Education and assessment are separate
The Learning Center explains published concepts and clinical reasoning. The protected CER-AI application performs structured, patient-specific assessment. Reading an educational page never calculates, changes, or substitutes for a clinical result.
Independent pathways
CER-AI keeps ERSS, Final BAD-D, CER-AI-adapted NICE, and PS3 visible as distinct assessment channels. It does not present them as one externally validated proprietary probability. Agreement and disagreement remain available for surgeon review.
Source provenance and uncertainty
Decision-critical extracted values remain tied to defined source fields. Missing, unreadable, conflicting, and surgeon-completed data are not intended to disappear inside a final label. This supports auditability and reduces silent source interchange.
Validation language
Publications cited by CER-AI support specific variables, systems, or background concepts. They do not automatically validate CER-AI as a complete product. CER-AI-specific diagnostic performance, clinical outcome, superiority, or regulatory claims require their own supporting evidence.
How independent pathways become the final result
| Condition | Final disposition | Priority |
|---|---|---|
| Any STOP-DEFER driver | STOP-DEFER | Overrides all other results |
| No stop, but critical data incomplete | ASSESSMENT INCOMPLETE | Missingness is never converted to PASS |
| Four systems complete; 0 or 1 CAUTION | PASS | Individual caution remains visible in its section |
| Four systems complete; 2 CAUTION | PASS WITH CAUTION | Both drivers are listed |
| Four systems complete; 3 or 4 CAUTION | CAUTION | All drivers are listed |
| Independent non-system caution | CAUTION | Retains its own authority |
The four counted systems are ERSS, Final BAD-D, NICE, and PS3. Tissue and clinical-eligibility stops remain independent; bilateral aggregation preserves the worse eye and does not add caution counts across eyes.
What the structured report carries
| Report block | Content | Why it matters |
|---|---|---|
| ERSS | Five component rows, points, total, topography category, disposition | Auditable weighted score |
| NICE | Four inputs, points, total, category, disposition | Auditable cumulative score |
| PS3 | Every factor, exact finding, Moderate/High counts, procedure disposition | Shows why a procedure is allowed, deferred, or incomplete |
| BAD-D | Final D, classification, status, component/PPI/ART context | Separates the decision signal from context |
| Safety and provenance | RSB/RST, PTA, final-K checks, source fields, confirmations, versions | Preserves independent constraints and traceability |
PDF and Word reports consume the same already-computed canonical report payload; report rendering does not recalculate clinical rules.
Selected sources
See the complete CER-AI medical reference registry for broader context.