William Rudman, Jack Merullo, L. Mercurio, Carsten Eickhoff
In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train neural networks for anomaly classification. In the George B. Moody 2022 Challenge, our team, BrownBAI, received an official score of 0.406 (38/40) for murmur classification and a score of 16773 (39/40) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data.
近年来，深度学习重新定义了检测心脏异常的算法。然而，许多最先进的算法仍然依赖于从给定的心脏信号中计算手工制作的特征，然后将其输入浅一维卷积网络或变压器架构。我们提出了ACQuA(拟吸引子异常分类)，这是一种任务不可知算法，可用于各种心脏设置，从从ECG信号中分类心律失常到从PCG信号中检测心脏杂音。利用动态分析和拓扑数据分析中的定理，我们创建了信息丰富的吸引子图像，这些吸引子图像1)是人类可识别的，2)可用于训练神经网络进行异常分类。在George B. Moody 2022挑战赛中，我们的团队BrownBAI在杂音分类中获得了0.406(38/40)的官方分数，在结果分类中获得了16773(39/40)的分数。此外，我们在cnc 2017挑战赛数据上评估了我们的模型，该数据要求从业者从ECG信号中分类心律失常。在cnc 2017挑战赛数据上，我们在隐藏验证数据上将获胜的F1分数提高了约14%。
{"title":"ACQuA: Anomaly Classification with Quasi-Attractors","authors":"William Rudman, Jack Merullo, L. Mercurio, Carsten Eickhoff","doi":"10.22489/CinC.2022.201","DOIUrl":"https://doi.org/10.22489/CinC.2022.201","url":null,"abstract":"In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train neural networks for anomaly classification. In the George B. Moody 2022 Challenge, our team, BrownBAI, received an official score of 0.406 (38/40) for murmur classification and a score of 16773 (39/40) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123484183","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}
As a contribution to the George B. Moody PhysioNet Challenge 2022 we (team listNto_urHeart) propose a phonocardiogram classifier. Based on the assumption that these recordings bear similarity to music, we borrow methods from the field of computational music analysis. In contrast to end-to-end machine learning approaches, we propose a carefully-crafted processing pipeline for automatically detecting single heartbeats in phonocardiogram recordings which are then classified by a bi-directional long short-term memory network. Our approach has the advantage of not requiring manual annotations during training, therefore alleviating the lack of annotated training data. In murmur detection, we reached a weighted accuracy of 0.68 in validation, 0.668 in test (rank: 25/40) and 0.64 $pm 0.08$ during training. In predicting patient outcome, we reached 10,362 in validation, 13,866 in test (rank: 27 /39) and 11, $386pm 2,108$ during training. The results indicate that borrowing algorithms from computational music analysis could bear the potential to address challenges in phonocardiogram classification successfully.
作为对George B. Moody PhysioNet Challenge 2022的贡献，我们(团队listNto_urHeart)提出了一个心音图分类器。基于这些录音与音乐相似的假设，我们借用了计算音乐分析领域的方法。与端到端机器学习方法相比，我们提出了一个精心设计的处理管道，用于自动检测心音图记录中的单次心跳，然后通过双向长短期记忆网络对其进行分类。我们的方法的优点是在训练过程中不需要手动标注，因此减轻了标注训练数据的缺乏。在杂音检测中，我们在验证中达到了0.68的加权精度，在测试中达到了0.668(排名:25/40)，在训练中达到了0.64 $pm 0.08$。在预测患者预后方面，我们在验证中达到了10,362，在测试中达到了13,866(排名:27 /39)，在训练期间达到了11,386美元/ 2,108美元。结果表明，从计算音乐分析中借鉴算法可以成功地解决声心图分类中的挑战。
{"title":"An LSTM-based Listener for Early Detection of Heart Disease","authors":"Philip Gemke, Nicolai Spicher, T. Kacprowski","doi":"10.22489/CinC.2022.151","DOIUrl":"https://doi.org/10.22489/CinC.2022.151","url":null,"abstract":"As a contribution to the George B. Moody PhysioNet Challenge 2022 we (team listNto_urHeart) propose a phonocardiogram classifier. Based on the assumption that these recordings bear similarity to music, we borrow methods from the field of computational music analysis. In contrast to end-to-end machine learning approaches, we propose a carefully-crafted processing pipeline for automatically detecting single heartbeats in phonocardiogram recordings which are then classified by a bi-directional long short-term memory network. Our approach has the advantage of not requiring manual annotations during training, therefore alleviating the lack of annotated training data. In murmur detection, we reached a weighted accuracy of 0.68 in validation, 0.668 in test (rank: 25/40) and 0.64 $pm 0.08$ during training. In predicting patient outcome, we reached 10,362 in validation, 13,866 in test (rank: 27 /39) and 11, $386pm 2,108$ during training. The results indicate that borrowing algorithms from computational music analysis could bear the potential to address challenges in phonocardiogram classification successfully.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125270263","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}
