Real-time estimation of perceptual thresholds based on the electroencephalogram using a deep neural network

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Perceptual thresholds assess nervous system function in vision, hearing, touch, and pain.
• Current methods rely on subjective participant responses to stimuli, requiring accuracy and consistency.
• Estimating perceptual thresholds is crucial for scientific and clinical evaluations.

Purpose of the Study:

• To demonstrate the feasibility of measuring nociceptive perception thresholds using only non-invasively recorded brain activity.
• To develop a novel method for objective perceptual threshold estimation.
• To overcome limitations of traditional psychophysical methods dependent on subject reporting.

Main Methods:

• A deep neural network was trained to perform a 2-interval forced choice procedure on electroencephalogram (EEG) data.
• The network identified post-stimulus brain activity within recorded EEG signals.
• Perceptual thresholds were estimated in real-time using a psychophysical method of limits based on network classifications.

Main Results:

• The deep neural network accurately matched human participants' perception reports.
• Network-estimated perceptual thresholds aligned with those derived from participant responses.
• The neural network reliably distinguished brain responses from non-responses and estimated thresholds during various tasks.

Conclusions:

• Deep neural networks can accurately predict stimulus perception by analyzing non-invasively recorded brain activity.
• Real-time estimation of perceptual thresholds is achievable without requiring verbal or motor responses.
• This approach offers an objective and efficient method for assessing sensory perception, particularly for nociception.