cbim-framework

CBI-M (Characterization-Based Injury Model) Framework — the NINDS TBI reclassification ontology and controlled vocabularies. Open reference implementation for the Manley et al. Lancet Neurology 2025 consensus framework.

Status

Stable specification. Ontology files are production-used in EvidenceOS BRIDGE-TBI and NeuroFusion. Community contributions welcome.

What's here

• ontology/ — 7 YAML ontology files (3,381+ lines total)
  • cbim_framework.yaml — Core 16-class patient classification (CBIM-01 to CBIM-16)
  • temporal_phases.yaml — Acute / subacute / chronic phase definitions
  • clinical_entities.yaml — GCS, imaging, biomarker, outcome entities
  • biomarker_kinetics.yaml — Biomarker trajectory definitions (GFAP, UCH-L1, S100B, NSE, NfL)
  • imaging_cdes.yaml — CT/MRI Common Data Elements aligned to NINDS
  • outcome_mapping.yaml — GOSE + GOCS + extended outcome mappings
  • resource_contexts.yaml — T×M×P facility/mechanism/physiology stratification
• schemas/ — JSON Schema and FHIR profiles derived from YAML ontology
• examples/ — Patient classification worked examples (synthetic data)
• validation/ — CI validation scripts for ontology consistency

The CBI-M classification system

CBI-M defines 16 patient classes (CBIM-01 to CBIM-16) based on:

• Structural injury pattern (focal / diffuse / mixed)
• Mechanism (blunt / penetrating / blast)
• Severity (GCS + imaging features + physiological modifiers)
• Temporal phase at presentation

CBIM naming is distinct from BC- biomarker combo naming. CBIM-01 to CBIM-16 = patient classes. BC-000 to BC-127 = biomarker combinations. Never conflate.

NINDS alignment

This ontology operationalises the 2025 NINDS TBI Reclassification Working Group consensus (8 Working Groups, ~580-paper citation corpus). It is designed to be the machine-readable implementation of the Manley et al. Lancet Neurology 2025 framework.

Contributing

Contributions from clinical neurologists and neurotrauma researchers are particularly valuable. All changes to CBIM class definitions require clinical validation evidence. See CONTRIBUTING.md for the review pathway.

License

CC-BY-4.0 — freely usable with attribution. Cite as: EvidenceOS CBI-M Ontology v[version], 2026. Based on: Manley et al., Lancet Neurology, 2025.

Maintainer

EvidenceOS Research Lab — ontology@evidenceos.com

Related repos

• evidence-commons — EvidenceOS canonical evidence layer
• evidenceos-bridge-tbi — Clinical deployment using CBI-M
• multiverse-analysis-toolkit — Ontology-constrained multiverse analysis
• tripod-ai-templates — TRIPOD+AI reporting using CBI-M