Physiological machine learning methods have a unique opportunity to augment deep-learning engineered features with additional features derived from prior pathological knowledge. We propose an phonocardiogram (PCG) classifier that combines raw spectrogram features with crafted, physician-inspired features with an end-to-end neural network architecture. Learned spectrogram features were obtained by training a convolutional neural network (CNN) directly on the raw mel-spectrogram representation of the PCG time-series. Crafted features were based on the four stages of the cardiac cycle (S1, systole, S2, and diastole). The spectrogram features have the advantage of introducing flexibility for the model to learn abstract, low-level information that captures a variety of different rhythmic abnormalities and the latter has the advantage of using segmentation to elucidate specific, high-level, human-interpretable information. Combined features are fed into a fully connected neural network which is able to learn the relationship between the two feature types. In the George B. Moody PhysioNet Challenge 2022 test set, our team (“lubdub”) received a weighted accuracy score of 0.835 with a cost of 14905 in the clinical outcome task (ranked 31/39). For the murmur prediction task, our model received a weighted accuracy score of 0.525 and a cost of 15083 (ranked 33/40).
生理机器学习方法有一个独特的机会，可以通过从先前的病理知识中获得的额外特征来增强深度学习工程特征。我们提出了一种心音图(PCG)分类器，它将原始谱图特征与精心制作的、医生启发的特征与端到端神经网络架构相结合。通过直接在PCG时间序列的原始mel谱图表示上训练卷积神经网络(CNN)来获得学习到的谱图特征。根据心脏周期的四个阶段(S1、收缩期、S2和舒张期)绘制特征。谱图特征的优点是为模型引入了灵活性，可以学习捕获各种不同节奏异常的抽象、低级信息，后者的优点是使用分割来阐明特定的、高级的、人类可解释的信息。将组合的特征输入到一个全连接的神经网络中，该网络能够学习两种特征类型之间的关系。在George B. Moody PhysioNet Challenge 2022测试集中，我们的团队(“lubdub”)在临床结果任务(排名31/39)中获得了0.835的加权准确率评分，成本为14905。对于杂音预测任务，我们的模型的加权准确率得分为0.525，成本为15083(排名33/40)。
{"title":"Classification of Murmurs in PCG Using Combined Frequency Domain and Physician Inspired Features","authors":"Julia Ding, Jing-Jing Li, Max Xu\"","doi":"10.22489/CinC.2022.065","DOIUrl":"https://doi.org/10.22489/CinC.2022.065","url":null,"abstract":"Physiological machine learning methods have a unique opportunity to augment deep-learning engineered features with additional features derived from prior pathological knowledge. We propose an phonocardiogram (PCG) classifier that combines raw spectrogram features with crafted, physician-inspired features with an end-to-end neural network architecture. Learned spectrogram features were obtained by training a convolutional neural network (CNN) directly on the raw mel-spectrogram representation of the PCG time-series. Crafted features were based on the four stages of the cardiac cycle (S1, systole, S2, and diastole). The spectrogram features have the advantage of introducing flexibility for the model to learn abstract, low-level information that captures a variety of different rhythmic abnormalities and the latter has the advantage of using segmentation to elucidate specific, high-level, human-interpretable information. Combined features are fed into a fully connected neural network which is able to learn the relationship between the two feature types. In the George B. Moody PhysioNet Challenge 2022 test set, our team (“lubdub”) received a weighted accuracy score of 0.835 with a cost of 14905 in the clinical outcome task (ranked 31/39). For the murmur prediction task, our model received a weighted accuracy score of 0.525 and a cost of 15083 (ranked 33/40).","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125440366","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}
Daniel Enériz, Antonio Rodriguez-Almeida, H. Fabelo, S. Ortega, Francisco Balea-Fernández, N. Medrano, Belén Calvo, G. Callicó
This work presents the advances of the UZ-ULPGC team in the Heart Murmur Detection from Phonocardiogram Recordings: The George B. Moody PhysioNet Challenge 2022. As the 2016 challenge proved the success of the combination of a segmentation algorithm and a classifier, a deep learning-based murmur detector is developed using the sequence segmentation-classification. A U-Net-based segmentation model is used to extract each cardiac cycle from the PCG with state-of-the-art accuracy. Three deep models are tested for the classification: a model based on four independent 1D-convolutional feature extractors; its variation enabling combination of the features; and