Retina Layout and Photoreceptors

In the past two weeks, I’ve focused on replicating the retina with appropriate modeling and mechanisms to mirror the human eye - specifically the retinal layout and individual photoreceptor “cell” functionality. Please find more on that below -

Development Activity - https://github.com/akhil-reddy/beads/graphs/commit-activity

Please note that some code (class templates, function comments, etc) is AI generated, so that I spend more of my productive time thinking and designing. However, I cross-verify each block of generated code with its corresponding design choice before moving ahead.

Algorithms / Development

Push implementation and transportation

1. Setup the two-dimensional retinal layout / manifold with proportions closely mirroring the human eye
   • The surface is made up of closely tied hexagonal structures without any gaps in between them
   • At the center of this (approximately) rectangular surface, a circle of a certain radius represents the fovea
   • Inside the fovea, each hexagon is divided into three equal parallelograms which represent individual cones
   • Outside the fovea, each hexagon is divided into six equal triangles representing the rods
2. Setup the mechanisms and components for cones
   • High activation threshold (around 100 photon–equivalents) so that low-intensity noise (many false negatives) is ignored
   • They have a nearly one-to-one mapping onto bipolar cells i.e., Push implementation
   • They show distinct spectral sensitivity (using a narrow Govardovskii nomogram). Imagine a friend who considers red and crimson the same colour and someone with better color sensitivity. This difference is primarily due to the stimulus registration in cones
   • They SHOULD have faster activation and deactivation kinetics
3. Setup the mechanisms and components for rods
   • Very low activation threshold (near 1 photon–equivalent) so that even dim light is detected (many false positives)
   • They converge many-to-one onto bipolar cells i.e., Push implementation
   • They use a broader Govardovskii nomogram (flatter, overlapping curve) for spectral sensitivity
   • They integrate signals over multiple iterations (to reduce false positives from noise). They SHOULD have slower activation and deactivation kinetics. This slower response aids in integrating photon signals over time, boosting sensitivity but sacrificing temporal resolution. In the biological retina, this is done to ensure that the single photon from the scenery picked up by Rh* is not noise / a stray from the environment. HOWEVER, as modern digital sensor already incorporate this concept for low light vision, we don't need to implement it YET
4. Collect the photoreceptor cell objects and store them in memory

Next Steps

Push implementation and transportation

1. Overlaying frames from a video onto the retina
2. Horizontal cell implementation
3. Push cell implementation
4. Neurotransmitters
5. Transportation

Reading

I was reading through a scientific journal paper on why cones are different from rods and I was wondering why biology has many complex (and sometimes complicated) mechanisms for relatively simple functions. Here are my thoughts -

1. To tackle cellular noise - Visual (and other sensory) signals have to be amplified in rods, for example, to counter intrinsic cellular noise. While I understand this rationale for biological systems, the amplifications serve little purpose in digital models
2. Chemical reaction regulation - The advantages of digital models are that they’re abstract and exact (sometimes, way too exact without flexibility). Precise representation and processing is a boon in digital systems (although the underlying hardware has its own physical complexities). However, nature relies on chemical reactions for processing and storing information which brings the need for many error correction mechanisms
3. Natural evolution is slow and incremental, so many complexities come from millions of upgrades to protein expressions. These complexities bring unnecessary complications as well