Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662903
Petra Novotna, Tomáš Vičar, Jakub Hejc, M. Ronzhina
Detection of cardiac arrhythmias is still an ongoing challenge. Here we focus on premature ventricular contraction (PVC) and premature atrial contraction (PAC) and introduce a deep-learning-based method for PVC/PAC localization in ECG. Our method is based on involving the time series with non-zero values corresponding to the ground truth PVC/PAC positions into the training process. To improve the efficiency of deep model training, the transition between the non-zero and zero areas in the train output time series was smoothed by introducing a Gaussian function. When applied to the new ECGs, the output signal (time series including Gaussians) is processed by a robust peak detector with Bayesian optimization of threshold, minimal distance and peak prominence. Positions of the detected peaks correspond to the desired PVC/PAC positions. The proposed method was evaluated on China Physiological Signal Challenge 2018 (CPSC2018) using own-created ground truth positions of PVC/PAC. The proposed method reached F1 score 0.923 and 0.688 for PAC and PVC, respectively, which is better than our previous results obtained via multiple instance learning-based method.
{"title":"Deep-Learning Premature Contraction Localization Using Gaussian Based Predicted Data","authors":"Petra Novotna, Tomáš Vičar, Jakub Hejc, M. Ronzhina","doi":"10.23919/cinc53138.2021.9662903","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662903","url":null,"abstract":"Detection of cardiac arrhythmias is still an ongoing challenge. Here we focus on premature ventricular contraction (PVC) and premature atrial contraction (PAC) and introduce a deep-learning-based method for PVC/PAC localization in ECG. Our method is based on involving the time series with non-zero values corresponding to the ground truth PVC/PAC positions into the training process. To improve the efficiency of deep model training, the transition between the non-zero and zero areas in the train output time series was smoothed by introducing a Gaussian function. When applied to the new ECGs, the output signal (time series including Gaussians) is processed by a robust peak detector with Bayesian optimization of threshold, minimal distance and peak prominence. Positions of the detected peaks correspond to the desired PVC/PAC positions. The proposed method was evaluated on China Physiological Signal Challenge 2018 (CPSC2018) using own-created ground truth positions of PVC/PAC. The proposed method reached F1 score 0.923 and 0.688 for PAC and PVC, respectively, which is better than our previous results obtained via multiple instance learning-based method.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126868599","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662872
H. Jessen, R. V. D. Leur, P. Doevendans, R. V. Es
Portable ECG devices with a reduced number of leads are increasingly being used in clinical practice. As part of the PhysioNet/Computing in Cardiology Challenge 2021, this study aims to develop an algorithm for automated diagnosis of reduced-lead ECGs. We compared separate baseline classifiers for the different lead-subsets with our newly proposed shared classifier. The different models were pre-trained on a physician-annotated dataset of 269,72612-lead ECGs. Fine-tuning was done on the challenge dataset, consisting of 88,243 ECGs. Even though different models showed promising results on the internal pre-training dataset, optimal scores were achieved by the baseline model on the hidden test set. Our team, UMCU, received scores of 0.47, 0.40, 0.41, 0.41, and 0.41 (ranked 14th, 17th, 17th, 17th, and 16th out of 39 teams) for the 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead versions of the hidden test set.
