Multimodal Biosignal Analysis Algorithm for the Classification of Cardiac Rhythms During Resuscitation

Abstract

Monitoring the heart rhythm during out-of-hospital cardiac arrest (OHCA) is important to improve treatment quality. OHCA rhythms fall into five categories: asystole (AS), pulseless electrical activity (PEA), pulse-generating rhythms (PR), ventricular fibrillation (VF) and ventricular tachycardia (VT). This paper introduces an algorithm to classify these OHCA rhythms using the ECG and the thorax impedance (TI) signals recorded by the defibrillation pads. The dataset consisted of 100 OHCA patient files from which 2833 4-s signal segments were extracted: 423 AS, 912 PE, 689 PR, 643 VF, and 166 VT. The Stationary Wavelet Transform (SWT) was used to obtain 95 features from the ECG and the TI. Random Forest classifiers were used, features were ranked during training using random forest importance, and models with increasing number of features were evaluated. The optimal classifier was obtained combining 50 ECG and TI features, with a median (80% confidence interval) average recall of 86.5%

A Machine Learning Model for the Prognosis of Pulseless Electrical Activity during Out-of-Hospital Cardiac Arrest

“…TI has been successfully used to discriminate PEA from rhythms associated with ROSC, by extracting the impedance circulation component (ICC), which reflects blood flow during ROSC [ 22 , 23 ]. In fact, models combining ECG and TI have been proposed to predict immediate rhythm transitions during OHCA [ 24 ] and to discriminate rhythms in OHCA [ 25 ], and in a preliminary study, a model combining an ECG and a TI feature showed promising results for the discrimination of faPEA and unPEA on a limited dataset [ 26 ].…”