Enhancing the accuracy of shock advisory algorithms in automated external defibrillators during ongoing cardiopulmonary resuscitation using a deep convolutional Encoder-Decoder filtering model

Artificial intelligence to support out-of-hospital cardiac arrest care: A scoping review

“…The majority were carried out through collaborations between multinational groups in Europe and the United States utilizing ECG waveform data from multicenter databases. Among these studies, six (25%) aimed to predict defibrillation outcomes in ventricular fibrillation (VF) 30 , 31 , 32 , 33 , 34 , 35 , five (21%) focused on rhythm classification 36 , 37 , 38 , 39 , 40 , five (21%) developed algorithms to advise defibrillation versus no defibrillation 41 , 42 , 43 , 44 , 45 , four (17%) focused specific on the classification of pulseless electrical activity (PEA) (i.e., PEA vs pseudo-PEA vs pulsed rhythm, or favorable-PEA vs non-favorable-PEA) 46 , 47 , 48 , 49 , two (8%) aimed to predict survival outcomes 30 , 50 and two (8%) aimed to suppress CPR artifact to improve ECG segment analysis 36 , 37 . Additional studies aimed to develop ECG based classification algorithms to predict rearrest in the immediate post-ROSC period 51 , predict the presence of a pulse during CPR 52 , and predict myocardial infarction/acute coronary artery occlusion during CPR 53…”

“…Among the studies of ECG segment analysis, fourteen used ML 35 , 36 , 30 , 31 , 32 , 38 , 39 , 40 , 47 , 48 , 49 , 50 , 51 , 52 , nine used DL 33 , 34 , 37 , 38 , 41 , 42 , 43 , 44 , 45 , 46 , and one used both ML and DL 38 . All studies utilized ECG waveform characteristics (i.e., VF amplitude/frequency, QRS size/amplitude) obtained from manual defibrillators, automated external defibrillators, and/or Holter monitors.…”

“…Coult et al 50 also added clinical variables including age (<60 or ≥60 years) and sex. The analyzed ECG segments ranged from 1-second to greater than 40-seconds and were primarily segments from continuous ECG waveform captured during OHCA resuscitation (10/24, 42%) 36 , 37 , 40 , 41 , 42 , 43 , 44 , 45 . Other studies analyzed more specific ECG segments including VF segments before and after defibrillator (9/24, 38%) 38 , 50 , 31 , 32 , 33 , 34 , 35 , pulseless electrical activity (PEA) segments (4/24, 17%) 46 , 47 , 48 , 49 , and immediate post-ROSC ECG segments (1/24, 4%) 51 …”

Deep Learning Strategy for Sliding ECG Analysis during Cardiopulmonary Resuscitation: Influence of the Hands-Off Time on Accuracy

“…Several approaches seeking a simpler AED implementation use only one ECG input, where periodic CC artefacts are suppressed by pattern matching algorithms [22], coherent line removal [23], and Kalman filters [24]. Short-time Fourier transform images of the ECG spectrum have been shown to be effective for filtering out CC artefacts while processed by a condition-based filtering algorithm [25] and deep convolutional encoder/decoder [26]. Whether with or without reference signals, the aforementioned methods have common disabilities in providing filtered ECG signals either with insufficiently suppressed CC artefact components or distorted ECG waves.…”

“…Whether with or without reference signals, the aforementioned methods have common disabilities in providing filtered ECG signals either with insufficiently suppressed CC artefact components or distorted ECG waves. These warped ECG signals limit the accuracy of the automated shock advisory algorithms whether they are conventionally trained for ECG signals without artefacts [13,[15][16][17]19,20,25,26] or apply specially optimized decision rules for detection of ventricular fibrillation during CPR [22,27].…”