Probabilistic Cognitive State Modeling (PCSM): Decoding Latent Spatiotemporal Dynamics to Reveal Serial-Parallel...

Area of Science:

• Cognitive Neuroscience
• Neuroimaging Analysis
• Computational Psychiatry

Background:

• Understanding flexible cognitive task transitions and serial-parallel processing is key to human cognition.
• Neuroimaging advances enable linking brain function with cognition, facilitating methods to quantify emergent cognitive properties.

Purpose of the Study:

• To introduce and validate Probabilistic Cognitive State Modeling (PCSM) for quantifying dynamic brain states during task-based fMRI.
• To derive interpretable cognitive properties like processing modes and demand from brain activity.

Main Methods:

• PCSM integrates Finite Impulse Response (FIR) modeling of BOLD signals with Gaussian Mixture Model-Hidden Markov Models (GMM-HMM).
• A ground-truth simulation with varied noise and transition probabilities validated PCSM's accuracy and stability.

Main Results:

• PCSM accurately recovered ground-truth latent states (>98%) and provided stable parameter estimates.
• Threshold analyses reliably distinguished parallel, mixed, and serial processing modes.
• PCSM identified expected relationships between cognitive demand, resource availability, and bottleneck persistence.

Conclusions:

• PCSM effectively reveals dynamic brain states underlying adaptive processing architectures.
• This framework enables mapping individual cognitive dynamics and analyzing cognitive processing, demand, and serial bottlenecks in fMRI studies.