Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Feb 14:16:787576.
doi: 10.3389/fnhum.2022.787576. eCollection 2022.

Combining Neural and Behavioral Measures Enhances Adaptive Training

Md Lutfor Rahman  1   2   3 , Benjamin T Files  2 , Ashley H Oiknine  4 , Kimberly A Pollard  2 , Peter Khooshabeh  2 , Chengyu Song  3 , Antony D Passaro  2
Affiliations

Combining Neural and Behavioral Measures Enhances Adaptive Training

Md Lutfor Rahman et al. Front Hum Neurosci. .

Abstract

Adaptive training adjusts a training task with the goal of improving learning outcomes. Adaptive training has been shown to improve human performance in attention, working memory capacity, and motor control tasks. Additionally, correlations have been observed between neural EEG spectral features (4-13 Hz) and the performance of some cognitive tasks. This relationship suggests some EEG features may be useful in adaptive training regimens. Here, we anticipated that adding a neural measure into a behavioral-based adaptive training system would improve human performance on a subsequent transfer task. We designed, developed, and conducted a between-subjects study of 44 participants comparing three training regimens: Single Item Fixed Difficulty (SIFD), Behaviorally Adaptive Training (BAT), and Combined Adaptive Training (CAT) using both behavioral and EEG measures. Results showed a statistically significant transfer task performance advantage of the CAT-based system relative to SIFD and BAT systems of 6 and 9 percentage points, respectively. Our research shows a promising pathway for designing closed-loop BCI systems based on both users' behavioral performance and neural signals for augmenting human performance.

Keywords: EEG; adaptive training; behavioral adaptive training; brain-computer interface; combined adaptive training; electroencephalography; theta-alpha ratio; theta-alpha ratio percentage (TARP).

PubMed Disclaimer

Conflict of interest statement

AO was employed by DCS Corporation. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
Method. (A) A flow chart of the training experiment. (B) Stimuli used in the go/no-go training task. (C) Trained (left column) and untrained (right column) threat (top row) and non-threat (bottom row) stimuli in the transfer task. (D) A screenshot from the transfer task.
Figure 2
Figure 2
Average transfer task performance for the three training conditions. The combined adaptive training (CAT) resulted in better transfer task performance compared to behavior adaptive training (BAT) and fixed difficulty (SIFD) conditions. The large dots show the sample average, and the small dots show individual data points. Error bars show 95% confidence intervals from bias corrected, accelerated percentile bootstrap.
Figure 3
Figure 3
Scalp maps of online and offline data. The upper row shows scalp maps (one dot per electrode) for theta-band power, alpha-band power, and TAR (left-to-right, respectively) from the original data. The lower row shows the same maps computed from the offline data after artifact rejection. Note the differences in scale.
Figure 4
Figure 4
Differences in training difficulty do not account for the transfer performance advantage of CAT. Here, we excluded SIFD condition as it was fixed at easiest level. Dashed lines show 95% confidence intervals of the fit.
Figure 5
Figure 5
Differences in training task scores do not account for the transfer performance advantage of CAT. Curved lines show 95% confidence intervals of the fit.
See this image and copyright information in PMC

References

    1. Ablin P., Cardoso J.-F., Gramfort A. (2018). Faster independent component analysis by preconditioning with hessian approximations. IEEE Trans. Signal Process. 66, 4040–4049. 10.1109/TSP.2018.2844203 - DOI - PubMed
    1. Anania J. (1982). The Effects of Quality of Instruction on the Cognitive and Affective Learning of Students. (Doctoral dissertation: ). The University of Chicago.
    1. Benjamini Y., Hochberg Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Stat Soc Series B (Methodol). 57, 289–300.
    1. Bloom B. S. (1984). The 2 sigma problem: the search for methods of group instruction as effective as one-to-one tutoring. Educ. Res. 13, 4–16.
    1. Chen F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Struct Eq Model Multidiscipl J. 14, 464–504. 10.1080/10705510701301834 - DOI

LinkOut - more resources