BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Poster Flash Talks Group A: Predictive EEG Signatures via Ornstein
 –Uhlenbeck Decomposition and Digital Twins. - Pierre-Olivier Michel (Uni
 versité PSL)
DTSTART:20251202T151500Z
DTEND:20251202T152000Z
UID:TALK241108@talks.cam.ac.uk
DESCRIPTION:EEG rhythms are rarely stationary\; they emerge as brief\, bur
 st-like events whose timing and shape carry information about brain state.
  We developed models to segment transient &nbsp\;patterns&mdash\;especiall
 y &alpha\;-spindles(8&ndash\;12 Hz)&mdash\;as trajectories of a two-dimens
 ional Ornstein&ndash\;Uhlenbeck (OU) process with a stable focus. The appr
 oach yields three interpretable parameters (decay\, intrinsic frequency\, 
 noise) that can be estimated in real time via two complementary routes: (i
 ) global estimators from the signal&rsquo\;s amplitude distribution\, auto
 correlation\, and Hilbert-phase dynamics\; and (ii) a segmentation-driven 
 fit that matches empirical spindle duration/amplitude distributions to OU 
 simulations. &nbsp\;We will show applications to anesthesia monitoring\, w
 here &alpha\;&ndash\;&delta\; dynamics and spindle fragmentation capture t
 ransitions in depth of anesthesia beyond conventional band-power metrics\,
  and discuss how this representation can interface with interpretable fore
 casting (e.g.\, early warnings of &nbsp\;suppressions).
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR