Hidden Markov multiple event sequence models: A paradigm for the spatio-temporal analysis of fMRI data

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Functional magnetic resonance imaging (fMRI) analysis faces challenges with hemodynamic response function (HRF) variability, timing, and non-linearity.
• Existing methods often require prior assumptions about activation patterns, limiting unsupervised discovery.
• Accurate brain mapping is crucial for understanding neural processes and neurological disorders.

Purpose of the Study:

• To present a novel, completely unsupervised fMRI brain mapping method.
• To address key challenges in fMRI analysis: HRF variability, event timing, and response non-linearity.
• To directly incorporate spatial and temporal information into activation detection.

Main Methods:

• Developed a method that formulates activation detection as temporal alignment of hemodynamic response onsets (HROs) between voxels and their spatial neighborhoods.
• Utilized hidden Markov multiple event sequence models (HMMESMs), a novel class of hidden Markov models, to solve the multiple event sequence alignment problem.
• Considered both event-related and epoch experimental paradigms.

Main Results:

• The HMMESM-based mapping approach demonstrated superior performance compared to Statistical Parametric Mapping (SPM2) on both real and synthetic fMRI data.
• The method successfully detects brain activation without any prior definition of expected patterns, proving its unsupervised nature.
• Spatial and temporal information were effectively integrated into the core of the activation detection process.

Conclusions:

• The proposed unsupervised HMMESM approach offers a significant advancement in fMRI brain mapping.
• This method overcomes limitations of existing techniques by handling HRF variability, timing, and non-linearity without prior assumptions.
• The findings suggest a more robust and flexible tool for neuroimaging research and clinical applications.