an autoencoder. Furthermore, to enable unique patient diagnostic, a decision model gathering all the patient-related cardiac cycles information is added. All classifiers show limited performance, probably due to the heavy class imbalance of the data at the cardiac cycle level and the minimal preprocessing chosen in the architecture. Note that our models have not been tested in the hidden challenge data and therefore we are not ranked. Hence, a 10-fold cross-validation over the training set is used to evaluate their performance, with the best model getting a weighted accuracy score in the presence task of $0.58pm 0.10$ and 10 $735pm 2208$ in Challenge cost score for the outcome.
这项工作展示了UZ-ULPGC团队在心音记录心脏杂音检测方面的进展:George B. Moody PhysioNet挑战2022。由于2016年的挑战证明了分割算法和分类器结合的成功，因此使用序列分割分类开发了基于深度学习的杂音检测器。使用基于u - net的分割模型以最先进的精度从PCG中提取每个心动周期。对三种深度模型进行了分类测试:基于四个独立的一维卷积特征提取器的模型;它的变化使特征组合;还有一个自动编码器。此外，为了实现独特的患者诊断，增加了一个决策模型，收集所有与患者相关的心脏周期信息。所有分类器都显示出有限的性能，这可能是由于在心脏周期水平上数据的严重类不平衡以及在体系结构中选择的最小预处理。请注意，我们的模型没有在隐藏挑战数据中进行测试，因此我们没有排名。因此，在训练集上进行10倍交叉验证来评估它们的性能，最佳模型在存在任务中的加权精度得分为0.58pm 0.10$和10$ 735pm 2208$的挑战成本得分。
{"title":"Exploring a Segmentation-Classification Deep Learning-based Heart Murmurs Detector","authors":"Daniel Enériz, Antonio Rodriguez-Almeida, H. Fabelo, S. Ortega, Francisco Balea-Fernández, N. Medrano, Belén Calvo, G. Callicó","doi":"10.22489/CinC.2022.266","DOIUrl":"https://doi.org/10.22489/CinC.2022.266","url":null,"abstract":"This work presents the advances of the UZ-ULPGC team in the Heart Murmur Detection from Phonocardiogram Recordings: The George B. Moody PhysioNet Challenge 2022. As the 2016 challenge proved the success of the combination of a segmentation algorithm and a classifier, a deep learning-based murmur detector is developed using the sequence segmentation-classification. A U-Net-based segmentation model is used to extract each cardiac cycle from the PCG with state-of-the-art accuracy. Three deep models are tested for the classification: a model based on four independent 1D-convolutional feature extractors; its variation enabling combination of the features; and an autoencoder. Furthermore, to enable unique patient diagnostic, a decision model gathering all the patient-related cardiac cycles information is added. All classifiers show limited performance, probably due to the heavy class imbalance of the data at the cardiac cycle level and the minimal preprocessing chosen in the architecture. Note that our models have not been tested in the hidden challenge data and therefore we are not ranked. Hence, a 10-fold cross-validation over the training set is used to evaluate their performance, with the best model getting a weighted accuracy score in the presence task of $0.58pm 0.10$ and 10 $735pm 2208$ in Challenge cost score for the outcome.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125497539","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}
Jesús Cano, V. Bertomeu-González, Lorenzo Fácia, J. Moreno-Arribas, R. Alcaraz, J. J. Rieta
Blood pressure (BP) fluctuates throughout the day, mainly due to circadian oscillations as well as a response to physical and mental stimuli. This study aims at investigating whether machine learning (ML) classifiers can detect hypertension pathology regardless of absolute BP values. The goal is to identify HTS patients from non-HTS recordings and NTS subjects from non-NTS recordings using photoplethysmography (PPG) and electrocardiography (ECG). 803 simultaneous PPG, ECG and invasive BP recordings from 51 subjects were analyzed. 668 were coherent BP segments, with high BP for HTS patients and normal BP for NTS subjects, and 135 were incoherent segments, with normal BP for HTS patients and high BP for NTS subjects. PPG and BP relationship was evaluated with discriminant features and classification models were employed to classify incoherent segments. Using the discriminant features of coherent segments for training and the set of incoherent segments for validation, K-nearest neighbors provided the best outcomes, with F1-score of 88.30%. Combining PPG and ECG recordings with ML-based methodologies would be of high interest for hypertension screening, so that HTS and NTS subjects could be properly discerned even in the case of incoherent or altered BP values. This method could be used as a support for clinical decision-making when diagnosing hypertension.
