Use of thermodynamic indices to distinguish between good and bad task performers in terms of working-memory load while performing mental arithmetic tasks.

Nonlinear dynamics is widely used in neuroscience to characterize complex systems, including human brain activity derived from electroencephalography (EEG). A reliable and non-invasive measurement of memory load, obtained while performing an analytical task, could minimize cognitive overload and decision-making errors, particularly in occupations requiring a high mental workload.

Publicly available PhysioNet EEG database was analyzed. Based on the number of calculations performed in 4 minutes, 36 subjects were divided into good (21 ± 7.4 calculations) and bad (7 ± 3.6 calculations) counters. Subsequently, linear and nonlinear dynamic features were extracted from their EEG recordings. The Synthetic Minority Over-sampling Technique (SMOTE) algorithm was used to balance the sample size. The top ten prominent features were selected using the random forest algorithm, with the Higuchi’s fractal dimension being the most critical.

The fuzzy entropy, sample entropy, and fractal dimension of the theta and beta bands were lower in the frontal lobes of the good counters; however, the entropy and fractal dimensions in the occipital and temporal lobes were higher.

With increasing cognitive load, the good counters had lower entropy and fractal dimension. This could be construed as a state of increased regularity of brain waves. The brain wave entropy behaves in congruence with the fractal dimension of EEG waves and thus, can be used as a biomarker to identify people with good cognitive capacity.