Plataforma computacional para análise de sinais EEG baseada em python aplicada a indivíduos em coma e estimulação musical de indivíduos normais

Abstract

The electroencephalogram (EEG) allows the non-invasive recording of brain electrical dynamics from the scalp, being crucial for investigating cerebral processes. However, the complexity of the signals and the need to broaden access to analyses motivated the migration of a processing software originally developed on a computational platform to Python, an openSource programming language. During the transcription, functional improvements were incorporated, including customized preprocessing routines, criteria for discarding noisy channels, selection of quantifiers, adjustable parameters, and the development of a graphical interface that facilitated adoption and training for researchers with different levels of experience. Although the computational platform version presented superior performance in the preprocessing and processing stages, being about 60 times faster, and generated processed data files five times smaller, the Python version guaranteed functional equivalence in the calculated quantifiers, with negligible relative error (on the order of 10^(-12)), except for the coherence quantifier, which has a slightly different algorithm. Statistical routines were developed to evaluate the effects of musical stimulation, in which the percentage of power contribution (PCP) quantifier, especially in the delta and alpha bands, proved to be more informative than the median frequency. Artificial intelligence techniques were applied to classify signals in clinical and musical tasks. In the coma prognosis database, a macro F1 Score of 90% was obtained; for coma etiology classification, 63% accuracy was achieved, increasing to 70% after removing exams in the "others" category; and in musical stimulus classification, 67% accuracy was reached. This work contributes to the dissemination of open and reproducible tools in neurophysiological signal analysis and paves the way for future improvements and research in the field.