减少导联数量的便携式心电设备越来越多地用于临床实践。作为PhysioNet/Computing in Cardiology Challenge 2021的一部分，该研究旨在开发一种自动诊断低导联心电图的算法。我们将不同铅子集的单独基线分类器与我们新提出的共享分类器进行了比较。不同的模型在医生注释的269,72612导联心电图数据集上进行预训练。对挑战数据集进行了微调，该数据集由88,243个心电图组成。尽管不同的模型在内部预训练数据集上显示出很好的结果，但基线模型在隐藏测试集上获得了最优分数。我们的UMCU团队在12-lead, 6-lead, 4-lead, 3-lead和2-lead版本的隐藏测试集中获得了0.47,0.40,0.41,0.41和0.41的分数(在39个团队中排名第14,17,17,17和16)。
{"title":"Automated Diagnosis of Reduced-Lead Electrocardiograms Using a Shared Classifier","authors":"H. Jessen, R. V. D. Leur, P. Doevendans, R. V. Es","doi":"10.23919/cinc53138.2021.9662872","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662872","url":null,"abstract":"Portable ECG devices with a reduced number of leads are increasingly being used in clinical practice. As part of the PhysioNet/Computing in Cardiology Challenge 2021, this study aims to develop an algorithm for automated diagnosis of reduced-lead ECGs. We compared separate baseline classifiers for the different lead-subsets with our newly proposed shared classifier. The different models were pre-trained on a physician-annotated dataset of 269,72612-lead ECGs. Fine-tuning was done on the challenge dataset, consisting of 88,243 ECGs. Even though different models showed promising results on the internal pre-training dataset, optimal scores were achieved by the baseline model on the hidden test set. Our team, UMCU, received scores of 0.47, 0.40, 0.41, 0.41, and 0.41 (ranked 14th, 17th, 17th, 17th, and 16th out of 39 teams) for the 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead versions of the hidden test set.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127895556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662847
Monika Butkuvienė, A. Petrėnas, Andrius Sološenko, A. Martín-Yebra, V. Marozas, L. Sörnmo
Existing studies offer little insight on how atrial fibrillation (AF) detection performance is influenced by the properties of AF episode patterns. The aim of this study is to investigate the influence of AF burden and median AF episode length on detection performance. For this purpose, three types of AF detectors, using either information on rhythm, rhythm and morphology, or ECG segments, were investigated on 1-h simulated ECGs. Comparing AF burdens of 20% and 80% for a median episode length of 167 beats, the sensitivity of the rhythm- and morphology-based detector increases only slightly whereas the specificity drops from 99.5% to 93.3%. The corresponding figures of specificity are 99.0% and 90.6% for the rhythm-based detector; 88.1% and 70.7% for the segment-based detector. The influence of AF burden on specificity becomes even more pronounced for AF patterns with brief episodes (median episode length set to 30 beats). Therefore, patterns with briefepisodes and high AF burden imply higher demands on detection performance. Future research should focus on how well episode patterns are captured.
{"title":"Atrial Fibrillation Episode Patterns and Their Influence on Detection Performance","authors":"Monika Butkuvienė, A. Petrėnas, Andrius Sološenko, A. Martín-Yebra, V. Marozas, L. Sörnmo","doi":"10.23919/cinc53138.2021.9662847","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662847","url":null,"abstract":"Existing studies offer little insight on how atrial fibrillation (AF) detection performance is influenced by the properties of AF episode patterns. The aim of this study is to investigate the influence of AF burden and median AF episode length on detection performance. For this purpose, three types of AF detectors, using either information on rhythm, rhythm and morphology, or ECG segments, were investigated on 1-h simulated ECGs. Comparing AF burdens of 20% and 80% for a median episode length of 167 beats, the sensitivity of the rhythm- and morphology-based detector increases only slightly whereas the specificity drops from 99.5% to 93.3%. The corresponding figures of specificity are 99.0% and 90.6% for the rhythm-based detector; 88.1% and 70.7% for the segment-based detector. The influence of AF burden on specificity becomes even more pronounced for AF patterns with brief episodes (median episode length set to 30 beats). Therefore, patterns with briefepisodes and high AF burden imply higher demands on detection performance. Future research should focus on how well episode patterns are captured.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"348 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132651393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662906
B. D. Maria, F. Perego, G. Cassetti, V. Bari, B. Cairo, F. Gelpi, Monica Parati, L. Vecchia, A. Porta
Among the analytical methods estimating the complexity of the heart period (HP), the linear model-based multiscale complexity (MSC) approach allows the estimation of the complexity over time scales linked to the cardiac autonomic control, i.e. in the low frequency (LF, 0.04-0.15 Hz) and high frequency $(HF, 0.15-0.4 Hz)$ bands. In this study we exploited MSC to evaluate the differences in the HP variability complexity during daytime (DAY) and nighttime (NIGHT) in 23 healthy females (WOMEN, age $36pm 6yrs)$ ) and 21 males (MEN, age $35pm 5yrs)$ performing a 24-hour Holter electrocardiogram. Parametric power spectral analysis was applied as well for comparison. Complexity indexes were computed regardless of the temporal scale (CI) and in the LF and HF bands ( $CI_{LF}$ and $CI_{HF}$, respectively). We found that the power spectral indexes did not differentiate WOMEN and MEN, while CI and $CI_{LF}$ were higher in WOMEN during DAY. The higher HP complexity in females could be explained by a lower sympathetic drive and more complex hormonal regulation than males. We conclude that MSC was more powerful than power spectral analysis in detecting gender differences in HP variability. In addition, as cardiac control differs between females and males, preventive and therapeutic interventions should take gender differences into account.
