Pub Date : 2019-09-01DOI: 10.23919/CinC49843.2019.9005838
Rasmus G. Sæderup, H. Zimmermann, Dagbjört Helga Eiriksdóttir, J. Hansen, J. Struijk, S. Schmidt
Non-invasive fetal ECG (NI-FECG) extraction algorithms enable long-term continuous beat-to-beat monitoring of the fetal heart rate (FHR), as opposed to the gold standard in FHR monitoring, cardiotocography (CTG). We investigate how NI-FECG extraction algorithms selected from the CinC 2013 Challenge (CinC13) perform on data with low quality signals and how performance can be evaluated using CTG, when FQRS annotation is not possible.Four-channel NI-FECG was recorded simultaneously with a CTG trace on 22 pregnant women, gestational age 29-41 weeks. Seven algorithms were tested: The winning CinC13 entry from Varanini et al. and six algorithms from the unofficial top-scoring CinC13 entry by Behar et al. Two accuracy measures were used: 1) The RMSE between the FECG-based FHR and CTG traces; 2) The Pearson correlation coefficient r between the FECG-based FHR and CTG trace and its average over all recordings, $bar r$.The algorithms with the lowest RMSE’s are Behar’s FUSE-SMOOTH, a constant FHR, and Varanini, while the Varanini algorithm delivers the best correlation with the CTG trace $(bar r = 0.73)$ with 41% of the recordings having r > 0.8, whereas the other algorithms have $bar r leq 0.59$ and ≤ 29% of the recordings with r > 0.8. FHR was estimated accurately in some recordings and poorly in others, believed to be due to large differences in signal quality.
无创胎儿心电图(NI-FECG)提取算法能够长期连续监测胎儿心率(FHR)，而不是FHR监测的金标准，心脏造影(CTG)。我们研究了从cinc2013挑战赛(CinC13)中选择的NI-FECG提取算法在低质量信号数据上的表现，以及在不可能进行FQRS注释时如何使用CTG评估性能。对22例孕龄29-41周的孕妇同时记录四通道NI-FECG和CTG。测试了七种算法:来自Varanini等人的获奖CinC13条目和来自Behar等人的非官方得分最高的CinC13条目的六种算法。采用两种精度测量方法:1)基于FHR和CTG迹线的RMSE;2)基于feg的FHR与CTG迹线的Pearson相关系数r及其在所有记录上的平均值，$bar r$。RMSE最低的算法是Behar的FUSE-SMOOTH算法，FHR恒定，Varanini算法，而Varanini算法与CTG迹线的相关性最好$(bar r = 0.73)$为41% of the recordings having r > 0.8, whereas the other algorithms have $bar r leq 0.59$ and ≤ 29% of the recordings with r > 0.8. FHR was estimated accurately in some recordings and poorly in others, believed to be due to large differences in signal quality.
{"title":"Comparison of Cardiotocography and Fetal Heart Rate Estimators Based on Non-Invasive Fetal ECG","authors":"Rasmus G. Sæderup, H. Zimmermann, Dagbjört Helga Eiriksdóttir, J. Hansen, J. Struijk, S. Schmidt","doi":"10.23919/CinC49843.2019.9005838","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005838","url":null,"abstract":"Non-invasive fetal ECG (NI-FECG) extraction algorithms enable long-term continuous beat-to-beat monitoring of the fetal heart rate (FHR), as opposed to the gold standard in FHR monitoring, cardiotocography (CTG). We investigate how NI-FECG extraction algorithms selected from the CinC 2013 Challenge (CinC13) perform on data with low quality signals and how performance can be evaluated using CTG, when FQRS annotation is not possible.Four-channel NI-FECG was recorded simultaneously with a CTG trace on 22 pregnant women, gestational age 29-41 weeks. Seven algorithms were tested: The winning CinC13 entry from Varanini et al. and six algorithms from the unofficial top-scoring CinC13 entry by Behar et al. Two accuracy measures were used: 1) The RMSE between the FECG-based FHR and CTG traces; 2) The Pearson correlation coefficient r between the FECG-based FHR and CTG trace and its average over all recordings, $bar r$.The algorithms with the lowest RMSE’s are Behar’s FUSE-SMOOTH, a constant FHR, and Varanini, while the Varanini algorithm delivers the best correlation with the CTG trace $(bar r = 0.73)$ with 41% of the recordings having r > 0.8, whereas the other algorithms have $bar r leq 0.59$ and ≤ 29% of the recordings with r > 0.8. FHR was estimated accurately in some recordings and poorly in others, believed to be due to large differences in signal quality.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"162 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73922468","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005892
C. Tobón, Diana C. Pachajoa, J. P. Ugarte, J. Saiz
