Current brain activity is a predictor of longitudinal motor imagery performance

Autor(es)

Trambaiolli, Lucas R. ; Dean, Philip J. A. ; Cravo, Andre M. ; Sterr, Annette ; Sato, Joao R. ; IEEE

• Tipo de documento: Artigo Científico
• Fonte: 2020 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC); v. N/A, p. 6-pg., 2020-01-01.

Resumo

This study aimed to evaluate whether current electroencephalographic spectral measures can predict participant's performance during future sessions of a motor imagery task. By investigating this point, we hope to understand which spectral components are related to MI "literacy". Twelve healthy subjects performed a neurofeedback task whereby a cursor was moved to one of two targets (left and right) using only motor imagery of the corresponding hands. To evaluate the effect of aptitude, we measured the Mahalanobis' distances between whole-scalp spectral patterns in four frequency bands (theta, alpha, beta, and gamma) during the first session of left and right motor imagery. Later, we used these features as inputs in a Support Vector Regressor to predict performance during the following two sessions. The performance was measured as the percentage of trials where the cursor correctly reached the target. Since our sample was balanced, this approach predicted performance on sessions two and three with mean absolute errors of 15.07 +/- 12.94% and 11.98 +/- 11.40%, respectively. The most relevant feature in both cases was the Mahalanobis' distance in alpha. These results suggest that participants who can not evoke different patterns of alpha power during left- and right-hand motor imagery during the first session, also are less likely to improve during the following training sessions. The investigation of whole-scalp differences allows a holistic comprehension of the neural basis of motor imagery. This method also characterizes a potential predictor of performance for future applications of MI-based neurofeedback and brain-computer interfaces. (AU)

• Processo FAPESP: 15/17406-5 - Decodificação emocional e neuromodulação do córtex prefrontal com registros simultâneo NIRS-EEG
• Beneficiário: Lucas Remoaldo Trambaiolli
• Modalidade de apoio: Bolsas no Brasil - Doutorado
• Processo FAPESP: 18/21934-5 - Estatística de redes: teoria, métodos e aplicações
• Beneficiário: André Fujita
• Modalidade de apoio: Auxílio à Pesquisa - Temático
• Processo FAPESP: 18/04654-9 - Séries temporais, ondaletas e dados de alta dimensão
• Beneficiário: Pedro Alberto Morettin
• Modalidade de apoio: Auxílio à Pesquisa - Temático
• Processo FAPESP: 13/10952-9 - Avaliação do substrato neural do aprendizado de controle de interface cérebro-computador em indivíduos saudáveis
• Beneficiário: Lucas Remoaldo Trambaiolli
• Modalidade de apoio: Bolsas no Brasil - Mestrado