DISCRIMINATION OF MENTAL WORKLOAD LEVELS IN HUMAN SUBJECTS WITH FUNCTIONAL NEAR-INFRARED SPECTROSCOPY

Angelo Sassaroli, Feng Zheng, Leanne Hirshfield, Audrey Girouard, Erin Solovey, Robert J. K. Jacob, Sergio FantiniJournal of Innovative Optical Health Sciences 2008 · DOI

TL;DR

This paper explores using functional near-infrared spectroscopy (fNIRS) to detect different levels of mental workload by measuring hemodynamic changes in the prefrontal cortex. They achieved classification accuracies of 44-72% using a simple 3-nearest neighbor algorithm, significantly outperforming chance, suggesting fNIRS's potential for real-time mental workload assessment.

What Problem Does This Paper Address?

Accurately and non-invasively measuring mental workload in real-time is a significant challenge with implications for human-computer interaction, cognitive load monitoring, and understanding brain function. Traditional methods like fMRI are bulky and expensive, while subjective reports are unreliable. The authors aim to leverage functional near-infrared spectroscopy (fNIRS), a more portable and accessible neuroimaging technique, to differentiate varying levels of cognitive demand, specifically focusing on working memory tasks. They investigate whether hemodynamic responses in the prefrontal cortex, as measured by fNIRS, can serve as reliable indicators of these workload levels.

Methods

The researchers employed a functional near-infrared spectroscopy (fNIRS) system to collect continuous wave (CW) optical data from the foreheads of human subjects. The fNIRS signals, which reflect changes in oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (Hb) concentrations, were translated into these concentration changes using the modified Beer-Lambert law.

The core of their approach involved a 3-nearest neighbor (3-NN) classification algorithm. For each trial, they extracted the maximum amplitude of the HbO and Hb concentration changes during a workload period and a subsequent rest period. This amplitude was then compared to the amplitudes of other trials. A classification was deemed successful if at least two of the three nearest neighbor data points belonged to the same workload category as the query data point. This simple yet effective algorithm was applied to distinguish between different levels of mental workload.

Datasets

The study utilized data collected from three healthy human subjects. These subjects participated in a protocol involving a working memory task designed to elicit different levels of mental workload. The data was collected during these tasks, and no pre-existing public datasets were used or introduced.

Key Insights and Ideas

  • fNIRS as a viable tool for mental workload detection: The study demonstrates that fNIRS can indeed detect and differentiate between varying levels of mental workload, a significant step towards more accessible neuroimaging applications.
  • Hemodynamic changes correlate with cognitive load: The observed changes in HbO and Hb concentrations in the prefrontal cortex are directly linked to the cognitive demands of the working memory tasks, providing a physiological basis for workload assessment.
  • Simplicity can be effective: The use of a straightforward 3-NN classifier, based solely on the amplitude of hemodynamic changes, yielded promising results, suggesting that complex algorithms may not always be necessary for initial discrimination.
  • Deoxyhemoglobin (Hb) might be a stronger indicator: The classification success rates were generally higher when using Hb changes compared to HbO changes, hinting that Hb might be a more sensitive marker for workload-related brain activity in this context.
  • Importance of probe placement: The authors noted that signal quality varied between channels (left vs. right forehead), suggesting that precise probe placement and potential lateralization of brain activity are crucial considerations for fNIRS studies.

Key Results

The study achieved classification success rates ranging from 44.4% to 72.2% across the three subjects and for both HbO and Hb measurements. Specifically:

  • HbO-based classification: Success rates ranged from 44.4% to 50.0%.
  • Hb-based classification: Success rates ranged from 55.6% to 72.2%.

These results are significantly higher than the chance level of 19.1% calculated for a 3-NN classifier with the given data distribution. This indicates a genuine ability to discriminate between workload levels.

Limitations and Future Work

The authors acknowledge several limitations and propose future directions:

  • Limited number of subjects: The study was conducted on only three subjects, necessitating larger-scale studies for broader validation.
  • Simple classification algorithm: While effective, the 3-NN algorithm is basic. More sophisticated machine learning algorithms could potentially improve accuracy.
  • Single parameter for classification: The current approach relies solely on the amplitude of concentration changes. Incorporating additional parameters, such as the temporal dynamics of the signals or data from multiple channels, could enhance discrimination.
  • Motion artifacts: fNIRS is sensitive to motion, and two of the initial five subjects had data that was unusable due to motion artifacts. Future work should focus on developing optical probes that minimize motion effects.
  • Spatial mapping: The study used point recordings. Developing methods for spatial mapping with fNIRS could provide a more comprehensive understanding of brain activity.
  • Application to real-world scenarios: The authors envision applications in human-computer interaction, suggesting future research into real-time, non-invasive mental workload assessment in dynamic environments.

Summary updated

Abstract

We have applied functional near-infrared spectroscopy (fNIRS) to the human forehead to distinguish different levels of mental workload on the basis of hemodynamic changes occurring in the prefrontal cortex. We report data on 3 subjects from a protocol involving 3 mental workload levels based on to working memory tasks. To quantify the potential of fNIRS for mental workload discrimination, we have applied a 3-nearest neighbor classification algorithm based on the amplitude of oxyhemoglobin ( HbO 2 ) and deoxyhemoglobin ( HbR ) concentration changes associated with the working memory tasks. We have found classification success rates in the range of 44–72%, which are significantly higher than the corresponding chance level (for random data) of 19.1%. This work shows the potential of fNIRS for mental workload classification, especially when more parameters (rather than just the amplitude of concentration changes used here) and more sophisticated classification algorithms (rather than the simple 3-nearest neighbor algorithm used here) are considered and optimized for this application.
BibTeX
@article{sassaroli2008discrimination,
  author = {Angelo Sassaroli and Feng Zheng and Leanne Hirshfield and Audrey Girouard and Erin Solovey and Robert J. K. Jacob and Sergio Fantini},
  title = {{DISCRIMINATION OF MENTAL WORKLOAD LEVELS IN HUMAN SUBJECTS WITH FUNCTIONAL NEAR-INFRARED SPECTROSCOPY}},
  journal = {Journal of Innovative Optical Health Sciences},
  year = {2008},
  doi = {10.1142/s1793545808000224},
}
Figure