Cardiac arrest is the main cause of death worldwide and is of great significance for the accurate prognosis of patients. The aim of this study is to predict neurological recovery from coma in patients after cardiac arrest. We utilized physiological features of EEG signals from the PhysioNet Challenge 2023 dataset and clinical information of patients, and constructed a prediction model through convolutional neural networks (CNNs) and bidirectional long short-term memory networks (Bi-LSTMs) for feature extraction. In the feature extraction stage, we passed the raw signals into two separate CNN models to extract features, and then concatenated the outputs from the two models. We then used these features as inputs to the Bi-LSTM model to predict neurological recovery. In the experiment, we randomly divided the dataset into 80% for training and 20% for testing. The challenge score is 0.299. The results showed that our model achieved an accuracy of 83.3%, an AUC of 0.88, and an F1-score of 0.77 on the test set. Our model can effectively predict neurological recovery in patients after cardiac arrest. In addition, we will try to conduct in-depth experimental analysis, including adjusting the hyperparameter, introducing attention mechanism, etc, such as the size of the convolution kernel in the CNN model, the number of layers in the CNN model, and the size of the hidden layer in the Bi LSTM model. According to theoretical analysis, increasing the number of layers in the CNN model and the size of hidden layers in the Bi LSTM model can improve the predictive performance of the model. In conclusion, we proposed a method based on CNNs and Bi-LSTMs that can effectively predict neurological recovery in patients after cardiac arrest. These results provide valuable references for clinical medicine and have the potential to provide more accurate and effective guidance for patient treat-ment.