A Machine Learning-Based Decoder Framework for the Cortical Voltage-Sensitive Dye Responses to Retinal Neuromorphic...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Vision

Background:

• Intracortical microstimulation (ICMS) shows promise for visual prostheses.
• Neuromorphic spike trains derived from images are proposed for ICMS pulse sequences.
• Cortical voltage-sensitive dye (VSD) responses to ICMS require further investigation for image information content.

Purpose of the Study:

• To explore the feasibility of machine-learning-based decoding of image information from cortical responses to ICMS.
• To address the lack of large-scale experimental datasets linking visual images, spike trains, and cortical responses.

Main Methods:

• Generated surrogate data using a Wiener-system model to simulate VSD responses to ICMS pulse trains.
• Trained a convolutional neural network (CNN) on synthetic datasets.
• Evaluated the CNN's ability to reconstruct images from simulated cortical responses.

Main Results:

• A CNN successfully reconstructed images from simulated VSD responses.
• The study demonstrated computational feasibility of decoding visual information from simulated cortical activity.
• The findings provide a foundation for future physiological validation.

Conclusions:

• Machine learning-based decoding of visual information from simulated cortical activity evoked by neuromorphic ICMS is computationally feasible.
• This proof-of-concept study supports the potential of using neuromorphic ICMS for visual prostheses.
• Further research is needed for physiological validation.