{"title":"Gender Differences in Short-Term Multiscale Complexity of the Heart Rate Variability","authors":"B. D. Maria, F. Perego, G. Cassetti, V. Bari, B. Cairo, F. Gelpi, Monica Parati, L. Vecchia, A. Porta","doi":"10.23919/cinc53138.2021.9662906","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662906","url":null,"abstract":"Among the analytical methods estimating the complexity of the heart period (HP), the linear model-based multiscale complexity (MSC) approach allows the estimation of the complexity over time scales linked to the cardiac autonomic control, i.e. in the low frequency (LF, 0.04-0.15 Hz) and high frequency $(HF, 0.15-0.4 Hz)$ bands. In this study we exploited MSC to evaluate the differences in the HP variability complexity during daytime (DAY) and nighttime (NIGHT) in 23 healthy females (WOMEN, age $36pm 6yrs)$ ) and 21 males (MEN, age $35pm 5yrs)$ performing a 24-hour Holter electrocardiogram. Parametric power spectral analysis was applied as well for comparison. Complexity indexes were computed regardless of the temporal scale (CI) and in the LF and HF bands ( $CI_{LF}$ and $CI_{HF}$, respectively). We found that the power spectral indexes did not differentiate WOMEN and MEN, while CI and $CI_{LF}$ were higher in WOMEN during DAY. The higher HP complexity in females could be explained by a lower sympathetic drive and more complex hormonal regulation than males. We conclude that MSC was more powerful than power spectral analysis in detecting gender differences in HP variability. In addition, as cardiac control differs between females and males, preventive and therapeutic interventions should take gender differences into account.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130430927","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662736
F. Margara, B. Rodríguez, Christopher N Toepfer, A. Bueno-Orovio
Hypertrophic cardiomyopathy (HCM) is a common genetic heart disease characterised by hyperdynamic contraction and slowed relaxation. It has been proposed that cellular hypercontractility can derive from mutations that destabilise the energy-conserving myosin super relaxed state, SRX. A new drug, Mavacamten, has been shown to re-stabilise myosin SRX. Here we develop a human-based in-silico model to investigate how disease and drug-induced SRX changes alter cardiac contractility. We do this to mechanistically investigate how Mavacamten restores function in a HCM causing mutation. Our simulations show that hypercontractility is accounted for by an increased availability of crossbridges due to a reduced abundance of myosin SRX, but cellular diastolic dysfunction is only recapitulated if there is a direct crossbridge contribution to thin filament activation. Our model replicates reduced cellular contractility with Mavacamten treatment, which also rescues the hypercontractile phenotype in HCM Our model demonstrates that Mavacamten is effective in correcting HCM abnormalities caused by mutations that destabilise SRX. However, genotypes that cause HCM via other molecular pathways may be incompletely salvaged by Mavacamten.
