Anupriya Pathania1, Melynda Schreiber2, Matthew W Miller3, Matthew J Euler4, Keith R Lohse5. 1. Department of Health, Kinesiology, and Recreation, University of Utah, United States of America. 2. Mechanical Engineering, University of Utah, United States of America. 3. School of Kinesiology, Auburn University, United States of America; Center for Neuroscience Initiative, Auburn University, United States of America. 4. Department of Psychology, University of Utah, United States of America. 5. Department of Health, Kinesiology, and Recreation, University of Utah, United States of America; Department of Physical Therapy and Athletic Training, University of Utah, United States of America. Electronic address: rehabinformatics@gmail.com.
Abstract
INTRODUCTION: The slope of the electroencephalography (EEG) power spectrum (also referred to as 1/f noise) is an important consideration when calculating narrow-band power. However, psychophysiological data also suggest this slope is a meaningful signal itself, not merely background activity or noise. We present two different methods for quantifying the slope of the power spectrum and assess their reliability and sensitivity. METHODS: We used data from N = 60 participants who had EEG collected during rest, a videogame task, and a second period of rest. At all phases of the experiment, we calculated the "spectral slope" (a regression-based method fit to all datapoints) and the "aperiodic slope" (estimated with the fitting oscillations with 1/f algorithm FOOOF). For both methods we assessed: their reliability, their sensitivity to the transition from rest to task, their sensitivity to changes during the videogame task itself, and the agreement between the two measures. RESULTS: Across resting phases, both spectral and aperiodic slopes showed a high degree of reliability. Both methods also showed a steepening of the power spectrum on-task compared to rest. There was also a high degree of consistency between the two methods in their estimate of the underlying slope, but FOOOF explained more variance in the power spectra across regions and type of activity (rest versus task). CONCLUSION: The slope of the power spectrum is a highly reliable individual difference and sensitive to within-subject changes across two different methods of estimation. Moving forward, we generally recommend the use of the FOOOF algorithm for its ability to account for narrow-band signals, but these data show how regression-based approaches produce similar estimates of the spectral slope, which may be useful in some applications.
INTRODUCTION: The slope of the electroencephalography (EEG) power spectrum (also referred to as 1/f noise) is an important consideration when calculating narrow-band power. However, psychophysiological data also suggest this slope is a meaningful signal itself, not merely background activity or noise. We present two different methods for quantifying the slope of the power spectrum and assess their reliability and sensitivity. METHODS: We used data from N = 60 participants who had EEG collected during rest, a videogame task, and a second period of rest. At all phases of the experiment, we calculated the "spectral slope" (a regression-based method fit to all datapoints) and the "aperiodic slope" (estimated with the fitting oscillations with 1/f algorithm FOOOF). For both methods we assessed: their reliability, their sensitivity to the transition from rest to task, their sensitivity to changes during the videogame task itself, and the agreement between the two measures. RESULTS: Across resting phases, both spectral and aperiodic slopes showed a high degree of reliability. Both methods also showed a steepening of the power spectrum on-task compared to rest. There was also a high degree of consistency between the two methods in their estimate of the underlying slope, but FOOOF explained more variance in the power spectra across regions and type of activity (rest versus task). CONCLUSION: The slope of the power spectrum is a highly reliable individual difference and sensitive to within-subject changes across two different methods of estimation. Moving forward, we generally recommend the use of the FOOOF algorithm for its ability to account for narrow-band signals, but these data show how regression-based approaches produce similar estimates of the spectral slope, which may be useful in some applications.
Authors: Viktoriya O Manyukhina; Andrey O Prokofyev; Ilia A Galuta; Dzerassa E Goiaeva; Tatiana S Obukhova; Justin F Schneiderman; Dmitrii I Altukhov; Tatiana A Stroganova; Elena V Orekhova Journal: Mol Autism Date: 2022-05-12 Impact factor: 6.476