Artificial neural network-boosted Cardiac Arrest Survival Post-Resuscitation In-hospital (CASPRI) score accurately predicts outcome in cardiac arrest patients treated with targeted temperature management

Figures

• 

  Figure 1. Artificial neural network (ANN) model in the present study. Schema showing the input, hidden, and output layers of the ANN model. The number of neurons in the hidden layer were set empirically and ranged from 1 to 50. The output layer contains two neurons—the favorable and unfavorable neurological outcome at hospital discharge. ANN artificial neural network, CPC cerebral performance category, MAP mean arterial pressure, ROSC restoration of spontaneous circulation.

• 

  Figure 2. Predictive performance of the CASPRI score. (A) Graphical visualization of the corresponding percentage of cardiac arrest patients who received TTM in current cohort who survived to hospital discharge and who had favorable neurological outcomes for every 5-points increases of CASPRI score. (B) ROC curve with indicated AUC of the CASPRI score univariable logistic regression model to predict favorable neurological outcomes in cardiac arrest patients who received TTM. AUC area under the curve, ROC receiver operating characteristic, CASPRI Cardiac Arrest Survival Post-resuscitation In-hospital, TTM targeted temperature management.

• 

  Figure 3. Predictive performance of ANN models. ROC curves with AUCs of the (A) training and (B) validation sets of ANN model to predict favorable neurological outcomes in cardiac arrest patients who received TTM using baseline parameters of CASPRI score. AUC values are presented as mean ± SD of the five training and validation sets during five-fold cross-validation. ANN artificial neural network, AUC area under the curve, ROC receiver operating characteristic, CASPRI Cardiac Arrest Survival Post-resuscitation In-hospital, TTM targeted temperature management, SD standard deviation.

• 

  Figure 4. Significance of variables in the ANN model. Graphical representation of the relative significance of the individual parameters in the ANN model. The numbers in each color-coded bar indicate the calculated indices of the total effect of the predicting factors, with a higher value representing a greater significance attributed to the model. ANN artificial neural network, CPC cerebral performance category, MAP mean arterial pressure, ROSC restoration of spontaneous circulation.