Vision from Phosphenes: Principles and Requirements for Artificial Optometry

Introduction to Phosphene-Based Vision

  • The Concept of Novel Visual Stimuli: The presentation explores artificial vision through phosphenes, described as vision "but not as we know it."
  • Mechanism of Delivery: Vision is delivered via a grid of stimulating electrodes.
  • Visual Representation: This system creates a pattern composed of light and dark dots.
    • Light Dot (Phosphene): Occurs when an electrode is turned "on."
    • Dark Dot: Occurs when an electrode is turned "off," resulting in no phosphene.

Identifying Objects with Pixelated Vision

  • Recognition Challenges: A central question is what a pattern of phosphenes actually tells the user and how much detail is required for confident identification of objects.
  • Standardized Recognition Set: The transcript lists a series of objects used for testing visual recognition:
    • Dog
    • Tomato
    • Car
    • Cup
    • Pear
    • Horse
    • Apple
    • Cow
  • Levels of Detail (Sampling Density): Identification confidence is tested across various grid resolutions:
    • 8×88 \times 8
    • 16×1616 \times 16
    • 24×2424 \times 24
    • 32×3232 \times 32
    • 48×4848 \times 48
    • 64×6464 \times 64

Sampling Density and Resolution

  • Definition of Sampling Density: This refers to the number of pixels in an array. Each pixel can be binary (on/off) or provide shades of grey.
  • Complexity Progression: The simplest form is 1 pixel, followed by grids such as 2×22 \times 2 or 3×33 \times 3.
  • Operational Constraints: A key factor in developing these devices is determining how many electrodes can be made functional within the array.
  • Retinal Resolution Comparison:
    • Normal Limit: In biological vision, the resolution limit is determined by the spacing of foveal cones.
    • The Resolution Threshold: To distinguish two points, there must be at least one "unstimulated" receptor between two stimulated receptors.
    • Angular Subtence: A foveal receptor has an angular subtence of 1 min arc1\,\text{min arc}, which corresponds to a Visual Acuity (VA) of 6/66/6.

Calculating Visual Acuity for Electrode Arrays

  • Calculation Example: Consider a grid of 2525 electrodes arranged in a 5×55 \times 5 configuration.
  • Projection Area: The array projects to an area of 10∘×10∘10^{\circ} \times 10^{\circ}.
  • Individual Electrode Subtence: In this array, each electrode subtends 2∘2^{\circ}.
  • Conversion to Minutes of Arc:
    • 2∘=120 min2^{\circ} = 120\,\text{min}
  • Calculated Visual Acuity (VA):
    • VA=6/720VA = 6/720
  • logMAR Measurement:
    • logMAR=2.08logMAR = 2.08

Simulation and Functional Requirements

  • The Sophisticated Research Approach: Researchers use "sampled" or "pixelated" text and images to evaluate visual requirements in normally sighted individuals by simulating phosphene patterns.
  • Utility Assessment: If performance with simulated pixelated vision is not functionally useful, then developing the physical prosthetic devices is not considered worthwhile.
  • Mobility Studies: A specific indoor mobility course was used to test the functionality of sampled vision:
    • Room Layout: The room was divided into square blocks measuring 1.4×1.4 m1.4 \times 1.4\,\text{m}.
    • Obstacles: Suspended paper cylinders were used as obstacles. These were 5 cm5\,\text{cm} in diameter and 1.8 m1.8\,\text{m} in length.
    • Experiment Parameters: The course used cloth screens (some open, some closed) to create different paths between specific start gates (Gate 1 or Gate 2) and an end goal (Gate 3).

Optimal Parameters and System Limitations

  • Key Findings for Success:
    • Pixel Density: Optimum performance is achieved with an array of approximately 25×2525 \times 25 (totaling 625625 pixels).
    • Placement: The stimulus should be delivered close to the fovea or the foveal projection within the visual cortex.
    • Field of View (FOV): For effective mobility, a field of view of approximately 30∘30^{\circ} is required.
  • Inherent Limitations of Current Implants:
    • Color: Implants cannot provide color vision.
    • Depth Perception: There is little or no depth perception or Binocular Vision (BV).
    • Localization: If a camera captures the image, it may not align with where the eye is actually pointing. Furthermore, the electrode array might be placed anywhere on the retina, displacing the visual input.
    • Visual Field Constraints:
      • The size of the field sampled depends on the camera's capabilities.
      • The size of the field actually stimulated depends on the physical size of the implant.