{"title":"ECG and PPG-Based Hypertension Screening Under Non-Hypertensive Blood Pressure Recordings","authors":"Jesús Cano, V. Bertomeu-González, Lorenzo Fácia, J. Moreno-Arribas, R. Alcaraz, J. J. Rieta","doi":"10.22489/CinC.2022.093","DOIUrl":"https://doi.org/10.22489/CinC.2022.093","url":null,"abstract":"Blood pressure (BP) fluctuates throughout the day, mainly due to circadian oscillations as well as a response to physical and mental stimuli. This study aims at investigating whether machine learning (ML) classifiers can detect hypertension pathology regardless of absolute BP values. The goal is to identify HTS patients from non-HTS recordings and NTS subjects from non-NTS recordings using photoplethysmography (PPG) and electrocardiography (ECG). 803 simultaneous PPG, ECG and invasive BP recordings from 51 subjects were analyzed. 668 were coherent BP segments, with high BP for HTS patients and normal BP for NTS subjects, and 135 were incoherent segments, with normal BP for HTS patients and high BP for NTS subjects. PPG and BP relationship was evaluated with discriminant features and classification models were employed to classify incoherent segments. Using the discriminant features of coherent segments for training and the set of incoherent segments for validation, K-nearest neighbors provided the best outcomes, with F1-score of 88.30%. Combining PPG and ECG recordings with ML-based methodologies would be of high interest for hypertension screening, so that HTS and NTS subjects could be properly discerned even in the case of incoherent or altered BP values. This method could be used as a support for clinical decision-making when diagnosing hypertension.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125562418","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}
Federico M. Muscato, V. Corino, M. Rivolta, P. Cerveri, A. Sanzo, A. Vicentini, R. Sassi, L. Mainardi
We developed an end-to-end automatic algorithm for the detection of signs of COVID-19 virus infection in ECGs. We analyzed 12-lead ECGs from patients infected by COVID-19 (C-group) and from a control group (NC-group). The C-group (896 cases) included patients (age range [19–96] years) hospitalized at Ospedale San Matteo in Pavia (Italy) during the first 2020 pandemic outbreak. Infection was confirmed by nasal swab testing. The NC-group (also 896 cases) was built by collecting ECG in sinus rhythm from 3 datasets: Georgia ECG (USA), PTB-XL (Germany) and CPSC 2018 (China). Control ECGs were matched by gender, age and heart rate. An additional control group, only used for testing, was extracted from the Ningbo (China) database. A 4-layers convolutional neural network (CNN), with increasing filter size plus a final fully connected (FC) layer, was designed to classify C vs NC-group. The CNN was trained and k-fold cross validated $(k=7)$ on 1536 ECGs (1316 for testing-220 for validation). Every fold model was used to classify the remaining, separate common test set of 256 ECGs. The accuracy was $0.86pm 0.01$ on validation, $0.86pm 0.01$ on the test set. The FPR on the NC-group was $0.14pm 0.03$ on validation, $0.13pm$ 0.02 on test and $0.10pm 0.01$ on the Ningbo test set $(p > 0.05,ns)$ showing that no bias was induced by the selection of datasets.