Exposure to gaseous air pollutants such as carbon monoxide (CO), nitric oxide (NO) and sulfur dioxide (SO2) promotes the occurrence of cardiac diseases. Investigations have shown that CO and SO2 block the calcium channel (ICaL) of myocytes. The SO2 also increases the sodium channel (INa), the transient outward (Ito) and inward rectifying (IK1) potassium currents. The NO blocks INa and increases ICaL. We developed concentration dependent equations to simulate the gaseous pollutants effects on the ionic currents. They were incorporated in the Courtemanche model of human atrial cell and in a 2D tissue model. A train of 10 stimuli was applied. The action potential duration (APD) was measured. S1-S2 cross-field protocol was applied to initiate a rotor. The CO and SO2 concentrations from 0 to 1000 uM and NO concentration from 0 to 500 nM were implemented. Six concentration combinations were simulated (cases 1 to 6). The gaseous air pollutants caused an APD shortening and loss of plateau phase of the action potential in a fraction that increases as the pollutant concentration increases. When the highest concentration was applied, the APD decreased by 81%. In the 2D model, from case 4 conditions it was possible to generate rotor, propagating with high stability. These results show pro-arrhythmic effects of gaseous air pollutants.
{"title":"In Silico Study of Gaseous Air Pollutants Effects on Human Atrial Tissue","authors":"C. Tobón, Diana C. Pachajoa, J. P. Ugarte, J. Saiz","doi":"10.23919/CinC49843.2019.9005892","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005892","url":null,"abstract":"Exposure to gaseous air pollutants such as carbon monoxide (CO), nitric oxide (NO) and sulfur dioxide (SO2) promotes the occurrence of cardiac diseases. Investigations have shown that CO and SO2 block the calcium channel (ICaL) of myocytes. The SO2 also increases the sodium channel (INa), the transient outward (Ito) and inward rectifying (IK1) potassium currents. The NO blocks INa and increases ICaL. We developed concentration dependent equations to simulate the gaseous pollutants effects on the ionic currents. They were incorporated in the Courtemanche model of human atrial cell and in a 2D tissue model. A train of 10 stimuli was applied. The action potential duration (APD) was measured. S1-S2 cross-field protocol was applied to initiate a rotor. The CO and SO2 concentrations from 0 to 1000 uM and NO concentration from 0 to 500 nM were implemented. Six concentration combinations were simulated (cases 1 to 6). The gaseous air pollutants caused an APD shortening and loss of plateau phase of the action potential in a fraction that increases as the pollutant concentration increases. When the highest concentration was applied, the APD decreased by 81%. In the 2D model, from case 4 conditions it was possible to generate rotor, propagating with high stability. These results show pro-arrhythmic effects of gaseous air pollutants.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"29 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75757067","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005811
Shaun Eisner, F. Fenton, I. Uzelac
Optical imaging methods on ex-vivo hearts have had large impact in furthering our understanding of cardiac electrophysiology. One common problem in this method is a baseline wandering of the fluorescence signals over time, caused by dye photo-bleaching, small variation of the excitation light source, or other similar artifacts. Due to its relative magnitude, the removal of baseline wandering can be a nontrivial task and has major implications for analyzing important physiological dynamics such as traveling waves and alternans. Here we present a computational technique for the removal of such baseline wandering based on Proportional-Integral-Derivative (PID) closed loop feedback. The PID method applied a continuous control stimulus to the input Vm based on an error value which is defined by Euclidean distance from a pre-computed setpoint in phase space. We quantify and validate the PID control method by adding a linear combination of arbitrary sinusoidal drift, of frequency less than the signal pacing frequency, to the system signal Vm. The PID control loop effectively removed the baseline wandering with minimal degradation to the input Vm, and thus provides a viable tool for baseline wandering removal when implemented in an appropriate phase space. The computational simplicity of the method also lends itself to implementation in embedded systems, such as Arduinos and Raspberry-Pis.
