Frontiers in Neuroinformatics (Nov 2019)

Analysis of Decision-Making Process Using Methods of Quantitative Electroencephalography and Machine Learning Tools

  • Grzegorz M. Wojcik,
  • Jolanta Masiak,
  • Andrzej Kawiak,
  • Lukasz Kwasniewicz,
  • Piotr Schneider,
  • Filip Postepski,
  • Anna Gajos-Balinska

DOI
https://doi.org/10.3389/fninf.2019.00073
Journal volume & issue
Vol. 13

Abstract

Read online

The electroencephalographic activity of particular brain areas during the decision making process is still little known. This paper presents results of experiments on the group of 30 patients with a wide range of psychiatric disorders and 41 members of the control group. All subjects were performing the Iowa Gambling Task that is often used for decision process investigations. The electroencephalographical activity of participants was recorded using the dense array amplifier. The most frequently active Brodmann Areas were estimated by means of the photogrammetry techniques and source localization algorithms. The analysis was conducted in the full frequency as well as in alpha, beta, gamma, delta, and theta bands. Next the mean electric charge flowing through each of the most frequently active areas and for each frequency band was calculated. The comparison of the results obtained for the subjects and the control groups is presented. The difference in activity of the selected Brodmann Areas can be observed in all variants of the task. The hyperactivity of amygdala is found in both the patients and the control group. It is noted that the somatosensory association cortex, dorsolateral prefrontal cortex, and primary visual cortex play an important role in the decision-making process as well. Some of our results confirm the previous findings in the fMRI experiments. In addition, the results of the electroencephalographic analysis in the broadband as well as in specific frequency bands were used as inputs to several machine learning classifiers built in Azure Machine Learning environment. Comparison of classifiers' efficiency is presented to some extent and finding the most effective classifier may be important for planning research strategy toward finding decision-making biomarkers in cortical activity for both healthy people and those suffering from psychiatric disorders.

Keywords