{"title":"A CNN for COVID-19 Detection Using ECG signals","authors":"Federico M. Muscato, V. Corino, M. Rivolta, P. Cerveri, A. Sanzo, A. Vicentini, R. Sassi, L. Mainardi","doi":"10.22489/CinC.2022.196","DOIUrl":"https://doi.org/10.22489/CinC.2022.196","url":null,"abstract":"We developed an end-to-end automatic algorithm for the detection of signs of COVID-19 virus infection in ECGs. We analyzed 12-lead ECGs from patients infected by COVID-19 (C-group) and from a control group (NC-group). The C-group (896 cases) included patients (age range [19–96] years) hospitalized at Ospedale San Matteo in Pavia (Italy) during the first 2020 pandemic outbreak. Infection was confirmed by nasal swab testing. The NC-group (also 896 cases) was built by collecting ECG in sinus rhythm from 3 datasets: Georgia ECG (USA), PTB-XL (Germany) and CPSC 2018 (China). Control ECGs were matched by gender, age and heart rate. An additional control group, only used for testing, was extracted from the Ningbo (China) database. A 4-layers convolutional neural network (CNN), with increasing filter size plus a final fully connected (FC) layer, was designed to classify C vs NC-group. The CNN was trained and k-fold cross validated $(k=7)$ on 1536 ECGs (1316 for testing-220 for validation). Every fold model was used to classify the remaining, separate common test set of 256 ECGs. The accuracy was $0.86pm 0.01$ on validation, $0.86pm 0.01$ on the test set. The FPR on the NC-group was $0.14pm 0.03$ on validation, $0.13pm$ 0.02 on test and $0.10pm 0.01$ on the Ningbo test set $(p > 0.05,ns)$ showing that no bias was induced by the selection of datasets.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126769230","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}
Detection of slow electrical conduction areas is crucial for providing an effective ablation therapy in ventricular tachycardia. To this aim local activations and their duration should be accurately identified. Currently mapping systems identify the precocity or lateness of a local activation with respect to a fixed reference without considering its duration. In this study we developed an automatic approach to compute local activation durations from electrograms (EGMs) and electrographic signals (ECGs). EGMs were acquired during both sinus rhythm and ventricular tachycardia with a commercial mapping catheter (Abbott Advisor HD Grid) in six patients. EGMs were band-pass filtered before processing and the analysis was based on the EGMs histogram and similarity techniques, only when a repeatable rhythm was detected in the ECGs the proposed approach was validated against 2846 activations manually annotated (GS) by an expert electrophysiologist. The mean error in the computation of the activation durations over each signal for each patient was -0.1±1.8ms (GS activation duration: 54.4±9.3ms). The developed algorithm is accurate, and the 3D dynamic maps showing slow electrical conduction areas may represent a useful tool to be integrated with activation and voltage maps to plan and assist therapeutic interventions in ventricular arrhythmias.
检测慢电传导区域对于室性心动过速提供有效的消融治疗至关重要。为此，应该准确地识别局部激活及其持续时间。目前的映射系统根据一个固定的参考来识别一个局部激活的早熟或滞后，而不考虑它的持续时间。在这项研究中，我们开发了一种自动方法，从电图(EGMs)和电图信号(ECGs)中计算局部激活持续时间。6例患者在窦性心律和室性心动过速期间使用商用测绘导管(Abbott Advisor HD Grid)获得egm。处理前对eeg进行带通滤波，分析基于eeg直方图和相似性技术，只有当在eeg中检测到可重复的节律时，才会由电生理学专家对2846个手动注释(GS)的激活进行验证。每个患者每个信号的激活持续时间计算的平均误差为-0.1±1.8ms (GS激活持续时间:54.4±9.3ms)。所开发的算法是准确的，显示慢电传导区域的3D动态图可能是一个有用的工具，可以与激活和电压图相结合，以计划和辅助室性心律失常的治疗干预。
{"title":"A New Approach for Mapping Electrical Conduction in Ventricular Tachycardia","authors":"C. Fabbri, Matteo Diani, N. Trevisi, C. Corsi","doi":"10.22489/CinC.2022.381","DOIUrl":"https://doi.org/10.22489/CinC.2022.381","url":null,"abstract":"Detection of slow electrical conduction areas is crucial for providing an effective ablation therapy in ventricular tachycardia. To this aim local activations and their duration should be accurately identified. Currently mapping systems identify the precocity or lateness of a local activation with respect to a fixed reference without considering its duration. In this study we developed an automatic approach to compute local activation durations from electrograms (EGMs) and electrographic signals (ECGs). EGMs were acquired during both sinus rhythm and ventricular tachycardia with a commercial mapping catheter (Abbott Advisor HD Grid) in six patients. EGMs were band-pass filtered before processing and the analysis was based on the EGMs histogram and similarity techniques, only when a repeatable rhythm was detected in the ECGs the proposed approach was validated against 2846 activations manually annotated (GS) by an expert electrophysiologist. The mean error in the computation of the activation durations over each signal for each patient was -0.1±1.8ms (GS activation duration: 54.4±9.3ms). The developed algorithm is accurate, and the 3D dynamic maps showing slow electrical conduction areas may represent a useful tool to be integrated with activation and voltage maps to plan and assist therapeutic interventions in ventricular arrhythmias.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122110866","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}