{"title":"Mavacamten Efficacy in Mutation-specific Hypertrophic Cardiomyopathy: an In Silico Approach to Inform Precision Medicine","authors":"F. Margara, B. Rodríguez, Christopher N Toepfer, A. Bueno-Orovio","doi":"10.23919/cinc53138.2021.9662736","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662736","url":null,"abstract":"Hypertrophic cardiomyopathy (HCM) is a common genetic heart disease characterised by hyperdynamic contraction and slowed relaxation. It has been proposed that cellular hypercontractility can derive from mutations that destabilise the energy-conserving myosin super relaxed state, SRX. A new drug, Mavacamten, has been shown to re-stabilise myosin SRX. Here we develop a human-based in-silico model to investigate how disease and drug-induced SRX changes alter cardiac contractility. We do this to mechanistically investigate how Mavacamten restores function in a HCM causing mutation. Our simulations show that hypercontractility is accounted for by an increased availability of crossbridges due to a reduced abundance of myosin SRX, but cellular diastolic dysfunction is only recapitulated if there is a direct crossbridge contribution to thin filament activation. Our model replicates reduced cellular contractility with Mavacamten treatment, which also rescues the hypercontractile phenotype in HCM Our model demonstrates that Mavacamten is effective in correcting HCM abnormalities caused by mutations that destabilise SRX. However, genotypes that cause HCM via other molecular pathways may be incompletely salvaged by Mavacamten.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130471152","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662700
C. Bartolucci, P. Mesirca, Claire Belles, Eugenio Ricci, E. Torre, J. Louradour, M. Mangoni, S. Severi
Nowadays, mathematical modeling has been one of the improvements in technologically advanced science in supporting decision-making in different healthcare scenarios. In the field of numerical modelling of heart electrophysiology, several models of action potential (AP) have been developed for cardiac chambers of different species. The atrioventricular node (AVN) acts as a subsidiary pacemaker and controls impulse conduction between the atria and ventricles. Despite its physiological importance, limited data are available for computing AVN cellular electrophysiology. Further, the ionic mechanisms underlying the automaticity of AVN myocytes are incompletely understood. Only two computational models of AVN have been developed in the last decades (one for rabbit, the other for mouse but without calcium handling). We aimed to develop a new mouse AVN model. We thus build on the preliminary AP mouse AVN model published by Marger et al., which has been updated and improved, by implementing more realistic cellular compartments and calculation of dynamics and handling of intracellular $Ca^{2+}$. The new model reproduces almost all the AVN AP hallmarks and has been used to simulate the effects of blockade of ionic currents involved in AVN pacemaking.
{"title":"A Novel Computational Model of Pacemaker Activity in the Mouse Atrioventricular Node Cell","authors":"C. Bartolucci, P. Mesirca, Claire Belles, Eugenio Ricci, E. Torre, J. Louradour, M. Mangoni, S. Severi","doi":"10.23919/cinc53138.2021.9662700","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662700","url":null,"abstract":"Nowadays, mathematical modeling has been one of the improvements in technologically advanced science in supporting decision-making in different healthcare scenarios. In the field of numerical modelling of heart electrophysiology, several models of action potential (AP) have been developed for cardiac chambers of different species. The atrioventricular node (AVN) acts as a subsidiary pacemaker and controls impulse conduction between the atria and ventricles. Despite its physiological importance, limited data are available for computing AVN cellular electrophysiology. Further, the ionic mechanisms underlying the automaticity of AVN myocytes are incompletely understood. Only two computational models of AVN have been developed in the last decades (one for rabbit, the other for mouse but without calcium handling). We aimed to develop a new mouse AVN model. We thus build on the preliminary AP mouse AVN model published by Marger et al., which has been updated and improved, by implementing more realistic cellular compartments and calculation of dynamics and handling of intracellular $Ca^{2+}$. The new model reproduces almost all the AVN AP hallmarks and has been used to simulate the effects of blockade of ionic currents involved in AVN pacemaking.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"186 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116905751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662779
Federico M. Muscato, V. Corino, L. Mainardi
The automatic detection and classification of cardiac abnormalities can assist physicians in making diagnoses, saving costs in modern healthcare systems. In this study we present an automatic algorithm for classification of cardiac abnormalities included in the CinC's challenge 2021 dataset consisting of twelve-lead, six-lead, three-lead, and two-lead ECGs (team: Polimi_1). For each set of leads an ensemble of three deep learning models, trained on three different subsets, was developed. These subsets, obtained by splitting the recordings with the most frequent classes, had more balanced distributions for training and were used to train the 3 classifiers. The trained models were modified Residual Networks with a Squeeze-and-Excitation module. This module is based on the intuition of channel attention: the basic idea of this approach is to apply a weight to the Convolutional channels based on their relevance in learning before propagating to the next layer. For evaluation, we submitted our model to the official phase of the PhysioNet/Computing in Cardiology Challenge 2021. The model received scores of 0.47, 0.46, 0.45, 0.48 and 0.45 (ranked 14th, 13th, 15th, 10th, and 13th out of 39 teams) on 12-lead, 6-lead, 4-lead, 3-lead, 2-lead hidden test set, respectively; placing us in the 11th position for the mean of the 12-lead, 3-lead, and 2-lead scores.
