Format

Send to

Choose Destination
Behav Brain Res. 2015 Sep 1;290:131-42. doi: 10.1016/j.bbr.2015.04.053. Epub 2015 May 8.

Single-trial classification of near-infrared spectroscopy signals arising from multiple cortical regions.

Author information

1
Bloorview Research Institute, Holland Bloorview Kids Rehabilitation Hospital, 150 Kilgour Road, Toronto, Ontario, M4G 1R8, Canada; Institute of Biomaterials and Biomedical Engineering, University of Toronto, 164 College Street, Toronto, Ontario, M5S 3G9, Canada.
2
Bloorview Research Institute, Holland Bloorview Kids Rehabilitation Hospital, 150 Kilgour Road, Toronto, Ontario, M4G 1R8, Canada; Institute of Biomaterials and Biomedical Engineering, University of Toronto, 164 College Street, Toronto, Ontario, M5S 3G9, Canada. Electronic address: tom.chau@utoronto.ca.

Abstract

Near-infrared spectroscopy (NIRS) brain-computer interface (BCI) studies have primarily made use of measurements taken from a single cortical area. In particular, the anterior prefrontal cortex has been the key area used for detecting higher-level cognitive task performance. However, mental task execution typically requires coordination between several, spatially-distributed brain regions. We investigated the value of expanding the area of interrogation to include NIRS measurements from both the prefrontal and parietal cortices to decode mental states. Hemodynamic activity was monitored at 46 locations over the prefrontal and parietal cortices using a continuous-wave near-infrared spectrometer while 11 able-bodied adults rested or performed either the verbal fluency task (VFT) or Stroop task. Offline classification was performed for the three possible binary problems using 25 iterations of bagging with a linear discriminant base classifier. Classifiers were trained on a 10 dimensional feature set. When all 46 measurement locations were considered for classification, average accuracies of 80.4±7.0%, 82.4±7.6%, and 82.8±5.9% in differentiating VFT vs rest, Stroop vs rest and VFT vs Stroop, respectively, were obtained. Relative to using measurements from the anterior PFC alone, an overall average improvement of 11.3% was achieved. Utilizing NIRS measurements from the prefrontal and parietal cortices can be of value in classifying mental states involving working memory and attention. NIRS-BCI accuracies may be improved by incorporating measurements from several, distinct cortical regions, rather than a single area alone. Further development of an NIRS-BCI supporting combinations of VFT, Stroop task and rest states is also warranted.

KEYWORDS:

Attention; Brain–computer interface; Near-infrared spectroscopy; Parietal cortex; Prefrontal cortex; Working memory

PMID:
25960315
DOI:
10.1016/j.bbr.2015.04.053
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center