We study brain-computer interfaces (BCI) based on the decoding of motor imagery (MI) from electroencephalography (EEG) neuromonitoring. The robustness of MI-BCI is a major concern in practical applications, and hence various efforts in the literature have been made to enhance the MI classification accuracy from EEG signals. Recently, classifiers based on convolutional neural networks (CNN) have achieved state-of-the-art performance. In further exploration of applying CNNs to EEG data, we propose a spatial component-wise convolutional network (SCCNet), featuring an initial convolutional layer for spatial filtering, a common processing in EEG analysis for signal enhancement and noise reduction. Through a series of optimization and validation, we show the superiority of SCCNet in MI EEG classification, outperforming other existing CNNs.