고려대학교 뇌공학과 주관의 첨단 뇌공학 연구 협의체 및 뇌 신호처리 연구실에서 뇌공학 세미나를 개최합니다.
많은 참석 부탁드립니다.
<aside> 📢
일시: 2026년 9월 14일 (월) 9:30-10:30
장소: 고려대학교 하나스퀘어 멀티미디어룸 </aside>
연자: Yibei Chen, PhD
Post-Doctoral Associate, Massachusetts Institute of Technology, Cambridge, MA, USA
Short bio
Yibei Chen is a postdoctoral associate at MIT's McGovern Institute, working with Dr. Satra Ghosh. She did her Ph.D. in communication neuroscience at University of California, Santa Barbara, master's at Fudan University, and undergraduate at Wuhan University. Her research pairs naturalistic neuroimaging with computational modeling to uncover the recurring "brain states" behind how we experience continuous, real-world events, from watching movies to playing video games.
Introduction
Yibei Chen is a postdoctoral associate at the McGovern Institute for Brain Research at MIT, where she works with Dr. Satra Ghosh. She came to neuroscience from the social sciences, earning her Ph.D. in communication neuroscience at University of California, Santa Barbara. Her research combines naturalistic neuroimaging with computational modeling, for example hidden Markov models, to understand how the brain organizes complex, continuous experience into recurring "brain states," and how those states arise, persist, and transition over time.
Her work uses rich, naturalistic stimuli to characterize these dynamics. In a recent study, pulished on Imaging Neuroscience, 2026, she showed that contextual information reshapes brain-state trajectories during narrative comprehension. In her latest work, also the focus of today's talk, she models brain states in individuals who watched multiple seasons of Friends, asking how state repertoires recur and generalize across diverse narrative contexts over time. A parallel project extends this approach to active, goal-directed experience, using switching dynamical models to characterize brain states while people play a video game (Super Mario).
Looking ahead, Dr. Chen's proposed research aims to turn the temporal structure of these dynamics into quantitative phenotypes that can be compared across individuals, across development, and across interacting brains, with the longer-term goal of identifying markers of psychiatric risk that emerge before clinical symptoms. Alongside her scientific work, she contributes to open and reproducible science, including leading the development of ReproSchema, a standard for machine-readable behavioral and clinical assessment now adopted by large developmental studies.
Title
Brain states recur across diverse narrative contexts during longitudinal viewing
Abstract
Daily experience unfolds continuously, yet cognition remains stable and coherent as sensory input and cognitive demands change. Brain activity may support this balance by drawing on a limited repertoire of persistent patterns that recur across nonadjacent moments. Whether such a repertoire remains stable across hours of varied experience, and whether its momentary expression relates to ongoing content, remains unknown. We analyzed longitudinal fMRI data from six individuals who watched four to six seasons of Friends, providing 39–58 hours of data per participant, and fitted a hidden Markov model separately to each participant. Each model identified approximately 45 recurring brain states. The same states, with similar recurrence tendencies, emerged from different periods of viewing. States differed in the sources of their recurrence and in functional-network composition, while transitions preferentially linked similar states. States that recurred most during Friends also tended to be expressed most during independent films. Momentary state identity was partly decodable from visual, auditory, and language features of the stimulus, and these content–state relationships generalized to varying degrees across new material. Thus, stability across extended experience resided in a reproducible repertoire of recurring states, whereas change resided in their content-sensitive momentary expression. Across experiences, contentstate relationships generalized, but the generalization of the recurrence structure was limited.