Automated emotion classification in the early stages of cortical processing: An MEG study

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• The human brain differentiates emotional categories rapidly, within 100-300 milliseconds post-stimulus.
• Previous research has not focused on classifying emotions this early in neural processing.

Purpose of the Study:

• To investigate the possibility of automatically classifying human emotions within the first 300 milliseconds after stimulus presentation.
• To identify the specific time-interval within 300 ms that yields the highest classification performance for emotional stimuli.

Main Methods:

• Magnetoencephalography (MEG) signals were recorded from 17 healthy volunteers exposed to pleasant, unpleasant, and neutral images.
• Six Linear Discriminant Analysis (LDA) classifiers were employed, utilizing feature selection via Genetic Algorithm and LDA.
• Classifications were performed on distinct time-intervals: 100-150 ms, 150-200 ms, and 200-300 ms post-stimulus.

Main Results:

• Significant classification performance was achieved for both pleasant vs. neutral and unpleasant vs. neutral comparisons.
• The highest classification performance (median AUC of 0.83) was observed for unpleasant vs. neutral stimuli within the 100-150 ms interval.
• This represents a classification at least 850 ms earlier than previously reported methods.

Conclusions:

• Brain emotional responses can be reliably classified within 300 ms of stimulus onset using the developed algorithm.
• Emotional processing in the human brain appears to occur significantly within the initial 100-150 ms post-stimulus.