{"title":"Baseline Wandering Removal in Optical Mapping Measurements With PID Control in Phase Space","authors":"Shaun Eisner, F. Fenton, I. Uzelac","doi":"10.23919/CinC49843.2019.9005811","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005811","url":null,"abstract":"Optical imaging methods on ex-vivo hearts have had large impact in furthering our understanding of cardiac electrophysiology. One common problem in this method is a baseline wandering of the fluorescence signals over time, caused by dye photo-bleaching, small variation of the excitation light source, or other similar artifacts. Due to its relative magnitude, the removal of baseline wandering can be a nontrivial task and has major implications for analyzing important physiological dynamics such as traveling waves and alternans. Here we present a computational technique for the removal of such baseline wandering based on Proportional-Integral-Derivative (PID) closed loop feedback. The PID method applied a continuous control stimulus to the input Vm based on an error value which is defined by Euclidean distance from a pre-computed setpoint in phase space. We quantify and validate the PID control method by adding a linear combination of arbitrary sinusoidal drift, of frequency less than the signal pacing frequency, to the system signal Vm. The PID control loop effectively removed the baseline wandering with minimal degradation to the input Vm, and thus provides a viable tool for baseline wandering removal when implemented in an appropriate phase space. The computational simplicity of the method also lends itself to implementation in embedded systems, such as Arduinos and Raspberry-Pis.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"61 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74680236","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005729
T. Gerach, D. Weiß, O. Dössel, A. Loewe
In silico studies are often used to analyze mechanisms of cardiac arrhythmias. The electrophysiological cell models that are used to simulate the membrane potential in these studies range from highly detailed physiological models to simplistic phenomenological models.To effectively cover the middle ground between those cell models, we utilize the manifold boundary approximation method (MBAM) to systematically reduce the widely used O’Hara-Rudy ventricular cell model (ORd) and investigate the influence of parametrization of the model as well as different strategies of choosing input quantities, further called quantities of interest (QoI).As a result of the reduction process, we present three reduced model variants of the ORd model that only contain a fraction of the original model’s ionic currents resulting in a twofold speedup in computation times compared to the original model. We find that the reduced models show similar action potential duration restitution and repolarization rates. Additionally, we are able to initialize and observe stable spiral wave dynamics on a 3D tissue patch for 2 out of the 3 reduced models.
{"title":"Observation Guided Systematic Reduction of a Detailed Human Ventricular Cell Model","authors":"T. Gerach, D. Weiß, O. Dössel, A. Loewe","doi":"10.23919/CinC49843.2019.9005729","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005729","url":null,"abstract":"In silico studies are often used to analyze mechanisms of cardiac arrhythmias. The electrophysiological cell models that are used to simulate the membrane potential in these studies range from highly detailed physiological models to simplistic phenomenological models.To effectively cover the middle ground between those cell models, we utilize the manifold boundary approximation method (MBAM) to systematically reduce the widely used O’Hara-Rudy ventricular cell model (ORd) and investigate the influence of parametrization of the model as well as different strategies of choosing input quantities, further called quantities of interest (QoI).As a result of the reduction process, we present three reduced model variants of the ORd model that only contain a fraction of the original model’s ionic currents resulting in a twofold speedup in computation times compared to the original model. We find that the reduced models show similar action potential duration restitution and repolarization rates. Additionally, we are able to initialize and observe stable spiral wave dynamics on a 3D tissue patch for 2 out of the 3 reduced models.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"5 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75573830","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005899
Shaleka Agrawal, G. Ramlugun, Kevin Jamart, James Kennelly, Jesse L. Ashton, G. Sands, M. Zarzoso, Jichao Zhao
Individual components of metabolic syndrome (MetS) have been correlated with atrial fibrillation (AF), but as a whole, the exact mechanism underlying the increased susceptibility of AF still remains unclear. This study identifies key structural substrates in a robust obesogenic dietary model of MetS in the rabbits. The rabbit atria from both MetS and controls (N=3 each) were processed and incubated in wheat germ agglutinin (WGA) to label cell membranes and collagen. Confocal microscopy was used to image the tissue. The collagen and cell membranes were segmented using a robust machine learning architecture, V-net. Quantification of fibrosis was done by calculating the ratio of total pixels of collagen to those of atrial tissue in each of the segmented images. Cell hypertrophy measurements were calculated by measuring means of individual cell diameters. We discovered atrial dilation, increased collagen, cell hypertrophy and reduction in axial-tubules in MetS atria. These are established arrhythmogenic phenotypes which might lead to increased AF susceptibility.