Emmi Antikainen, R. Rehman, T. Ahmaniemi, M. Chatterjee
Daytime sleepiness impairs the activities of daily living, especially in chronic disease patients. Typically, daytime sleepiness is measured with subjective patient reported outcomes (PROs), which could be prone to recall bias. Objective measures of daytime sleepiness, which are sensitive to change, would benefit the assessment of disease states and novel therapies that impact the quality of life. The presented study aimed to predict daytime sleepiness from two hours of continuously measured respiratory rate using a 1-dimensional convolutional neural network. A wearable biosensor was used to continuously measure electrocardiography (ECG) based respiratory rate, while the participants $(N=82)$ were asked to fill in Karolinska Sleepiness Scale three times a day. Considering the need for a sleepiness measure for chronic diseases, neurodegenerative disease (NDD, $N=14)$ patients, immune-mediated inflammatory disease (IMID, $N=42$) patients, as well as healthy participants $(N=26)$ were included in the study. The diseaseagnostic model achieved an accuracy of 63% between non-sleepy and sleepy states. The result demonstrates the potential of using respiratory rate with deep learning for an objective measure of daytime sleepiness.
{"title":"Predicting Daytime Sleepiness from Electrocardiography Based Respiratory Rate Using Deep Learning","authors":"Emmi Antikainen, R. Rehman, T. Ahmaniemi, M. Chatterjee","doi":"10.22489/CinC.2022.100","DOIUrl":"https://doi.org/10.22489/CinC.2022.100","url":null,"abstract":"Daytime sleepiness impairs the activities of daily living, especially in chronic disease patients. Typically, daytime sleepiness is measured with subjective patient reported outcomes (PROs), which could be prone to recall bias. Objective measures of daytime sleepiness, which are sensitive to change, would benefit the assessment of disease states and novel therapies that impact the quality of life. The presented study aimed to predict daytime sleepiness from two hours of continuously measured respiratory rate using a 1-dimensional convolutional neural network. A wearable biosensor was used to continuously measure electrocardiography (ECG) based respiratory rate, while the participants $(N=82)$ were asked to fill in Karolinska Sleepiness Scale three times a day. Considering the need for a sleepiness measure for chronic diseases, neurodegenerative disease (NDD, $N=14)$ patients, immune-mediated inflammatory disease (IMID, $N=42$) patients, as well as healthy participants $(N=26)$ were included in the study. The diseaseagnostic model achieved an accuracy of 63% between non-sleepy and sleepy states. The result demonstrates the potential of using respiratory rate with deep learning for an objective measure of daytime sleepiness.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128528373","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}
"aline cabasson, Olivier Meste, S. Zeemering, U. Schotten, P. Bonizzi
In noninvasive studies of atrial fibrillation (AF), especially in body surface potential map (BSPM) measurements, the dominant frequency (DF) is usually defined as the highest peak in the power spectrum, after prior cancellation or removal of the ECG components related to the ventricular activity. However, the power spectrum is often hampered by phase breaks presence in atrial signals due to either signal concatenation or to chaotic behavior. Fourier analysis (including multiple frequency components models) is used as a starting point to develop methods adapted to handle phase breaks. Fourier analysis and the average frequency derived from the phase of the analytic signal (within an AF cycle or globally) were selected as estimators of the single frequency model, and compared by means of simulations. It is found that for large phase breaks (±T/2 every half-second), and for a SNR of 5db, the 95 % confidence interval were smaller for the estimates based on the phase, within an AF cycle, of the analytic signal. For the more realistic multiple frequency model, the Fourier decomposition is extended by using a Least Mean Squares (LMS) adaptive algorithm, with or without imposing a constant magnitude. Slight differences in performances of the presented methods are exemplified on a single AF subject where the DF is computed over all the leads of the BSPM records.
