SYSTEM / NEUROMORPHIC ARCHITECTURE

A brain,
built for the eye.

OCTON's deep AI mirrors the cognitive architecture of the human brain. A fast pattern-recall layer surfaces what's “off” in milliseconds — the way the hippocampus retrieves a memory. A slow, deliberate reasoning layer maps it across twelve retinal pathologies and systemic risk signals — the way the neocortex builds a posterior under uncertainty.

PIPELINE

Two modules.
One retina.

When a fundus image lands in OCTON, both modules fire in parallel. The Hippocampus emits a fast pattern recall and an urgency tier within seconds. The Neocortex then constructs a structured Bayesian posterior across twelve conditions and three systemic risk axes.

MODULE 01HippocampusPATTERN RECALL · 1.8sMODULE 02NeocortexDEEP REASONING · 14sRETINAL INPUTSTRUCTURED FINDINGS12 DISEASE PROBABILITIESEEG · γ-BAND 40-100 HzLIVE
MODULE 01

Hippocampus

OCTON Preliminary Analysis · streaming triage

Pattern-recall over a learned medical library. Surfaces first impression, observed anomalies, urgency tier, and a suggested next step in under three seconds. The hippocampus doesn't reason — it remembers.

> observe(fundus.image)
> recall(memory_dictionary)
> assoc(co_occurring_signs)
> emit(urgency_tier, next_step)
MODULE 02

Neocortex

OCTON In-Depth Analysis · structured reasoning

Bayesian reasoning across twelve conditions and three systemic axes. Emits calibrated probabilities, severities, key findings, and follow-up actions. The neocortex doesn't just recognize — it explains.

> reason(DR | AMD | GL | HR | CT | RD | RVO | RAO | ME | ERM | PAP | ON)
> P(D|I) = P(I|D)·P(D)/P(I)
> quantify(severity, prob, findings)
> emit(structured_findings + systemic_risk)
PRINCIPLES

Wired by Hebb.
Reasoned by Bayes.

Two foundational principles of neural computation power OCTON's deep AI. Association builds memory. Probability builds inference. Together they form a substrate that scales from a single retinal feature to a whole-patient risk profile.

Hebbian association

Donald Hebb's 1949 postulate: “when an axon of cell A repeatedly takes part in firing cell B, some growth process or metabolic change takes place such that A's efficiency in firing B is increased.”In OCTON, co-occurring retinal signs strengthen the link to a disease pattern.

wij(t+1) = wij(t) + η · xi(t) · xj(t)
where η is the learning rate, xi·xj the co-activation.
  • / Co-occurring features ignite linked memory traces
  • / Pattern recall < 3 seconds
  • / Provides the urgency tier and triage

Bayesian inference

Reverend Thomas Bayes' theorem: belief updates given evidence. OCTON's Neocortex computes a posterior probability for each of twelve conditions — weighing the visual likelihood against population priors and severity gradients.

P(D | I) = P(I | D) · P(D) / P(I)
posterior = likelihood × prior / evidence
  • / Calibrated probabilities, not gut feels
  • / Updates as new evidence arrives
  • / Weighs priors from population data
CORTICAL CORE

120,000,000
cortical columns.

The human neocortex is built from ~120 million cortical columns — repeating microcircuits, each a self-contained vertical stack of six neuron layers. They fire in coordinated waves at gamma frequencies. OCTON's deep reasoning models that scale and tempo.

Cortical columns modeled
1.2 × 10⁸
Mean activation density
~3.4%
Pattern dictionary size
12 conditions · 38 signs
Inference time
~14s
Live Cortical Projection

Beyond the fundus.
Inside your visual cortex.

Every finding traces back to the brain regions it would impact. You and your clinician see not just what the eye saw — but what the brain would lose.

8
pathway
nodes
12
diseases
mapped
<1s
projection
latency
Now mapping
Diabetic retinopathy
R · 85%ON · 65%LG · 50%HC · 30%

Microvascular ischemia stops upstream signal — V1 receives fewer photoreceptor packets.

R85%ON65%OXLG50%V1V4MTHC30%
DR · moderate
1 / 5
PIPELINE

End-to-end flow.

01
Capture

Smartphone live camera or upload.

02
Store

Encrypted object storage.

03
OCTON Analyses

Hippocampus → Neocortex streams to your screen.

04
Report

Severity, findings, follow-up.

PRINCIPLES

Built for screening,
not replacement.

OCTON is a clinical decision-support companion. It exists to raise flags early — at the point of the smartphone, before symptoms. Every finding should be reviewed by a licensed ophthalmologist.

OCTON

A neuromorphic retinal screening engine. Fast pattern triage by the Hippocampus module, deep diagnostic reasoning by the Neocortex. For early detection of diabetic retinopathy, AMD, glaucoma & more.

For early screening — Not a substitute for a licensed ophthalmologist.

System
  • Hippocampus / OCTON Preliminary Analysis
  • Neocortex / OCTON In-Depth Analysis
© 2026 OCTON AIv1.0.0 / EARLY ACCESS
Install OCTON

Add OCTON to your home screen for faster captures and offline-ready scans.

AVA
Return to AVA