A protocol for fMRI visual decoding

• Autor(a) principal: Peres, André Salles Cunha
• Data de Publicação: 2014
• Outros Autores: Sato, João Ricardo, dos Santos, Antônio Carlos, Hallak, Jaime Eduardo Cecílio, Ribeiro, Sidarta Tollendal Gomes, Araújo, Dráulio Barros de
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UFRN
• Texto Completo: https://repositorio.ufrn.br/jspui/handle/123456789/24027
• topic: Decoding; fMRI; BOLD distribution; Visual cortex

Introdução Functional magnetic resonance imaging (fMRI) is widely used to assess patterns of brain activity in response to specific tasks. Recent advances of signal processing tools opened the perspective of decoding information from different stimuli based on fMRI brain activity. Currently, the decoding of visual information is the most successful strategy. Typically, during the encoding phase the volunteers passively see a large number of images and a pattern of the fMRI signal is associated to each one of them. Based only on these BOLD signal patterns, statistical algorithms are used to infer what was the image seen by the subject. A common strategy used for visual cortex decoding is to separate the images into categories, with the intent of creating an average of BOLD distribution for each category. Thus, decoding refers to indicating the category to which an image belongs to. Objetivos Our purpose in this work is to evaluate the feasibility of implementing a visual cortex decoding protocol based on six categories: tree, car, house, food, person, and reptile. Métodos Two asymptomatic volunteers were invited to participate in the study. They were asked to passively watch a set of 1,440 images divided into these six categories, while fMRI data was continuously being acquired. Subjects participated in 13 sessions of 30 minutes each. fMRI analysis was based on the General Linear Model implemented in SPM8 (UCL ­ UK). A threshold was set at p < 0.05 (FWE, corrected). The BOLD distribution was compared for each pair of category, doing a subtraction between them, totaling 30 comparisons. Resultados e Conclusões We found significant differences in the BOLD distribution for all pairs analyzed, which indicate the feasibility to further perform visual cortex decoding using the protocol described above.