{"title":"Is the Dominant Frequency Accurate Enough for Atrial Fibrillation Signals?","authors":"\"aline cabasson, Olivier Meste, S. Zeemering, U. Schotten, P. Bonizzi","doi":"10.22489/CinC.2022.418","DOIUrl":"https://doi.org/10.22489/CinC.2022.418","url":null,"abstract":"In noninvasive studies of atrial fibrillation (AF), especially in body surface potential map (BSPM) measurements, the dominant frequency (DF) is usually defined as the highest peak in the power spectrum, after prior cancellation or removal of the ECG components related to the ventricular activity. However, the power spectrum is often hampered by phase breaks presence in atrial signals due to either signal concatenation or to chaotic behavior. Fourier analysis (including multiple frequency components models) is used as a starting point to develop methods adapted to handle phase breaks. Fourier analysis and the average frequency derived from the phase of the analytic signal (within an AF cycle or globally) were selected as estimators of the single frequency model, and compared by means of simulations. It is found that for large phase breaks (±T/2 every half-second), and for a SNR of 5db, the 95 % confidence interval were smaller for the estimates based on the phase, within an AF cycle, of the analytic signal. For the more realistic multiple frequency model, the Fourier decomposition is extended by using a Least Mean Squares (LMS) adaptive algorithm, with or without imposing a constant magnitude. Slight differences in performances of the presented methods are exemplified on a single AF subject where the DF is computed over all the leads of the BSPM records.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128694729","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}
Jonas Sandelin, T. Koivisto, Jukka Sirkiä, A. Anzanpour
The purpose of this study is to attempt to identify acute myocardial infarction with high frequency serial electrocardiogram which both are ECG analyzing techniques. The idea is to combine these two techniques and see if changes between different ECGs from the same person can provide us with some information, whether it being in the high frequency or normal frequency range of ECG. A heart attack can occur at any time and therefore the possibility of using a wearable device was also researched. To answer the questions, an existing database which contained multiple ECGs for each person with high sampling frequency was used. On top of this, a new serial ECG database was gathered using a wearable device designed by the University of Turku. Using multiple ECGs, features were extracted from the signals and then used in different machine learning methods in order to classify the subjects. All of the methods seem to be relevant. High frequency ECG was found to be useful, while serial ECG provided us good results with both databases. The device was also found to be able to produce good quality ECG.
{"title":"Identification of Myocardial Infarction by High Frequency Serial ECG Measurement","authors":"Jonas Sandelin, T. Koivisto, Jukka Sirkiä, A. Anzanpour","doi":"10.22489/CinC.2022.185","DOIUrl":"https://doi.org/10.22489/CinC.2022.185","url":null,"abstract":"The purpose of this study is to attempt to identify acute myocardial infarction with high frequency serial electrocardiogram which both are ECG analyzing techniques. The idea is to combine these two techniques and see if changes between different ECGs from the same person can provide us with some information, whether it being in the high frequency or normal frequency range of ECG. A heart attack can occur at any time and therefore the possibility of using a wearable device was also researched. To answer the questions, an existing database which contained multiple ECGs for each person with high sampling frequency was used. On top of this, a new serial ECG database was gathered using a wearable device designed by the University of Turku. Using multiple ECGs, features were extracted from the signals and then used in different machine learning methods in order to classify the subjects. All of the methods seem to be relevant. High frequency ECG was found to be useful, while serial ECG provided us good results with both databases. The device was also found to be able to produce good quality ECG.","PeriodicalId":117840,"journal":{"name":"2022 Computing in Cardiology (CinC)","volume":"498 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128986205","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：