized by their lower heart rate response during dynamic exercise, the responses in the variability of PP, PR, and RR intervals were fundamentally identical with those in sedentary subjects. We conclude that the AV nodal mechanism that operates at a higher level of heart rate during dynamic exercise may cancel fluctuation in atrial excitation interval and keep ventricular excitation rhythm at the predetermined cardiac cycle length.

This paper analyzes the optimization features of machine learning (ML) model training procedures using multi-GPU systems to enhance cyber security in telecommunication networks. A key aspect of the study is the use of data parallelism, which allows the distribution of the training load across multiple GPUs, significantly reducing training time and improving model accuracy-critical factors for rapid threat detection in cyberspace. A novel approach for optimizing data batch size using Mutual Information (MI) is proposed, which harmonizes the utilization of computational resources with the information content of the training data. MI helps to determine the optimal data batch size that minimizes training errors and improves model accuracy without a significant increase in training time. Experimental results demonstrate the substantial advantages of multi-GPU configurations compared to single-GPU setups, providing faster training and improved model accuracy. It was particularly emphasized that MI-guided batch size tuning significantly outperforms traditional manual tuning methods, ensuring higher validation accuracy and reducing training time. The study showed that the MI-based approach is an effective tool for addressing the problem of optimizing ML model training processes in real-world scenarios where cyber security is critical. The proposed methods allow ML models to train faster and more accurately identify potential threats, making them particularly relevant for telecommunication networks where a rapid response to new threats in real time is required. The implementation of modern computational technologies such as multi-GPU systems and MI-optimized training enhances the efficiency and accuracy of machine learning models. This, in turn, improves cyber security measures and ensures a more reliable defence of telecommunication networks against malicious attacks. It is noted that the proposed approaches can be adapted not only for cyber security but also for other areas where high model accuracy and fast training are important. Future research prospects include the development of new machine learning methods, particularly deep neural networks, the exploration of alternative computational architectures such as quantum computing or distributed systems, and their integration into real-time systems. Special attention should be paid to the ethical aspects of implementing automated cyber security systems, particularly in preventing bias in algorithms and ensuring fairness in their application.