After successful resuscitation from cardiac arrest, many patients remain comatose due to hypoxic-ischemic brain injury. EEG is an effective and non-invasive technique commonly used to monitor brain activity and assist in outcome prediction for comatose patients post cardiac arrest. However, prior approaches largely focus on the temporal and quantitative features, without fully taking advantage of EEG spatio-spectral information. This study presents a network that combines temporal-spatio-spectral and quantitative EEG features to address the limitations of single-mode feature extraction networks. The network comprises three blocks: the time-space block processes time-series EEG as input and extracts temporal-spatial features with channel attention, time convolutional layer, and spatial convolutional layer; time-frequency representation and feature extraction blocks transform raw EEG signals into time-frequency spectrograms and extract temporal-spectral features using the continuous wavelet transform and three feature extraction layers; the feature fusion block integrated the temporal-spatio-spectral features, the quantitative EEG features and clinical features to obtain the effective cardiac arrest features. The model is trained on I-CARE dataset includes 607 cardiac arrest subjects from seven hospitals. Model performance is evaluated with AUC and the challenge official score metric, with poor outcome defined as CPC score 3-5 and good outcome as CPC score 1-2 at 3-6 months post-cardiac arrest. We divided the data set into training set, validation set and test set according to 0.7, 0.15, 0.15. In our splited test set, the accuracy rates are 85.6%. Our model was evaluated on the PhysioNet Challenge 2023 and got 0.48 point with team name "SHE Lab”. The experimental results showed the potential of integrate temporal-spatio-spectral and quantitative EEG features in extracting potential cardiac arrest biomarkers, and may indicating the promising abilities in automatic neurological outcome prediction.