An amputee due to birth trauma represents a very unique case, in which amputation occurred from delivery process itself. Providing appropriate prosthetics is crucial for enabling such case to improve their daily workout and provide entirely new experiences. However, in such case, conventional surface electromyography (sEMG)-based methods for prosthetic control are unsuitable due to the lack of sufficient sEMG signals, as a case has never used the amputated limb since birth. As an alternative to sEMG, we apply an electroencephalography (EEG)-based approach for prosthetic control. In this study, we demonstrated the feasibility of EEG-based motor imagery decoding for hand grasp types to control a prosthetic hand. We collected 10 days of EEG data about one subject and achieved a 10x5-fold cross-validation accuracy of 0.8740 ± 0.0965 in classifying cylindrical and lateral grasps using time delay embedding (TDE) and tangent space mapping (TSM). Our results demonstrate the effectiveness of TSM as an EEG decoder and the performance improvement achieved by TDE. Future work will focus on adapting this approach for real-world brain-computer interface applications.