Visual cortex speckle imaging for shape recognition

Area of Science:

• Neuroscience
• Biophotonics
• Artificial Intelligence

Background:

• Non-invasive methods for monitoring brain activity are crucial for understanding visual processing.
• Laser-speckle patterns offer a potential window into neural dynamics within the visual cortex.

Purpose of the Study:

• To introduce and validate a non-invasive method for classifying geometric shapes based on laser-speckle patterns from the human striate cortex.
• To assess the efficacy of deep neural networks (DNNs) in decoding these patterns for shape recognition.

Main Methods:

• Utilized a fast digital camera to capture laser-speckle patterns reflected from the striate cortex during visual stimulation.
• Employed optimized deep neural networks (DNNs) for the classification of distinct speckle patterns corresponding to different shapes.
• Tested classification accuracy for single shapes (rectangles, triangles) and simultaneous presentation of multiple shapes.

Main Results:

• Distinct laser-speckle patterns were reliably detected for visual stimuli like rectangles and triangles.
• The DNN classifier achieved high recall rates: 98% for rectangles and 91% for triangles in single-shape trials.
• Robust performance (82% recall) was maintained when multiple shapes were presented simultaneously.
• Circular stimuli yielded less distinct patterns, resulting in lower classification accuracy.

Conclusions:

• This non-invasive neurovisual classification technique effectively decodes visual shape information from striate cortex laser-speckle patterns.
• The approach, combining low-cost optics and AI, has significant potential for real-time, portable monitoring of visual cortex activity.
• Applications include cognitive neuroscience research, brain-machine interfaces, and clinical assessment of visual processing disorders.