Abstract
INTRODUCTION: This paper describes an experimental work using machine learning (ML) as a "decoding for interpretation" to understand the brain's physiology better. METHODS: Multivariate pattern analysis (MVPA) was used to decode the patterns of event-related potentials (ERPs, brain responses to stimuli) in a visual oddball task. The ERPs were measured before (run 1) and after (30 min-run 2, 90 min-run 3) a single dose of an energy dietary supplement with only a small amount of caffeine. RESULTS: Its effect on ERPs was successfully decoded. Above-chance decoding accuracies were obtained between ∼350 and 450 ms (corresponds to P3 peak) after stimulus onset for both the placebo and study groups, whereas between ∼200 and 260 ms (corresponds to P2 waveform) only in the placebo group. Moreover, the decoding accuracies were significantly higher in the placebo than in the study group in the 200-250 ms and 450-500 ms time bins. Our previously reported findings showed an increase in P3 amplitude among the runs only in the placebo group, indicating a reduction of mental fatigue caused by the supplementation. DISCUSSION: Thus, this paper extends these results, showing that the dietary supplement affected the brain's neural activity related to the attention-related processing of the visual stimuli in the oddball task already at the early processing stage. This implies that inhibiting the fatigue-related brain changes after only a single dose of a dietary neurostimulant acts on early and late processing stages. This emphasizes the value of decoding for interpretation in ERP research. The results also point out the necessity of controlling the uptake of dietary supplements before the neurophysiological examinations.