Every recording processed by the ERG Analysis API produces one report built in four layers. Each layer is written for a different reader — from a general clinician glancing at a result, through to a researcher auditing exactly how the model reached its conclusion. Nothing in a higher layer is hidden from a lower one; the layers simply add detail and technical depth as you go deeper.
Layer 1 — Traffic Signal Summary
The fastest layer to read. It shows a single colour-coded signal — Green, Amber, or Red — alongside one plain-language sentence, a signal quality score, and a confidence percentage. Any clinical user can act on this layer alone.
Layer 2 — Clinical Summary
Written for optometrists, GPs, and junior clinicians. It lists the key measured parameters next to the patient’s values and the expected normal range, flags which ones fall outside that range, gives the model’s disease probability breakdown, and recommends a next action with a timeframe.
Layer 3 — Specialist Report
Built for retinal specialists and electrophysiologists. It includes the full waveform and time-frequency spectrogram, a SHAP “waterfall” chart showing which measured features most influenced the result, a comparison against normative reference values, and — where a prior recording exists — a longitudinal comparison over time.
Layer 4 — Technical / Audit Report
For researchers and regulatory reviewers. It records the model version used, the training sample size behind each disease category, accuracy statistics with confidence intervals, a full signal-quality audit, the processing configuration, a timestamped log, and the report’s HL7 FHIR identifier — the record that allows every result to be traced and reviewed after the fact.
Why four layers, not one report?
A single flat report either overwhelms a busy clinician with technical detail they don’t need, or hides the evidence a specialist needs to check the model’s reasoning. Structuring the output this way means a GP can act on Layer 1 in seconds, while a specialist reviewing a borderline case can drill down to Layer 3 or 4 without anyone needing a second, separate document.