A Multivariate Functional Connectivity Approach to Mapping Brain Networks and Imputing Neural Activity in Mice

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Traditional functional connectivity (FC) analysis using Pearson correlation is widely used to study brain organization but is susceptible to physiological noise.
• Confounding factors like breathing, motion, and anesthesia can distort FC measurements, complicating interpretation in both healthy and diseased states.
• A multivariate approach could offer improved accuracy and robustness in mapping neural networks.

Purpose of the Study:

• To develop and validate a multivariate functional connectivity (MFC) method for analyzing spontaneous brain activity.
• To compare the performance of MFC against traditional FC in detecting neural network alterations under various physiological conditions and after stroke.

Main Methods:

• Analyzed neural calcium imaging data from Thy1-GCaMP6f mice under different states (awake, asleep, anesthetized, motion) and after stroke.
• Employed a linear support vector regression model to calculate optimal weights for predicting neural activity in a region of interest (ROI) from surrounding signals.
• Interpreted the resulting weight maps as multivariate functional connectivity (MFC).

Main Results:

• MFC maps showed a sparser, more focused set of strong positive connections compared to traditional FC, resembling anatomical connectivity.
• MFC analysis demonstrated greater resilience to global data variations that significantly impact standard FC.
• MFC analysis was more effective than traditional FC in detecting connectivity deficits after photothrombotic stroke.

Conclusions:

• Multivariate functional connectivity (MFC) provides a more robust and accurate method for mapping brain functional networks than traditional bivariate FC analysis.
• MFC is less confounded by physiological noise and offers superior sensitivity for detecting changes in brain connectivity, particularly in disease models.
• This approach advances our understanding of functional connectivity and its alterations in neurological conditions.