心脏异常的自动检测和分类可以帮助医生做出诊断，节省现代医疗保健系统的成本。在这项研究中，我们提出了一种用于心脏异常分类的自动算法，该算法包含在CinC的挑战2021数据集中，该数据集由12导联、6导联、3导联和2导联心电图组成(团队:Polimi_1)。对于每组线索，开发了三个深度学习模型的集合，在三个不同的子集上进行训练。这些子集是通过将记录与最频繁的类分开得到的，具有更平衡的训练分布，并用于训练3个分类器。训练后的模型是带有挤压-激励模块的改进残差网络。该模块基于通道注意的直觉:该方法的基本思想是在传播到下一层之前，根据卷积通道在学习中的相关性对其应用权重。为了进行评估，我们将我们的模型提交给了PhysioNet/Computing in Cardiology Challenge 2021的官方阶段。模型在12-lead、6-lead、4-lead、3-lead、2-lead隐藏测试集上的得分分别为0.47、0.46、0.45、0.48、0.45(在39支队伍中排名第14、13、15、10、13);这让我们在12分，3分和2分的平均得分中排名第11位。
{"title":"Ensemble Learning of Modified Residual Networks for Classifying ECG with Different Set of Leads","authors":"Federico M. Muscato, V. Corino, L. Mainardi","doi":"10.23919/cinc53138.2021.9662779","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662779","url":null,"abstract":"The automatic detection and classification of cardiac abnormalities can assist physicians in making diagnoses, saving costs in modern healthcare systems. In this study we present an automatic algorithm for classification of cardiac abnormalities included in the CinC's challenge 2021 dataset consisting of twelve-lead, six-lead, three-lead, and two-lead ECGs (team: Polimi_1). For each set of leads an ensemble of three deep learning models, trained on three different subsets, was developed. These subsets, obtained by splitting the recordings with the most frequent classes, had more balanced distributions for training and were used to train the 3 classifiers. The trained models were modified Residual Networks with a Squeeze-and-Excitation module. This module is based on the intuition of channel attention: the basic idea of this approach is to apply a weight to the Convolutional channels based on their relevance in learning before propagating to the next layer. For evaluation, we submitted our model to the official phase of the PhysioNet/Computing in Cardiology Challenge 2021. The model received scores of 0.47, 0.46, 0.45, 0.48 and 0.45 (ranked 14th, 13th, 15th, 10th, and 13th out of 39 teams) on 12-lead, 6-lead, 4-lead, 3-lead, 2-lead hidden test set, respectively; placing us in the 11th position for the mean of the 12-lead, 3-lead, and 2-lead scores.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128412730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662854
Andrea Rozo, J. Buil, Jonathan Moeyersons, John F. Morales, Roberto Garcia van der Westen, L. Lijnen, C. Smeets, S. Jantzen, V. Monpellier, D. Ruttens, C. Hoof, S. Huffel, W. Groenendaal, C. Varon
Thoracic bio-impedance (BioZ) measurements have been proposed as an alternative for respiratory monitoring. Given the ambulatory nature of this modality, it is more prone to noise sources. In this study, two pre-trained machine learning models were used to classify BioZ signals into clean and noisy classes. The models were trained on data from patients suffering from chronic obstructive pulmonary disease, and their performance was evaluated on data from patients undergoing bariatric surgery. Additionally, transfer learning (TL) was used to optimize the models for the new patient cohort. Lastly, the effect of different breathing patterns on the performance of the machine learning models was studied. Results showed that the models performed accurately when applying them to another patient population and their performance was improved by TL. However, different imposed respiratory frequencies were found to affect the performance of the models.
