A stacked sparse auto-encoder and back propagation network model for sensory event detection via a flexible ECoG

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Current prostheses lack direct sensory feedback, limiting amputee functionality.
• Restoring tactile sensation is challenging due to unnatural sensations from existing methods.
• Understanding somatosensory system input during movement is crucial for effective sensory feedback.

Purpose of the Study:

• To propose an efficient model for detecting sensory events from electrocorticography (ECoG) signals.
• To optimize the model using advanced algorithms for improved performance.
• To validate the model's effectiveness in a rodent model.

Main Methods:

• Utilized a stacked sparse autoencoder and backpropagation neural network for signal detection.
• Employed Particle Swarm Optimization and Adaptive Moment Estimation (Adam) for model optimization.
• Recorded tactile afferent signals from the primary somatosensory cortex (S1) during mechanical stimulation.

Main Results:

• The model achieved high accuracy (98.8%), sensitivity (96.8%), and specificity (99.1%) in detecting sensory events.
• Identified significant differences in tactile sensitivity across different regions of the rat hind-paw.
• Demonstrated the model's superior performance compared to state-of-the-art methods.

Conclusions:

• The proposed AI model effectively detects sensory events from flexible ECoG recordings.
• This technology shows promise for restoring somatosensory feedback in prosthetic limbs.
• Further research could lead to more natural and intuitive sensory experiences for amputees.