Protocol for developing an Open Base of gRAduated EEG Signals (OBRAS)

Background . Artificial intelligence (AI) systems based on neural networks enabling the automatic analysis of electroencephalography (EEG) examinations, such as detecting episodes of paroxysmal activity are currently being actively developed. Large datasets are required for training neural network algorithms. Objective : To develop a roadmap for the Open Base of Graduated EEG Signals (OBRAS) project, designed to eliminate the shortage of high-quality, annotated EEG recordings suitable for developing and training AI algorithms, as well as for creating educational programs and digital EEG atlases. Material and methods . The project protocol was drafted, which includes the collection of native EEG signals from open sources and from clinical practice of the participants, their deidentification and structured mapping. A two-tier data organization system was proposed: i) primary grouping into folders (e.g., Normal, Epi, NonEpi) and ii) main, flexible classification using a system of tags for age, sex, examination type, pathology, presence of specific EEG patterns, etc. Eligibility criteria and technical requirements for the recordings (format, number of channels, sampling frequency) were defined. Conclusion . The OBRAS project roadmap outlines the protocol, logical sequence of work, annotation methodology, funding model (non-profit partnership), and solutions in the areas of ethics and information security. The creation of such publicly accessible database will accelerate development of AI algorithms for automated EEG analysis. Implementation of the OBRAS project may significantly contribute to developing medical, research, and information technologies in neurophysiology.