Linear parameter-varying model and adaptive filtering technique for detecting neuronal activities: an fNIRS study.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Functional near-infrared spectroscopy (fNIRS) is a non-invasive brain imaging technique using near-infrared light.
• fNIRS offers advantages like low cost, portability, and good temporal resolution for real-time imaging.
• fNIRS shows significant potential as a tool for brain-computer interfaces.

Purpose of the Study:

• To present a novel technique for fNIRS-based modeling of brain activities.
• To utilize the linear parameter-varying (LPV) method and adaptive signal processing for improved brain activity analysis.
• To enhance the capabilities of fNIRS for brain-computer interface applications.

Main Methods:

• Developed a novel fNIRS modeling technique using the linear parameter-varying (LPV) method.
• Employed adaptive signal processing, specifically the affine projection algorithm, to estimate unknown coefficients of the LPV system.
• Assumed a Gaussian distribution for the parameter vector within the LPV model.

Main Results:

• The proposed LPV model offers greater efficiency compared to the General Linear Model (GLM) by allowing more defined states.
• The model operates in an online fashion, unlike many previous approaches.
• Experimental validation using random finger-tapping tasks demonstrated improved results, with 24 states utilized.

Conclusions:

• The proposed technique, using t-statistics for activation maps, shows improved estimation of hemodynamic responses compared to existing GLM-based algorithms.
• The algorithm's convergence is evidenced by error reduction over consecutive iterations.
• This advancement holds promise for more effective fNIRS-based brain-computer interfaces.