Motor imagery (MI), a covert cognitive process where an individual mentally simulate an action without actually moving any body part, can provide an effective neurorehabilitation tool for motor function improvement or recovery. MI can become more efficient by providing feedback to the patient indicating whether he/she employs MI correctly or not. In this study, we applied 3D convolutional neural networks (3D-CNNs) to brain images acquired using functional MRI to classify the brain state. As the result, It was suggested that MI classification using our 3D-CNNs is possible and the classifier works for unknown subjects' data.