Week 5 - vision

Light enters via pupil; iris controls pupil diameter. Anatomy: cornea, lens focus image onto retina; retina contains photoreceptors; optic nerve = axons of retinal ganglion cells.
Primary refractive components: cornea ~ $42\,\text{diopters}$ and lens up to ~ $20\,\text{diopters}$ .
Normal eye focusing (emmetropia): distant objects form sharp images on retina; accommodation adjusts lens power via ciliary muscles (flat for distance; round for near).
Myopia: eyeball too long or lens too strong; corrected with concave lenses to diverge light for near focus.

Rods: high sensitivity, active in low light (scotopic). Cones: lower sensitivity, active in bright light (photopic); three cone types with different wavelength sensitivities -> color vision.
Phototransduction basics: light absorption by photopigments hyperpolarizes photoreceptors; darkness -> depolarized (around -30 mV) with open cGMP-gated Na+ channels; light reduces cGMP, channels close, hyperpolarization occurs; photoreceptors are non-spiking.
Cone distribution: central retina (fovea) dense cone inner segments; peripheral retina more cones and many rods; central/ peripheral differences support spatial detail vs sensitivity.
Color vision: red (L), green (M), blue (S) cones; color blindness arises when one cone type is missing.

Spatial acuity falls off in the periphery due to lower receptor density and convergence.
Fovea: high acuity due to 1:1 convergence (one foveal cone per midget ganglion cell). Peripheral retina relies on pooling to increase light sensitivity.
Eye movements actively place objects on the fovea to optimize acuity.

Photoreceptors -> bipolar cells -> ganglion cells (output neurons, capable of action potentials; other retinal neurons exist but are not the focus here).
Receptive field (RF): region of visual space where light affects a cell’s membrane potential.
OFF bipolar cells: center-off, surround-on; hyperpolarized by center light; depolarized by surround light; preserve photoreceptor sign; ionotropic glutamate receptors.
ON bipolar cells: center-on, surround-off; depolarized by center light; hyperpolarized by surround light; invert photoreceptor sign; metabotropic receptors.
Receptive fields are typically circular approximations for photoreceptors; center-surround organization is a key feature transmitted through bipolar and ganglion cells.
ON/OFF center-surround bipolar cells feed corresponding ON/OFF center ganglion cells; ganglion cells are spiking.
Parallel circuits: retina contains >12 distinct ganglion cell types, enabling parallel processing (e.g., different pathways from the same photoreceptors).
M-type (parasol) ganglion cells: large, fast responses, motion sensitivity; non-color sensitive; project to magnocellular LGN layers.
P-type (midget) ganglion cells: small, high-acuity, color-sensitive; project to parvocellular LGN layers.

Center-surround antagonism enhances edge detection and contrast; context affects perception (examples shown in class).
Conceptual models explain how multiple RFs (e.g., LGN inputs) converge onto a primary cell to generate a simple cells response in cortex.

Retinal ganglion cells project mainly to the LGN (gateway to cortex) and pretectum (pupil reflex and eye movements).
LGN has six layers, with parallel channels: magnocellular (layers 1-2) and parvocellular (layers 3-6).
Receptive-field properties in LGN resemble those in the retina; layers segregate inputs from contralateral vs. ipsilateral eyes.
Parasol (M) pathway -> magnocellular LGN -> dorsal stream (where/how).
Midget (P) pathway -> parvocellular LGN -> ventral stream (what).

Visual system uses parallel processing starting in retina and LGN and continuing in cortex.
Cortical hierarchy is parallel and distributed, not strictly serial:
- Dorsal (parietal, where/how): MT (motion), MST (complex motion), area involved in spatial processing and action.
- Ventral (temporal, what): V4 (shape and color), IT (object and face recognition).
Some processing is serial (e.g., V1 -> V4 -> IT for complex object recognition) but many computations occur in parallel earlier on.

V1 is retinotopically organized; foveal over-representation is large (more neurons for central vision).
V1 neurons show clustering by receptive field location; orientation tuning is a core feature.
Simple cells: elongated receptive fields with distinct ON and OFF subregions; often monocular; response depends on bar position within RF; orientation-selective.
Complex cells: more homogeneous RFs, less reliance on exact bar position; typically binocular; orientation-selective.
Putative circuit models explain simple and complex cell responses via convergent inputs from LGN.
Orientation maps form columns; neighboring columns tend to prefer similar orientations; orientation map is coarser than retinotopy.

MT (middle temporal): motion processing; neurons direction-tuned; lesions impair motion perception (akinetopsia); stimulation biases motion perception.
IT (inferotemporal): face perception; neurons tuned to facial features; lesions impair face recognition (prosopagnosia); microstimulation biases face perception.

Visual fields are mapped to hemispheres via the optic nerve, chiasm, and tracts; nasal retina fibers cross at the chiasm, temporal retina fibers do not cross.
Each LGN layer represents the contralateral visual field; input from the left visual field goes to right visual cortex, and vice versa.
Images are inverted on the retina due to optics; this inversion is mapped in the cortex with retinotopic organization.

Cone types provide color information; RGB-like separation supports color perception.
Color vision tests rely on cone sensitivity and alignment; color vision deficiencies arise when one cone type is missing or dysfunctional.
Spatial acuity tests include Snellen charts (6/6, 6/60, etc.), grating acuity, and vernier acuity; each has strengths/limitations for research vs. real-world testing.

Snellen chart: line legibility; line 6/6 should subtend approximately $5\,\text{arcmin}$ at the eye.
Grating acuity: threshold for detecting a grating; vernier acuity: threshold for judging alignment differences between lines.
Spatial acuity depends on receptor density and convergence; fovea supports high acuity with low convergence ratios; periphery relies on pooling for light sensitivity.

Active sensing: eye movements place objects on the high-acuity fovea for better perception.
Center-surround processing is a general principle across sensory systems, enabling efficient edge detection and contrast enhancement.
Retinal prostheses and future technologies aim to restore or augment visual function by interfacing with retinal circuits.

Parallel processing emerges early in the retina and LGN and continues in cortical areas.
Two major cortical streams: dorsal (where/how) and ventral (what), with partial serial progression (V1 -> higher areas) and extensive parallelism.
Receptive fields transition from center-surround in retina to orientation-selective in V1, then to feature- and object-selective responses in higher areas.
Visual perception integrates signals across multiple pathways, with center-surround and motion cues contributing to robust scene understanding.