Ophthalmic Signal Research
Research-only · Not for clinical use · Living document · Updated June 2026
OculoDx documents how consumer mobile hardware — iPhone TrueDepth, LiDAR, ARKit — can be repurposed as a calibrated screening instrument for 308 ocular and neurological conditions. The atlas pairs each condition with the evidence for what these sensors can recover, the evidence against, and the multi-modal signals that distinguish it from its differential diagnoses.
Phase-0 delivers 59 transparent, deterministic scoring models — fully auditable, no opaque ML classifiers. Conditions without a Phase-0 model are explicitly flagged. What we know, what we suspect, and what we cannot yet say.
A typed graph of elements — from raw sensor capture through observations, computed features, and clinically interpreted biomarkers to condition-specific scoring models. Each kind builds on the one below.
| Kind | Name | Count | Description |
|---|---|---|---|
| Condition | Conditions | 308 | Clinical taxonomy — the assessed targets |
| Model | Scoring Models | 59 | Rule-based, transparent, fully auditable |
| Biomarker | Biomarkers | 104 | Clinical interpretation of computed features |
| Feature | Features | 114 | Pure numbers computed from observations |
| Observation | Observations | 46 | Raw givens produced by tests |
| Test | Tests | 7 | Frozen capture protocols |
| Sensor | Sensors | 20 | Consumer mobile hardware (TrueDepth, LiDAR, ARKit) |
A sample of the atlas — one entry per modality, each with its signal chain and detection probability.
| Condition | ICD | Family | Detection | Phase-0 Model |
|---|---|---|---|---|
| Scleral icterus (jaundice) | R17 | systemic | 98% | ✅ jaundice-score-v1 |
| Corneal arcus | H18.41 | cornea | 97% | ✅ corneal-arcus-score-v1 |
| Aniridia (absent iris) | Q13.1 | pupil | 98% | ✅ iris-structure-score-v1 |
Each entry in the full atlas carries its required and auxiliary biomarkers, feature chain, evidence band, and detection probability. 308 entries total.
OculoDx applies a T0–T3 validation framework with evidence banding on every entry — distinguishing what is experimentally confirmed from what is theoretically grounded or clinically plausible.
- Deterministic scoring only. All 59 Phase-0 models are rule-based and fully auditable. No ML classifiers. No black boxes.
- Evidence-annotated throughout. All 114 features and 104 biomarkers carry literature citations and validation tier annotations.
- Honest coverage. Conditions without a wired Phase-0 model are explicitly flagged. Phase-1 extends the inventory to Android and wearables; Phase-2 adds prospective cohort validation and ML-based scoring.
The first open taxonomy mapping 308 conditions to mobile sensors through a typed graph of observations, features, and biomarkers.
OculoDx Research is invitation-only.
| Reviewer Access | Researcher Access |
|---|---|
| For peer reviewers evaluating the methodology pre-publication. Time-bounded. | For scholars working on adjacent methodology who want to use the atlas in their own work. |
| research@oculodx.ai | research@oculodx.ai |
Already have credentials? → Sign in to the research portal
OculoDx Research Team. (2026). OculoDx Condition Atlas and Signal Dictionary, Phase 0.
OculoDx, Inc. https://research.oculodx.ai
© 2026 OculoDx, Inc. · Research-only · Not for clinical use · research@oculodx.ai
