Parse
Vendor-aware DICOM handling for Heidelberg Spectralis and Zeiss Cirrus, including structured reports and private thickness tags when present.
About this site
A personal prototype that turns published OCT/MS thresholds into a report UI.
RetinaSignal is a research-grade demo I built in 2026. It shows how routine optical coherence tomography (OCT) measurements — peripapillary RNFL and macular GCIPL thickness — can be compared against published multiple-sclerosis criteria and shown as a structured, evidence-labeled report. It is not a company, not a medical device, and not for patient care.
What this demo does
The local Python pipeline and this hosted UI cover a six-screen clinician-facing workflow.
Measure
pRNFL and mGCIPL values, bilateral inter-eye difference, age-aware normative comparison, and a quality/confidence score.
Compare
Rules aligned to published OCT/MS literature, with evidence-strength labels on each flag and an abstention path for low-confidence scans.
Export
PDF, JSON, and CSV reports, plus a longitudinal rate-of-change view and a similar-case gallery for illustration only.
What is real on this hosted site
The public site is a static walkthrough. The full parser runs locally.
You can try
- The six-screen workflow with a built-in bilateral demo case.
- Threshold tables, risk-tier copy, confidence factors, and clinician-override logging in the browser.
- PDF / JSON / CSV download generated in the browser.
- US, UK, and Hong Kong report-layout variants.
Still illustrative
- Hosted file upload does not run the DICOM parser. Dropped files get the same synthetic biomarkers as “Load demo data.”
- Similar-cases cards use a public retinal OCT dataset (Kermany OCT2017) with no MS labels.
- Quality gating here is a demo proxy, not OSCAR-IB.
- Nothing on this site is validated against a clinical MS cohort.
Evidence the rules point at
The prototype implements published cutoffs. It does not invent biomarkers.
Diagnostic (McDonald 2024)
Optic nerve can contribute to dissemination in space. Consensus OCT support is an inter-eye difference of ≥ 6 µm in pRNFL or ≥ 4 µm in macular GCIPL, with quality control, and not during acute optic neuritis.
Prognostic (literature, not the diagnostic rule)
Absolute thinning (for example pRNFL ≤ 88 µm or mGCIPL < 77 µm) and annual thinning rates come from MS OCT meta-analysis and consensus papers. Those are labeled separately from the DIS inter-eye criterion.
- Saidha S, et al. Use of OCT and VEPs in the 2024 McDonald diagnostic criteria. Lancet Neurol. 2025.
- IMSVISUAL Consortium longitudinal OCT-MS meta-analysis, 2025.
- Austrian OCT-MS Consensus. J Neurol. 2025.
What this is not
- Not a diagnostic test, treatment, or FDA/CE-marked device.
- Not a hospital product, EHR integration, or clinical service.
- Not advice for patients. If you have symptoms, talk to a clinician.
Who built it
Tobias Chan. Built as a personal project to learn OCT workflows, medical-data parsing, and how tightly a decision-support UI has to be tied to published claims. The prototype is frozen; this site is the walkthrough.