Motor Imagery-based Brain-Computer Interfaces (MI-BCIs) enable control of external devices through imagined movements, showing promise for assistive technologies. However, the use of numerous EEG sensors poses practical challenges. This research proposes a method to reduce EEG electrodes in MI-BCIs using the Artificial Bee Colony (ABC) algorithm for channel selection, cupled with Common Spatial Pattern (CSP) for feature extraction. Experiments were conducted on a public BCI Competition dataset featuring four movement classes recorded via 22 EEG and 3 EOG channels. Our ABC-based channel selection method, combined with CSP and Support Vector Machine classification, achieved 70.22% accuracy using only 12 channels, compared to 65.51% with all channels on subject-independent BCI. These findings demonstrate the feasibility of developing more practical MI-BCIs with fewer EEG electrodes, potentially enhancing assistive technology usability for individuals with motor disabilities.