Abstract 364: Continuous Electrocardiogram Classification During Resuscitation Using Transfer Learning

Circulation ◽  
2020 ◽  
Vol 142 (Suppl_4) ◽  
Author(s):  
Jason Coult ◽  
Diya Sashidhar ◽  
Thomas Rea ◽  
Jennifer E Blackwood ◽  
Peter Kudenchuk ◽  
...  
Keyword(s):  
Cardiac Arrest ◽  
Transfer Learning ◽  
Learning Algorithm ◽  
Deep Convolutional Neural Networks ◽  
Class Prediction ◽  
Time Frequency ◽  
Probability Threshold ◽  
Rhythm Analysis ◽  
Morlet Wavelet Transform ◽  
Hospital Cardiac Arrest

Introduction: Cardiac arrest resuscitation requires CPR interruption for ECG rhythm analysis, but pausing CPR is adversely associated with survival. Ideally, automated rhythm analysis would occur agnostic of CPR state throughout resuscitation and discriminate non-shockable from shockable rhythms. Transfer learning of pre-trained deep convolutional neural networks (CNNs) may enable accurate ECG analysis when applied to time-frequency representations of the ECG. We designed and evaluated a transfer learning algorithm to identify ventricular fibrillation (VF), asystole (AS), and organized (OR) rhythms agnostic of CPR. Methods: In this observational study of out-of-hospital cardiac arrest, rhythms were manually diagnosed in continuous defibrillator ECG recordings. Non-overlapping adjacent 2-s ECG segments were extracted from the first 30 min of each case regardless of CPR during VF, AS, and OR. Each segment was represented by an intrafrequency-normalized Morlet wavelet transform from 4-40 Hz. Using a 2/3 subset of patients for training, a series of two ResNet-101 CNNs were retrained to perform a shock decision (VF vs. non-shockable) followed by a specific non-shockable prediction (AS, OR, or Indeterminate). Performance was evaluated in a 1/3 validation subset of patients using a range of probability decision thresholds to predict the class of each segment. Results: In total, 275100 segments were collected from 461 patients. Of 90962 segments from 152 validation patients, using a 0.7 probability threshold for class prediction, 21% (18930/90962) were indeterminate, shock vs. no-shock sensitivity and specificity were 90% (19702/21930) and 97% (48421/50102), and specificities among non-shockable rhythms for AS vs. OR were 84% (5032/5998) and 86% (37760/44104), respectively (Fig 1). Conclusion: Transfer learning may enable shock/no-shock and rhythm-specific ECG classification continuously throughout resuscitation regardless of CPR.

Abstract 184: AED Application in Non-Shockable Out-of-Hospital Cardiac Arrest Does Not Reduce Survival

Circulation ◽  
2018 ◽  
Vol 138 (Suppl_2) ◽  
Author(s):  
Ross A Pollack ◽  
Siobhan P Brown ◽  
Thomas Rea ◽  
Peter J Kudenchuk ◽  
Myron L Weisfeldt
Keyword(s):  
Cardiac Arrest ◽  
Primary Outcome ◽  
Patient Characteristics ◽  
Neurological Status ◽  
Relative Safety ◽  
Initial Rhythm ◽  
Bystander Cpr ◽  
Rhythm Analysis ◽  
In Transit ◽  
Hospital Cardiac Arrest

Introduction: It is well established that AEDs improve outcome in shockable out-of-hospital cardiac arrest (OHCA). An increasing proportion (now the majority) of OHCAs present with non-shockable rhythms. Survival from non-shockable OHCA depends on high-quality CPR in transit to definitive care. Studies of AED use in non-shockable in-hospital arrest (as opposed to OHCA) have shown reduced survival with AED application possibly due to CPR interruptions to apply pads and perform rhythm analysis. We sought to determine whether AED application in non-shockable public, witnessed OHCA has a significant association with survival to discharge. Methods: This is a retrospective analysis of OHCA from 2010-2015 at 10 Resuscitation Outcomes Consortium centers. All adult, public, witnessed non-shockable OHCAs were included. Non-shockable arrest was defined as no shock delivered by the AED or by review of defibrillator tracings (10%). The initial rhythm on EMS arrival was used to confirm the rhythm. The primary outcome was survival to hospital discharge with favorable neurological status (modified rankin score <3). The OR was adjusted for the Utstein variables. Results: During the study period there were 1,597 non-shockable public, witnessed OHCA, 9.8% of which had an AED applied. The initial rhythm on EMS arrival was PEA or asystole in 86% of cases. Significantly more OHCA in the AED applied group had CPR performed. 6.5% of those without an AED applied survived with favorable neurologic status compared to 9% with an AED. After adjustment for the Utstein variables including bystander CPR, the aOR for survival with favorable neurologic outcome was 1.38 (95% CI:0.72-2.65). Conclusion: After adjusting for patient characteristics and bystander CPR, the application of an AED in non-shockable public witnessed OHCA had no significant association with survival or neurological outcome supporting the relative safety and potential benefit of AED application in non-shockable OHCA.

