This systematic review and meta-analysis aims to study the various Artificial Intelligence (AI) algorithms used as classifiers and their novel integration with state-of-the-art technologies like Brain Computer Interface (BCI). The primary objective is to statistically explore the AI’s accuracy and other performance metrics with BCI concerning neurorehabilitation in patients with stroke and other neurological complications. 

BCIs augment traditional rehabilitation methods with cutting-edge biomedical engineering. BCIs utilise electroencephalographic (EEG) signals in conjunction with AI as classifiers, providing real time feedback. BCI approaches like P300 and Steady State Visual Evoked Potential (SSVEP) paradigms have been widely explored for diverse applications like virtual reality and gaming. Other major applications include robotic prostheses, communication and rehabilitation in people suffering from motor disabilities like spinal cord lesions, ALS and locked-in syndrome, which needs to be further established. 

An extensive search was conducted in all the major medical databases for relevant articles concerning Visual Evoked Potential based BCI. The statistical analysis was performed in R-Studio. The pooled means of accuracy and Information Transmission Rate (ITR) were analysed using the Inverse variance method and the heterogeneity was assessed using the I^2 test. Subgroup analysis was also conducted.

Both SSVEP and P300 BCI systems were statistically appreciated. However, fundamental and technical differences—such as the experimental paradigm, cognitive and sensory demands, attention types (covert vs overt), and synchronous versus asynchronous operation—may impact their suitability for neurorehabilitation applications.