{"title":"Structural Basis of Atrial Arrhythmogenesis in Metabolic Syndrome","authors":"Shaleka Agrawal, G. Ramlugun, Kevin Jamart, James Kennelly, Jesse L. Ashton, G. Sands, M. Zarzoso, Jichao Zhao","doi":"10.23919/CinC49843.2019.9005899","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005899","url":null,"abstract":"Individual components of metabolic syndrome (MetS) have been correlated with atrial fibrillation (AF), but as a whole, the exact mechanism underlying the increased susceptibility of AF still remains unclear. This study identifies key structural substrates in a robust obesogenic dietary model of MetS in the rabbits. The rabbit atria from both MetS and controls (N=3 each) were processed and incubated in wheat germ agglutinin (WGA) to label cell membranes and collagen. Confocal microscopy was used to image the tissue. The collagen and cell membranes were segmented using a robust machine learning architecture, V-net. Quantification of fibrosis was done by calculating the ratio of total pixels of collagen to those of atrial tissue in each of the segmented images. Cell hypertrophy measurements were calculated by measuring means of individual cell diameters. We discovered atrial dilation, increased collagen, cell hypertrophy and reduction in axial-tubules in MetS atria. These are established arrhythmogenic phenotypes which might lead to increased AF susceptibility.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"55 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78605699","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005733
Dagbjört Helga Eiriksdóttir, Rasmus G. Sæderup, Diana Riknagel, H. Zimmermann, Maciej Plocharski, J. Hansen, J. Struijk, S. Schmidt
Monitoring cardiovascular activity during pregnancy is of high importance for identifying abnormal development of the fetus. Automated cardiovascular auscultation of the abdomen in both infrasonic and audible frequencies is a non-invasive method for monitoring the maternal and fetal health, including blood flow to the placenta. However, the quality of such recordings is often compromised by artifacts. The purpose of this study was to automatically identify high-quality auscultation signals. 324 recordings were obtained with two microphones placed bilaterally on the abdomen of 90 pregnant women (gestational age of 28-41 weeks), with signal duration of 30 s - 180 s. The signals were band-pass filtered to infrasonic frequencies (2.5 Hz - 25 Hz) and audible low frequencies (25 Hz - 125 Hz), divided into 10 s segments, and areas with unwanted transients were removed. Five features were calculated for segments of at least five continuous seconds. A logistic regression model was trained and tested using the identified features, obtaining a maximum classification accuracy of 92.8% for the infrasonic frequencies (81.6% sensitivity and 97.0% specificity), and 96.1% accuracy for the audible frequencies (90.4% sensitivity and 97.2% specificity). These results demonstrate the feasibility of automatical identification of high-quality segments at infrasonic and audible frequencies.
{"title":"Quality Assessment of Maternal and Fetal Cardiovascular Sounds Recorded From the Skin Near the Uterine Arteries During Pregnancy","authors":"Dagbjört Helga Eiriksdóttir, Rasmus G. Sæderup, Diana Riknagel, H. Zimmermann, Maciej Plocharski, J. Hansen, J. Struijk, S. Schmidt","doi":"10.23919/CinC49843.2019.9005733","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005733","url":null,"abstract":"Monitoring cardiovascular activity during pregnancy is of high importance for identifying abnormal development of the fetus. Automated cardiovascular auscultation of the abdomen in both infrasonic and audible frequencies is a non-invasive method for monitoring the maternal and fetal health, including blood flow to the placenta. However, the quality of such recordings is often compromised by artifacts. The purpose of this study was to automatically identify high-quality auscultation signals. 324 recordings were obtained with two microphones placed bilaterally on the abdomen of 90 pregnant women (gestational age of 28-41 weeks), with signal duration of 30 s - 180 s. The signals were band-pass filtered to infrasonic frequencies (2.5 Hz - 25 Hz) and audible low frequencies (25 Hz - 125 Hz), divided into 10 s segments, and areas with unwanted transients were removed. Five features were calculated for segments of at least five continuous seconds. A logistic regression model was trained and tested using the identified features, obtaining a maximum classification accuracy of 92.8% for the infrasonic frequencies (81.6% sensitivity and 97.0% specificity), and 96.1% accuracy for the audible frequencies (90.4% sensitivity and 97.2% specificity). These results demonstrate the feasibility of automatical identification of high-quality segments at infrasonic and audible frequencies.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"36 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75947777","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005929