{"title":"Controlled Breathing Effect on Respiration Quality Assessment Using Machine Learning Approaches","authors":"Andrea Rozo, J. Buil, Jonathan Moeyersons, John F. Morales, Roberto Garcia van der Westen, L. Lijnen, C. Smeets, S. Jantzen, V. Monpellier, D. Ruttens, C. Hoof, S. Huffel, W. Groenendaal, C. Varon","doi":"10.23919/cinc53138.2021.9662854","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662854","url":null,"abstract":"Thoracic bio-impedance (BioZ) measurements have been proposed as an alternative for respiratory monitoring. Given the ambulatory nature of this modality, it is more prone to noise sources. In this study, two pre-trained machine learning models were used to classify BioZ signals into clean and noisy classes. The models were trained on data from patients suffering from chronic obstructive pulmonary disease, and their performance was evaluated on data from patients undergoing bariatric surgery. Additionally, transfer learning (TL) was used to optimize the models for the new patient cohort. Lastly, the effect of different breathing patterns on the performance of the machine learning models was studied. Results showed that the models performed accurately when applying them to another patient population and their performance was improved by TL. However, different imposed respiratory frequencies were found to affect the performance of the models.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131893804","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662753
J. Suh, Jimyeong Kim, Eunjung Lee, Jaeill Kim, Duhun Hwang, J. Park, Junghoon Lee, Jaeseung Park, Seo-Yoon Moon, Yeonsu Kim, Min-Ho Kang, Soo-Jung Kwon, E. Choi, Wonjong Rhee
The goal of PhysioNet/Computing in Cardiology Challenge 2021 was to identify clinical diagnoses from 12 -lead and reduced-lead ECG recordings, including 6-lead, 4-lead, 3-lead, and 2-lead recordings. Our team, snu_adsl, have used EfficientNet-B3 as the base deep learning model and have investigated methods including data augmentation, self-supervised learning as pre-training, label masking that deals with multiple data sources, threshold optimization, and feature extraction. Self-supervised learning showed promising results when the size of labeled dataset was limited, but the competition's dataset turned out to be large enough that the actual gain was marginal. In consequence, we did not include self-supervised pre-training in our final entry. Our classifiers received scores of 0.48, 0.48, 0.47, 0.47, and 0.45 (ranked 12th, 10th, 11th, 11th, and 13th 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.
