Preparing multi-channel and multi-patient EEG dataset for convnet model

I have a question about final stage of pre-processing EEG data which is preparing dataset for model.
I want to use a deep learning model to make a decision if patients have MCI or not according to the EEG data. I have 61 patient’s data and the recorded EEG signal (30 minutes long) has 19 channels. After pre-processing steps I want to feed the data to a kind of 1D-convolutional network. Actually, I’m not sure about the final form of input dataset to the model. Should it be the data frame of all epoched data of 61 patients or maybe there is another option.