Abstract P82: Implementation Of New Resuscitation Guidelines Takes One-and-a-half Year

Circulation ◽  
2008 ◽  
Vol 118 (suppl_18) ◽  
Author(s):  
Jocelyn Berdowski ◽  
Andra Schmohl ◽  
Rudolph W Koster
Keyword(s):  
Cardiac Arrest ◽  
World Wide ◽  
Prospective Observational Study ◽  
Spontaneous Circulation ◽  
Chest Compressions ◽  
Return Of Spontaneous Circulation ◽  
Resuscitation Guidelines ◽  
Rhythm Analysis ◽  
New Guidelines ◽  
Hospital Cardiac Arrest

Objective- In November 2005, updated resuscitation guidelines were introduced world-wide, and will be revised again in 2010. This study aims to determine how long it takes to implement new guidelines. Methods- This was a prospective observational study. From July 2005 to January 2008, we included all patients with a non traumatic out-of-hospital cardiac arrest. Ambulance paramedics sent all continuous ECG registrations with impedance signal by modem. We excluded ECGs from patients with Return Of Spontaneous Circulation at arrival, incomplete ECG registrations, ECGs with technical deficits or with continuous chest compressions. The same guidelines needed to be used in over 75% of the registration time in order to be labeled. We classified ECGs as guidelines 2000 if the c:v ratio was 15:2, shock blocks were present and there was rhythm analysis after each shock; guidelines 2005 if the c:v ratio was 30:2, a single shock protocol was used and chest compressions was immediately resumed after shock or rhythm analysis in a no shock scenario. We accepted 10% deviations in the amount of compressions (13–17 for 2000 guidelines, 27–33 for 2005). Results- Of the 1703 analyzable ECGs, we classified 827 (48.6%) as guidelines 2000 and 624 (36.6%) as guidelines 2005. In the remaining 252 ECGs (14.8%) 31 used guidelines 1992, 137 applied guidelines 2000 with c:v ratio of 30:2 and 84 did not show distinguishable guideline usage. Since the introduction in November 2005, it took 17 months to apply new guidelines in over 80% of the cases (figure 1 ). Conclusion- Guideline changes are slowly implemented by professionals. This needs to be taken in consideration when new guideline revisions are considered.

scholarly journals Pre-charging the defibrillator before rhythm analysis reduces hands-off time in patients with out-of-hospital cardiac arrest with shockable rhythm

Resuscitation ◽  
2021 ◽  
Author(s):  
Bo Nees Iversen ◽  
Carsten Meilandt ◽  
Ulla Væggemose ◽  
Christian Juhl Terkelsen ◽  
Hans Kirkegaard ◽  
...  
Keyword(s):  
Cardiac Arrest ◽  
Shockable Rhythm ◽  
Rhythm Analysis ◽  
Hospital Cardiac Arrest ◽  
Off Time

scholarly journals Evaluation of Scalability and Degree of Fine-Tuning of Deep Convolutional Neural Networks for COVID-19 Screening on Chest X-ray Images Using Explainable Deep-Learning Algorithm

2020 ◽  
Vol 10 (4) ◽  
pp. 213 ◽  
Author(s):  
Ki-Sun Lee ◽  
Jae Young Kim ◽  
Eun-tae Jeon ◽  
Won Suk Choi ◽  
Nan Hee Kim ◽  
...  
Keyword(s):  
Deep Learning ◽  
Learning Strategies ◽  
Transfer Learning ◽  
Learning Algorithm ◽  
Fine Tuning ◽  
Deep Convolutional Neural Networks ◽  
X Ray ◽  
Backbone Network ◽  
Deep Learning Algorithm ◽  
Chest X Ray

