CLINICAL

ONTOLOGY

REASONING

ENGINE

Verifiable clinical reasoning.
at scale.

Verifiable clinical reasoning.
at scale.

Encodes clinical knowledge into a structured ontology and knowledge graph that AI agents reason against deterministically.

REQUEST A DEMO

→

SEE HOW IT WORKS

→

CLINICAL

ONTOLOGY

REASONING

ENGINE

Every recommendation traceable.
Every decision auditable. Every output defensible.

Every recommendation traceable.
Every decision auditable. Every output defensible.

Every recommendation
traceable.
Every decision
auditable.
Every output
defensible.

THE PROBLEM

Healthcare AI has an accuracy problem.

General-purpose AI hallucinates. It generates plausible-sounding clinical outputs that are wrong in ways you can't detect without re-doing the entire analysis yourself. In regulated healthcare, that's not a risk. It's a liability.

01

It hallucinates

General-purpose LLMs produce clinical errors at rates of 58 to 82 percent. They generate confident, well-structured outputs that are factually wrong.

CORE uses deterministic reasoning against structured clinical knowledge. The ontology constrains what the system can conclude.

01

It hallucinates

General-purpose LLMs produce clinical errors at rates of 58 to 82 percent. They generate confident, well-structured outputs that are factually wrong.

CORE uses deterministic reasoning against structured clinical knowledge. The ontology constrains what the system can conclude.

02

It's a black box

You can't trace how the AI reached its conclusion. There's no audit trail, no evidence chain, no way to verify the reasoning or defend it at IMR.

CORE provides a Glass Box: every finding links to source text, every criterion to its guideline section, every decision to a complete evidence chain.

02

It's a black box

You can't trace how the AI reached its conclusion. There's no audit trail, no evidence chain, no way to verify the reasoning or defend it at IMR.

CORE provides a Glass Box: every finding links to source text, every criterion to its guideline section, every decision to a complete evidence chain.

03

It doesn't know
the rules

LLMs can read medical records but they don't understand clinical practice guidelines, the compounding rules that govern determinations, or the regulatory frameworks that vary by jurisdiction and line of business.

CORE encodes clinical guidelines into a structured knowledge graph with deterministic decision logic. It doesn't interpret the rules. It applies them.

03

It doesn't know
the rules

LLMs can read medical records but they don't understand clinical practice guidelines, the compounding rules that govern determinations, or the regulatory frameworks that vary by jurisdiction and line of business.

CORE encodes clinical guidelines into a structured knowledge graph with deterministic decision logic. It doesn't interpret the rules. It applies them.

0
1
0
0
%
0
1
0
0
%

Neuro-symbolic AI accuracy in peer-reviewed clinical extraction study Nature Communications Medicine, 2025 (Prenosil, Weitzel et al.)

0
1
2
3
4
5
6
0
1
2
3
%
0
1
2
3
4
5
6
0
1
2
3
%

PT-4 alone on the same task, same data, same reports

The only difference was adding a structured verification layer

+
0
1
2
3
0
1
2
3
4
5
6
7
pts
+
0
1
2
3
0
1
2
3
4
5
6
7
pts

Accuracy gain from the architecture CORE is built on

Neuro-symbolic AI: LLMs read, knowledge graphs verify

The LLM reads. The knowledge graph reasons. The human decides.

The LLM reads. The knowledge graph reasons. The human decides.

The LLM reads. The knowledge graph reasons. The human
decides.

HOW IT WORKS

Neuro-symbolic AI for clinical reasoning.

CORE combines the reading power of large language models with the reasoning precision of deterministic knowledge graphs. The LLM never makes the clinical determination. It reads. The graph decides.

ONTOLOGY

Clinical ontology and knowledge graph

Clinical guidelines encoded as structured, machine-actionable knowledge. Entity types, relationships, scoping rules, and decision logic. Not retrieved documents. Computable logic. CORE's ontology includes a foundation of standard medical guidelines and terminology alongside specialized regulatory frameworks like ACOEM and California MTUS. The architecture is designed to extend to additional clinical practice guidelines as new verticals and jurisdictions come online.

SEE HOW IT WORKS

→

AGENTS

Clinical AI agents

Specialized agents that work in sequence, each with a defined scope and measurable criteria. One identifies the request type and applicable rules. One reads and indexes the medical documents against the ontology. One extracts every relevant clinical entity. One generates a verified clinical summary. Each agent hands off to the next. Each output is traceable.

SEE HOW IT WORKS

→

GLASS BOX

Traceable clinical outputs

The system produces structured evidence chains, not summaries. Gap analysis with exact citations. Relevance-filtered timelines. Guideline-matched findings. Every conclusion has a verifiable path from the decision back through the knowledge graph to the source medical record. The reviewer evaluates the evidence chain. They don't re-read 400 pages.

