Software tools for assessing and enhancing quality of fNIRS experimental measurements
Luca Pollonini & Samuel Montero Hernandez, University of Houston
Duration: 90 min
Optional: Laptop; Matlab pre-installed
Synopses: This mini-course focuses on collection and verification of high-quality fNIRS signals prior to data analysis. In particular, it describes a quantitative method for assessing optode-scalp coupling and presence of movement artifacts in all optical channels and time samples of an fNIRS dataset. Notably, this approach is based on a combination of time- and frequency-domain measures of the physiological, systemic pulsation present in any raw fNIRS signals, and therefore it is applicable to datasets acquired with any fNIRS device. Of practical relevance, it has been implemented in two MATLAB applications freely available to the fNIRS community: (1) PHOEBE (Placing Headgear Optodes Efficiently Before Experimentation), a graphical tool that displays the scalp coupling of optodes in real time to facilitate the optimal placement of a fNIRS headgear, akin to electrical impedance used in electroencephalography, and (2) NIRSplot, a tool for post-hoc analysis of fNIRS datasets that informs about the data quality of the entire experiment by automatically identifying and intuitively displaying time period and optical channels with low quality data (due to artifacts or other contaminants) and, optionally, it restores compromised fNIRS signals to improve further analysis. Demonstrations of both software tools will be provided using datasets collected with different fNIRS devices. This mini-course welcomes new fNIRS researchers from different backgrounds seeking to strengthen their data collection and assessment skills, and it encourages all users to contribute ideas towards establishing a consensus on fNIRS signal quality control.
Rationale: Functional near-infrared spectroscopy (fNIRS) is an ever-growing optical technique that has seen a proliferation of new instruments addressing the research interests of scientists and clinicians alike. However, the assessment of fNIRS data quality and the comparison between data collected with different instruments remain challenging due to the lack of a standard method that defines and quantifies the signal-to-noise ratio (SNR) of an fNIRS signal. In addition, movement artifacts may temporarily compromise the quality of otherwise clean fNIRS signals and therefore may require the rejection of experimental trials with unrecoverable signals. This mini-course seeks to bring together new and seasoned fNIRS investigators to promote and further develop practical, easily interpretable methods for assessing fNIRS data quality. Specifically, it targets researchers with limited exposure to technical fNIRS training other than those provided by specific device manufacturers.
Course structure: At the course outset, attendants will learn about the typical features of high- vs. low-quality fNIRS signals, especially how movement artifacts may affect recordings (estimated course time: ~10 mins). Then, they will learn how physiological components, such as systemic oscillations otherwise deemed undesirable for cortical functional analysis, can be used as robust indicators of fNIRS signal quality (estimated course time: ~20 mins). Finally, ample time (estimated course time: ~60 mins) will be dedicated to demonstrate software tools for acquisition and assessment of fNIRS data quality. First, we will offer a quick review, and possibly live-demonstrate, PHOEBE for use before or during the experiment. Subsequently, we will describe and demonstrate NIRSplot for data assessment after the experiment using datasets collected with different fNIRS devices (NIRx, Shimadzu, Obelab, Techen). Time permitting, the mini course will conclude with a public Q&A session (~15 mins), although this could be easily replaced by offline discussions.
Learning objectives: Attendants will learn about the typical features of high- vs. low-quality fNIRS signals, especially how movement artifacts may affect recordings. Then, they will learn how physiological components, such as systemic oscillations otherwise deemed undesirable for cortical functional analysis, can be used as robust indicators of fNIRS signal quality Finally, ample time will be dedicated to demonstrate software tools for acquisition and assessment of fNIRS data quality.
Requirements: if you want to follow along you can bring a laptop with Matlab installed. https://bitbucket.org/lpollonini/phoebe/wiki/Home