According to recent studies, patients with COVID-19 have different feature characteristics on chest X-ray (CXR) than those with other lung diseases. This study aimed at evaluating the layer depths and degree of fine-tuning on transfer learning with a deep convolutional neural network (CNN)-based COVID-19 screening in CXR to identify efficient transfer learning strategies. The CXR images used in this study were collected from publicly available repositories, and the collected images were classified into three classes: COVID-19, pneumonia, and normal. To evaluate the effect of layer depths of the same CNN architecture, CNNs called VGG-16 and VGG-19 were used as backbone networks. Then, each backbone network was trained with different degrees of fine-tuning and comparatively evaluated. The experimental results showed the highest AUC value to be 0.950 concerning COVID-19 classification in the experimental group of a fine-tuned with only 2/5 blocks of the VGG16 backbone network. In conclusion, in the classification of medical images with a limited number of data, a deeper layer depth may not guarantee better results. In addition, even if the same pre-trained CNN architecture is used, an appropriate degree of fine-tuning can help to build an efficient deep learning model.

scholarly journals Association Between Survival and Early Versus Later Rhythm Analysis in Out-of-Hospital Cardiac Arrest: Do Agency-Level Factors Influence Outcomes?

2014 ◽  
Vol 64 (1) ◽  
pp. 1-8 ◽  
Author(s):  
Thomas Rea ◽  
David Prince ◽  
Laurie Morrison ◽  
Clifton Callaway ◽  
Tom Aufderheide ◽  
...  
Keyword(s):  
Cardiac Arrest ◽  
Rhythm Analysis ◽  
Hospital Cardiac Arrest

scholarly journals Early versus Later Rhythm Analysis in Patients with Out-of-Hospital Cardiac Arrest

2011 ◽  
Vol 365 (9) ◽  
pp. 787-797 ◽  
Author(s):  
Ian G. Stiell ◽  
Graham Nichol ◽  
Brian G. Leroux ◽  
Thomas D. Rea ◽  
Joseph P. Ornato ◽  
...  
Keyword(s):  
Cardiac Arrest ◽  
Rhythm Analysis ◽  
Hospital Cardiac Arrest

Abstract 253: Trajectory of Ventricular Fibrillation Prognostic Status During Cardiopulmonary Resuscitation

Circulation ◽  
2020 ◽  
Vol 142 (Suppl_4) ◽  
Author(s):  
Jason Coult ◽  
Shiv Bhandari ◽  
Diya Sashidhar ◽  
Thomas Rea ◽  
Jennifer E Blackwood ◽  
...  
Keyword(s):  
Cardiac Arrest ◽  
Ventricular Fibrillation ◽  
Cardiopulmonary Resuscitation ◽  
Time Dependent ◽  
Probability Of Survival ◽  
Predicted Probability ◽  
Myocardial Energetics ◽  
Rhythm Analysis ◽  
Initial Shock ◽  
Hospital Cardiac Arrest

Background: Resuscitation from out-of-hospital cardiac arrest (OHCA) due to ventricular fibrillation (VF) typically involves continuous CPR cycles interrupted every 2 minutes for rhythm analysis and potential defibrillation. Quantitative measures of the VF ECG waveform have been proposed to guide therapy for VF arrest because they are associated with myocardial energetics, are dynamic over the course of resuscitation, and predict outcome. However, while VF waveform measures have until recently have required CPR interruption to accurately gauge prognostic status, CPR interruptions are associated with a lower chance of survival. We used a novel waveform measure previously-validated during active CPR to estimate the course of VF status through the 2-minute CPR cycle between consecutive shocks. Methods: We conducted an observational study of patients with VF OHCA who experienced recurrent VF for at least 90 seconds following initial shock. We used the continuous defibrillator ECG to calculate the VF waveform measure as a function of predicted probability of survival-with-intact-neurologic-status at 1-s intervals over the course of resuscitation between shocks. Results: We collected 499 VF ECG segments (≥90 seconds) during CPR from 313 patients. The trajectory of the average prognostic VF measure had a 3-phase time-dependent pattern (Fig. 1). During CPR, the slope of the measure decreased during the initial 25 s of VF (slope = -12%/min) and was relatively flat during the subsequent 65-s interval of VF (slope = +1%/min). Furthermore, slope decreased sharply following the cessation of CPR for rhythm analysis, charge, and shock (slope = -23%/min). Conclusion: On average, a novel VF waveform measure assessed during the scheduled cycle of CPR and rhythm analysis between consecutive shocks was characterized by a period of decline, stabilization, and then decline. Whether these changes in VF status can be used to improve care for individual patients is uncertain.

Sign in / Sign up

Export Citation Format

Share Document