CORE Auditability

Open any determination to its full basis.

CORE writes an entry for each determination it issues. The log shows the outcome and whether a human reviewed it. Open a record and you see the facts it read, cited to a source page, and the deterministic rules it applied, each tied to the governing guideline.

CAPABLLITIES

The Glass Box: every decision traceable.

Where black-box AI gives you an output and asks you to trust it, CORE gives you the complete evidence chain and lets you verify it.

ZERO DATA EGRESS

CORE isn't SaaS. It runs inside your walls.

CORE deploys directly into your enterprise cloud infrastructure, VPC, or tenant. Your medical records, proprietary data, and PHI never leave your secure perimeter. No data egress. No third-party processing. No external API calls with patient information. Your environment, your firewall, your control.

Glass Box evidence mapping

Every guideline criterion matched to specific findings in the medical record. Source citations with exact page numbers. Highlighted text showing the evidence that supports or contradicts each criterion.

Relevance-filtered timeline

Patient history projected onto a chronological axis, filtered to the specific request. If the request is for a lumbar injection, only lumbar-related history is displayed. Gaps in therapy, symptom escalation, treatment progression.

Gap analysis

Specific gaps between the documented medical record and guideline requirements. Not "conservative therapy may not have been met." Instead: "ACOEM requires 6 weeks PT. Record shows 4 sessions over 3 weeks. Gap: duration not satisfied by 3 weeks."

Gap analysis

Specific gaps between the documented medical record and guideline requirements. Not "conservative therapy may not have been met." Instead: "ACOEM requires 6 weeks PT. Record shows 4 sessions over 3 weeks. Gap: duration not satisfied by 3 weeks."

Natural language queryrized deployment

Ask questions about the case file in plain language. Responses grounded in the knowledge graph and source documents. If the evidence doesn't exist, the system says so. It doesn't guess.

Clinical summarization

AI-generated summary containing all and only the information relevant to the specific case. Every statement traceable to its source document. Configurable for physician reviewers or publication.

Guideline ingestion

Complex text-based medical guidelines converted into computable logic. ACOEM, MTUS,, VA guidelines, state workers' compensation frameworks, and standard clinical practice guidelines. Evidence requirements, contraindications, and therapy durations automatically enforced. Extensible to new guideline sets as your needs grow.

Guideline ingestion

Complex text-based medical guidelines converted into computable logic. ACOEM, MTUS,, VA guidelines, state workers' compensation frameworks, and standard clinical practice guidelines. Evidence requirements, contraindications, and therapy durations automatically enforced. Extensible to new guideline sets as your needs grow.

SOLUTION

Ready-to-deploy clinical solutions.

CORE powers two types of solutions. Fusion products fuse AI reasoning with human expertise. Pipeline products automate clinical data processing end to end.

FUSION

AI-augmented human decision support

Fuses AI reasoning with human expertise to deliver trusted, verifiable outcomes at scale. The AI builds the evidence chain. The human makes the determination.

Fusion IMR

Fusion Litigation

PIPELINE

Fully automated agentic workflows

Autonomous agent compositions that process, transform, and structure data at scale. No human in the loop at runtime. Feeds into Fusion applications or delivers standalone output.

Pipeline Extract

Pipeline FHIR

PIPELINE

Fully automated agentic workflows

Autonomous agent compositions that process, transform, and structure data at scale. No human in the loop at runtime. Feeds into Fusion applications or delivers standalone output.

Pipeline Extract

Pipeline FHIR

THE PLATFORM

Build your own clinical solutions.

Build your own clinical solutions.

For sophisticated buyers who need the clinical reasoning platform to compose custom solutions for workflows we haven't productized yet.

as a platform

CORE's clinical ontology, knowledge graph, and agent pipeline are available as a platform for organizations that need to build custom clinical solutions. Bring your own workflow. Compose agents against the ontology. Deploy on COMAND. The hard part — encoding clinical knowledge into deterministic, auditable AI — is already done.

Clinical ontology access
A foundation of standard medical knowledge, established clinical practice guidelines, and specialized regulatory frameworks. Extend with your own guideline sets for new jurisdictions and lines of business.

Agent composition
Compose custom agent pipelines from CORE's modular clinical agents. Configure for your specific regulatory and workflow requirements.

Deploys in your cloud
CORE deploys directly into your VPC or tenant. Your PHI never leaves your secure perimeter. Full evaluation, monitoring, and governance infrastructure runs behind your firewall on COMAND.

©2026 Ground Truth Systems, Inc.