Low-dimensional chaos maps learning in a model neuropil (olfactory bulb)

Area of Science:

• Computational Neuroscience
• Nonlinear Dynamics
• Olfactory System Research

Background:

• Quantifying chaotic systems often uses correlation dimension (D2), but biological data's non-stationarity limits its application.
• Traditional D2 algorithms require data stationarity, a condition rarely met by biological signals.
• Previous work introduced the point correlation dimension (PD2) to track D2 in dynamic data.

Purpose of the Study:

• To mathematically validate the point correlation dimension (PD2) algorithm for stationary data.
• To demonstrate PD2's ability to reject noise and accurately assess biological data.
• To investigate neural dynamics in the olfactory bulb using PD2 before and after odor stimulation.

Main Methods:

• Developed and mathematically argued for the point correlation dimension (PD2) algorithm.
• Recorded olfactory bulb activity from conscious rabbits using 64 electrodes at 640 Hz.
• Applied both D2 and PD2 algorithms to analyze neural responses to novel and habituated odors.

Main Results:

• PD2 was calculable in 100% of trials, whereas D2 was only successful in 10% of novel-odor trials.
• Both algorithms showed a uniform dimensional increase with novel odors.
• PD2 uniquely identified dimensional decreases during inspiration and spatial gradients in control states, which changed with odor habituation.

Conclusions:

• The PD2 algorithm is sensitive, accurate, and suitable for dimensional analysis of biological data.
• Analysis of unfamiliar olfactory information transiently evokes a single global process in the neuropil.
• Odor experience leads to multiple, spatially specific processes that map learning sites.