Zhengling He, Xianxiang Chen, Zhen Fang, Weidong Yi, Chenshuo Wang, Li Jiang, Zhongkai Tong, Zhongrui Bai, Yueqi Li, Yichen Pan
Early prediction of sepsis can help to identify potential risks in time and help take necessary measures to prevent more dangerous situations from occurring. In PhysioNet/Computing in Cardiology Challenge 2019, we integrate Long Short Term Memory (LSTM) recurrent neural network and ensemble learning to achieve early sepsis prediction. Specifically, we tackle the problem of class imbalance and data missing firstly, and then we manually extract features according to the prior knowledge from the medical field. In addition, we regard the prediction of sepsis as a time series prediction problem and pre-train LSTM-based models as feature extractors to obtain the "deep" features on time series that might be related to the onset of sepsis. Manual features and "deep" features are then used to train prediction models under the framework of ensemble learning, including Extreme Gradient Boosting (XGBoost) and Gradient Boosting Decision Tree (GBDT) regressor. The final normalized utility score our team (UCAS_DataMiner) have obtained was 0.313 on full hidden test set.
对脓毒症的早期预测有助于及时发现潜在的风险，并采取必要的措施，防止更危险的情况发生。在PhysioNet/Computing In Cardiology Challenge 2019中，我们整合了长短期记忆(LSTM)递归神经网络和集成学习来实现早期脓毒症预测。具体来说，我们首先解决类别失衡和数据缺失的问题，然后根据医学领域的先验知识手动提取特征。此外，我们将脓毒症的预测视为一个时间序列预测问题，并将基于lstm的预训练模型作为特征提取器，以获得时间序列上可能与脓毒症发病相关的“深度”特征。然后使用人工特征和“深度”特征在集成学习框架下训练预测模型，包括极端梯度增强(XGBoost)和梯度增强决策树(GBDT)回归器。我们的团队(UCAS_DataMiner)在完全隐藏测试集上获得的最终标准化效用得分为0.313。
{"title":"Early Sepsis Prediction Using Ensemble Learning with Features Extracted from LSTM Recurrent Neural Network","authors":"Zhengling He, Xianxiang Chen, Zhen Fang, Weidong Yi, Chenshuo Wang, Li Jiang, Zhongkai Tong, Zhongrui Bai, Yueqi Li, Yichen Pan","doi":"10.23919/CinC49843.2019.9005929","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005929","url":null,"abstract":"Early prediction of sepsis can help to identify potential risks in time and help take necessary measures to prevent more dangerous situations from occurring. In PhysioNet/Computing in Cardiology Challenge 2019, we integrate Long Short Term Memory (LSTM) recurrent neural network and ensemble learning to achieve early sepsis prediction. Specifically, we tackle the problem of class imbalance and data missing firstly, and then we manually extract features according to the prior knowledge from the medical field. In addition, we regard the prediction of sepsis as a time series prediction problem and pre-train LSTM-based models as feature extractors to obtain the \"deep\" features on time series that might be related to the onset of sepsis. Manual features and \"deep\" features are then used to train prediction models under the framework of ensemble learning, including Extreme Gradient Boosting (XGBoost) and Gradient Boosting Decision Tree (GBDT) regressor. The final normalized utility score our team (UCAS_DataMiner) have obtained was 0.313 on full hidden test set.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"1049 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77643279","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005923
C. Pou-Prom, Zhen Yang, Maitreyee Sidhaye, David Dai
Sepsis is a life-threatening condition that is caused by infection, and is estimated to affects an estimated 1.7 million adults in the United States and contributes to 265,000 deaths annually. Identifying sepsis before it happens and treating it earlier leads to decreased mortality and decreased lengths of stay. As part of the PhysioNet/Computing in Cardiology Challenge 2019, we developed an ensemble-based approach for the early detection of sepsis in ICU patients.Our final model predicted sepsis using the previous 24 hours of data, and consisted of a combination of two con-volutional neural networks and a random forest trained on different subsets of the data. In training our models, we experimented with random undersampling and cluster-based undersampling as a means for addressing severe class imbalance. On validation data, our final model achieved a utility score of 0.432 on hospital A (AUROC: 0.794, AUPRC: 0.101), 0.247 on hospital B (AUROC: 0.816, AUPRC: 0.056), and a utility of 0.375 on combined data from both hospitals (AUROC: 0.809, AUPRC: 0.089). On the heldout test data, the model obtained a utility score of 0.266 and we received an official ranking of 31/79.
{"title":"Development of an Early Warning System for Sepsis","authors":"C. Pou-Prom, Zhen Yang, Maitreyee Sidhaye, David Dai","doi":"10.23919/CinC49843.2019.9005923","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005923","url":null,"abstract":"Sepsis is a life-threatening condition that is caused by infection, and is estimated to affects an estimated 1.7 million adults in the United States and contributes to 265,000 deaths annually. Identifying sepsis before it happens and treating it earlier leads to decreased mortality and decreased lengths of stay. As part of the PhysioNet/Computing in Cardiology Challenge 2019, we developed an ensemble-based approach for the early detection of sepsis in ICU patients.Our final model predicted sepsis using the previous 24 hours of data, and consisted of a combination of two con-volutional neural networks and a random forest trained on different subsets of the data. In training our models, we experimented with random undersampling and cluster-based undersampling as a means for addressing severe class imbalance. On validation data, our final model achieved a utility score of 0.432 on hospital A (AUROC: 0.794, AUPRC: 0.101), 0.247 on hospital B (AUROC: 0.816, AUPRC: 0.056), and a utility of 0.375 on combined data from both hospitals (AUROC: 0.809, AUPRC: 0.089). On the heldout test data, the model obtained a utility score of 0.266 and we received an official ranking of 31/79.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"18 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79146631","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005662
S. Carrasco-Sosa, A. Guillén-Mandujano
In 13 healthy subjects we assessed the effect of hypoglycemia (HG) provoked by insulin on: R-R intervals (RR), systolic pressure (SP), diastolic pressure (DP), pulse pressure (PP), respiratory frequency (RF) and tidal volume (VT) 5-min time series; the instantaneous time courses of their low-frequency (LFRR, LFSP, LFDP, LFPP), high-frequency (HFRR, HFRes) powers and their respective central frequencies (cfLFRR, cfLFSP, cfLFDP, cfLFPP), computed by a time-frequency distribution; instantaneous baroreflex (BRS) and respiratory sinus arrhythmia sensitivities (RSAS), obtained by alpha index, and their coherences (cBRS and cRSAS) by cross time-frequency analysis. Peak HG (2.7±0.3 mmol/l) induced: 1) decreases (p<0.03) in five 1-min epoch means (EM) of HFRR, LFRR, BRS and RSAS dynamics, three EM of CFLFPP and cBRS, two EM of CFLFRR and CFLFSP; 2) increases (p<0.02) in five EM of SP, DP, PP, VT and RF, three EM of HFRes, two EM of LFSP and LFDP, one EM of LFPP; 3) no change in CFLFDP, RR and cRSAS. In healthy subjects, insulin-provoked HG elicits changes in the fluctuating time courses of all measures studied, integrating a counterregulatory response of autonomic control mechanisms and vagal depression associated with sympathetic, cardiovascular and respiratory activation.
