Estimating fMRI timescale maps

Area of Science:

• Neuroscience
• Computational Neuroscience
• Data Science

Background:

• Brain activity occurs across hierarchical timescales, crucial for information processing.
• Current methods for estimating brain timescales in fMRI have limitations, including restrictive assumptions and lack of statistical inference.

Purpose of the Study:

• To develop and evaluate novel methods for mapping brain timescales in resting-state fMRI.
• To overcome limitations of existing methods by relaxing assumptions and enabling statistical inference.

Main Methods:

• Formalized and evaluated two methods: time-domain autoregressive (AR1) model fitting and autocorrelation-domain exponential decay model fitting.
• Defined timescales by projecting fMRI time series onto approximating models, requiring only stationarity and mixing conditions.
• Incorporated robust standard errors to account for model misspecification and provided theoretical properties of estimators.

Main Results:

• The time-domain method demonstrated more accurate estimates, especially under model misspecification.
• This method is computationally efficient for high-dimensional fMRI data.
• Results yielded timescale maps consistent with known functional brain organization.

Conclusions:

• The study successfully demonstrates valid statistical inference on fMRI timescale maps.
• Introduced novel, more robust methods for analyzing brain dynamics.
• Provided Python implementations for broader research application.