PhysioNet/Computing in Cardiology Challenge 2021的目标是从12导联和减少导联的心电图记录中识别临床诊断，包括6导联、4导联、3导联和2导联记录。我们的团队snu_adsl使用了EfficientNet-B3作为基础深度学习模型，并研究了包括数据增强、自监督学习作为预训练、处理多个数据源的标签屏蔽、阈值优化和特征提取在内的方法。当标记数据集的大小有限时，自监督学习显示出有希望的结果，但竞争对手的数据集足够大，实际收益是边际的。因此，我们在最终的条目中没有包括自我监督的预训练。我们的分类器获得了0.48,0.48,0.47,0.47和0.45的分数(在39个团队中排名第12，第10，第11，第11和第13)，用于12-lead, 6-lead, 4-lead, 3-lead和2-lead版本的隐藏测试集与挑战评估指标。
{"title":"Learning ECG Representations for Multi-Label Classification of Cardiac Abnormalities","authors":"J. Suh, Jimyeong Kim, Eunjung Lee, Jaeill Kim, Duhun Hwang, J. Park, Junghoon Lee, Jaeseung Park, Seo-Yoon Moon, Yeonsu Kim, Min-Ho Kang, Soo-Jung Kwon, E. Choi, Wonjong Rhee","doi":"10.23919/cinc53138.2021.9662753","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662753","url":null,"abstract":"The goal of PhysioNet/Computing in Cardiology Challenge 2021 was to identify clinical diagnoses from 12 -lead and reduced-lead ECG recordings, including 6-lead, 4-lead, 3-lead, and 2-lead recordings. Our team, snu_adsl, have used EfficientNet-B3 as the base deep learning model and have investigated methods including data augmentation, self-supervised learning as pre-training, label masking that deals with multiple data sources, threshold optimization, and feature extraction. Self-supervised learning showed promising results when the size of labeled dataset was limited, but the competition's dataset turned out to be large enough that the actual gain was marginal. In consequence, we did not include self-supervised pre-training in our final entry. Our classifiers received scores of 0.48, 0.48, 0.47, 0.47, and 0.45 (ranked 12th, 10th, 11th, 11th, and 13th 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.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134598659","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-13DOI: 10.23919/cinc53138.2021.9662864
Veronica Maidel, Maayan Lia Yizraeli Davidovich, Z. Shinar, Tal Klap
Lately, many health systems accelerated their initiatives of advanced remote monitoring systems. Moving to an unattended environment requires overcoming patients' compliance issues and demonstrating the effectiveness of remote monitoring technology. Current Early Warning Scores detection of deterioration, commonly based on spot check EMR data, demonstrates low translational impact from one facility to another. In this study we used vitals collected passively by a sensor, to build a Machine Learning model for timely prediction of deteriorating patients, within 24-hours of their transfer to ICU or death. Time series features, such as trends and vitals' variability were used in conjunction with age & comorbidity data. Evaluating the model yielded an AUROC of 0.81 on data from an inpatient setting, and an AUROC of 0.88 on an independent test set from a COVID-19 unit. The suggested model, based on passive measurement technology, performs equally well as models based on EMR that include nurse inputs. Applying the model on other acute settings (such as a COVID-19 unit) showed similar performance, increasing confidence of its robustness and transferability. The model performance combined with the fact that it does not require human compliance, makes it a good candidate for future testing on home settings.
{"title":"A Prediction Model of In-Patient Deteriorations Based on Passive Vital Signs Monitoring Technology","authors":"Veronica Maidel, Maayan Lia Yizraeli Davidovich, Z. Shinar, Tal Klap","doi":"10.23919/cinc53138.2021.9662864","DOIUrl":"https://doi.org/10.23919/cinc53138.2021.9662864","url":null,"abstract":"Lately, many health systems accelerated their initiatives of advanced remote monitoring systems. Moving to an unattended environment requires overcoming patients' compliance issues and demonstrating the effectiveness of remote monitoring technology. Current Early Warning Scores detection of deterioration, commonly based on spot check EMR data, demonstrates low translational impact from one facility to another. In this study we used vitals collected passively by a sensor, to build a Machine Learning model for timely prediction of deteriorating patients, within 24-hours of their transfer to ICU or death. Time series features, such as trends and vitals' variability were used in conjunction with age & comorbidity data. Evaluating the model yielded an AUROC of 0.81 on data from an inpatient setting, and an AUROC of 0.88 on an independent test set from a COVID-19 unit. The suggested model, based on passive measurement technology, performs equally well as models based on EMR that include nurse inputs. Applying the model on other acute settings (such as a COVID-19 unit) showed similar performance, increasing confidence of its robustness and transferability. The model performance combined with the fact that it does not require human compliance, makes it a good candidate for future testing on home settings.","PeriodicalId":126746,"journal":{"name":"2021 Computing in Cardiology (CinC)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132945262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：