Mental Workload Classification and Tasks Detection in Multitasking: Deep Learning Insights from EEG Study.

Main Methods:

• Conducted an EEG experiment with 50 participants performing the NASA Multi-Attribute Task Battery II (MATB-II) under four distinct task load levels.
• Developed and applied a convolutional neural network (CNN) for two classification tasks: differentiating task load levels and detecting active subtasks.
• Analyzed EEG segments to assess the CNN's ability to distinguish between different cognitive demands and task quantities.

Main Results:

• The CNN successfully classified EEG segments corresponding to the presence of individual MATB-II subtasks, indicating sensitivity to task-specific neural patterns.
• The CNN struggled to differentiate between the two highest task load levels, suggesting EEG patterns were not sufficiently distinct for higher MWL discrimination.
• Low error rates across participants indicated effective adaptation to increased task demands, potentially masking workload-related EEG variations.

Conclusions:

• EEG, even when analyzed with deep learning, may not provide sufficiently distinct patterns to differentiate higher levels of task load in multitasking scenarios.
• The experimental design, where higher task loads differed primarily in quantity, and participant adaptation, may limit the detectability of MWL-related EEG changes.
• Future research should consider alternative neuroimaging techniques or experimental paradigms to better capture nuanced differences in mental workload during complex tasks.