{"title":"Instantaneous Time Course of the Autonomic Cardiovascular and Respiratory Response of Healthy Subjects to Hypoglycemic Stimulus","authors":"S. Carrasco-Sosa, A. Guillén-Mandujano","doi":"10.23919/CinC49843.2019.9005662","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005662","url":null,"abstract":"In 13 healthy subjects we assessed the effect of hypoglycemia (HG) provoked by insulin on: R-R intervals (RR), systolic pressure (SP), diastolic pressure (DP), pulse pressure (PP), respiratory frequency (RF) and tidal volume (V<inf>T</inf>) 5-min time series; the instantaneous time courses of their low-frequency (LF<inf>RR</inf>, LF<inf>SP</inf>, LF<inf>DP</inf>, LF<inf>PP</inf>), high-frequency (HF<inf>RR</inf>, HF<inf>Res</inf>) powers and their respective central frequencies (cfLF<inf>RR</inf>, cfLF<inf>SP</inf>, cfLF<inf>DP</inf>, cfLF<inf>PP</inf>), computed by a time-frequency distribution; instantaneous baroreflex (BRS) and respiratory sinus arrhythmia sensitivities (RSAS), obtained by alpha index, and their coherences (cBRS and cRSAS) by cross time-frequency analysis. Peak HG (2.7±0.3 mmol/l) induced: 1) decreases (p<0.03) in five 1-min epoch means (EM) of HF<inf>RR</inf>, LF<inf>RR</inf>, BRS and RSAS dynamics, three EM of CFLFPP and cBRS, two EM of CFLFRR and CFLFSP; 2) increases (p<0.02) in five EM of SP, DP, PP, V<inf>T</inf> and RF, three EM of HF<inf>Res</inf>, two EM of LF<inf>SP</inf> and LF<inf>DP</inf>, one EM of LF<inf>PP</inf>; 3) no change in <inf>CF</inf>LF<inf>DP</inf>, RR and cRSAS. In healthy subjects, insulin-provoked HG elicits changes in the fluctuating time courses of all measures studied, integrating a counterregulatory response of autonomic control mechanisms and vagal depression associated with sympathetic, cardiovascular and respiratory activation.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"14 1","pages":"Page 1-Page 4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81269630","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 : 2019-09-01DOI: 10.23919/CinC49843.2019.9005872
Zhongyao Zhao, Xingyao Wang, Zhipeng Cai, Jianqing Li, Chengyu Liu
Progress in wearable techniques makes the long-term daily electrocardiogram (ECG) monitoring possible. Premature ventricular contraction (PVC) is one of the most common cardiac arrhythmias. This study proposed a method by combining the modified frequency slice wavelet transform (MFSWT) and convolutional neural network (CNN). Training data are from the 2018 China physiological signal challenge (934 PVC and 906 non-PVC recordings). The first 10-s ECG waveforms in each recording were transformed into 2-D time-frequency images (frequency range of 0-50 Hz and size of 300 × 100) using MFSWT. A 25-layer CNN structure was constructed, which includes five convolution layers with kernel size of 3×3, five dropout layers, five ReLU layers, five maximum pooling layers with kernel size of 2 × 2, a flatten layer, two fully connected layers, as well as the input and output layers. Test data were recorded from 12-lead Smart ECG vests, including 775 PVC and 742 non-PVC recordings. Results showed that, the proposed method achieved a high accuracy of 97.89% for PVC/non-PVC episodes classification, indicating that the combination of MFSWT and CNN provides new insight to accurately identify PVC from the wearable ECG recordings.
{"title":"PVC Recognition for Wearable ECGs Using Modified Frequency Slice Wavelet Transform and Convolutional Neural Network","authors":"Zhongyao Zhao, Xingyao Wang, Zhipeng Cai, Jianqing Li, Chengyu Liu","doi":"10.23919/CinC49843.2019.9005872","DOIUrl":"https://doi.org/10.23919/CinC49843.2019.9005872","url":null,"abstract":"Progress in wearable techniques makes the long-term daily electrocardiogram (ECG) monitoring possible. Premature ventricular contraction (PVC) is one of the most common cardiac arrhythmias. This study proposed a method by combining the modified frequency slice wavelet transform (MFSWT) and convolutional neural network (CNN). Training data are from the 2018 China physiological signal challenge (934 PVC and 906 non-PVC recordings). The first 10-s ECG waveforms in each recording were transformed into 2-D time-frequency images (frequency range of 0-50 Hz and size of 300 × 100) using MFSWT. A 25-layer CNN structure was constructed, which includes five convolution layers with kernel size of 3×3, five dropout layers, five ReLU layers, five maximum pooling layers with kernel size of 2 × 2, a flatten layer, two fully connected layers, as well as the input and output layers. Test data were recorded from 12-lead Smart ECG vests, including 775 PVC and 742 non-PVC recordings. Results showed that, the proposed method achieved a high accuracy of 97.89% for PVC/non-PVC episodes classification, indicating that the combination of MFSWT and CNN provides new insight to accurately identify PVC from the wearable ECG recordings.","PeriodicalId":6697,"journal":{"name":"2019 Computing in Cardiology (CinC)","volume":"159 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84992036","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：