Dynamic network analysis of electrophysiological task data

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Chetan Gohil¹ (chetan.gohil@psych.ox.ac.uk), Oliver Kohl¹, Rukuang Huang¹, Mats WJ van Es¹, Oiwi Parker Jones¹, Laurence Hunt¹, Andrew J Quinn², Mark W Woolrich¹; ¹University of Oxford, ²University of Birmingham

A popular approach for studying cognition is to use functional neuroimaging combined with a task. In electrophysiological data, we often study the oscillatory task response by averaging the time-frequency response epoched around the cognitive event of interest over trials. Whilst effective, the researcher must decide a trade-off between the time and frequency resolution, and sensors/brain regions of interest. Here, we show how the oscillatory task responses from conventional time-frequency approaches can be represented more parsimoniously at the network level using two state-of-the-art methods: the HMM (Hidden Markov Model) and DyNeMo (Dynamic Network Modes). In a face perception task recorded with MEG (N=19), where participants were shown an image of a famous, unfamiliar, and scrambled face. Comparing DyNeMo, HMM and traditional oscillatory response analysis, we show only DyNeMo is able to detect subtle transient differences in the oscillatory network response for famous vs unfamiliar faces. We observe a sequence of network activations at millisecond speeds that relate to bottom-up and top-down processes. Our results suggest the recognition of famous faces relies on the engagement of a prefrontal network and a suppression of the visual network. We argue DyNeMo offers a more sensitive and interpretable network perspective on the cognitive processes that are revealed in functional neuroimaging studies compared to conventional methods.

Topic Area: METHODS: Electrophysiology