Prediction of brain regions from single channel ECoG signals by deep learning
Ryosuke Negi, Tsukuba Univesity, Japan; Masaru Kuwabara, Ryota Kanai, Araya, Inc, Japan
Posters 3 Poster
Pacific Ballroom H-O
Sat, 27 Aug, 19:30 - 21:30 Pacific Time (UTC -8)
In Mountcastle's "The Mindful Brain" , it is argued that every part of the neocortex shares a common computational principle. In this study, we challenge this view by postulating that each cortical region has its own specific algorithm and latent features characteristics to the region. More specifically, we hypothesized that if each region functions differently, there should be distinct temporal patterns for each brain region. That is to say, if cortical regions operate differently depending on the area, those differences should be reflected in the the ECoG signals. To test this idea, We trained a deep learning model to predict the region of an ECoG electrode from single channel ECoG data. Furthermore, we probed similarities across brain regions based on the latent features of the ECoG signals. We found that our deep learning model can classify the position of the electrode at 51% of accuracy (the chance level was 14%).This suggests that ECoG signals from single channels contain characteristic temporal signatures specific to the region, supporting the hypothesis that each cortical region has a distinct functional feature as opposed to the notion of uniform functional structures across regions.