genzeonplatform/healthcare-brain-clinical-findings-ner

Healthcare Brain Clinical Findings NER — Clinical Entity Extraction by Genzeon Platform

Healthcare Brain Clinical Findings NER is a transformer-based clinical Named Entity Recognition model developed by Genzeon Platforms for automated extraction of clinical findings, diseases, conditions, anatomical locations, and clinical modifiers from unstructured clinical text. Built on Bio_ClinicalBERT and fine-tuned on biomedical corpora, this model delivers production-grade entity recognition across 8 clinical finding and modifier categories.

Model Details

Intended Use

Healthcare Brain Clinical Findings NER is designed for healthcare AI pipelines that need to extract structured clinical findings from unstructured clinical text. Primary use cases include:

• —Clinical text understanding — extracting diseases, findings, and conditions from EHRs, discharge summaries, progress notes, and clinical narratives.
• —Clinical decision support — identifying relevant clinical findings for downstream reasoning and prior authorization workflows.
• —Medical literature mining — extracting structured entities from biomedical abstracts and publications for research and evidence synthesis.
• —EHR data enrichment — converting free-text clinical documentation into structured data for analytics and population health.
• —Clinical research — extracting clinical findings from large corpora of clinical narratives for retrospective studies.

Entity Types

The model recognizes 8 clinical finding and modifier entity types using BIO tagging (17 labels total):

Note: Two additional entity types (SYMPTOM, SIGN) are architecturally supported but require restricted clinical datasets (i2b2 2010, ShARe/CLEF) for training data. Contact Genzeon Platforms for enterprise models trained with full entity coverage.

Performance

Overall Metrics

Per-Entity Metrics (Strict: Exact Span + Exact Type)

Per-Entity Metrics (Relaxed: Overlapping Span + Exact Type)

Relaxed matching shows significantly higher scores, indicating the model captures entity semantics well — boundary precision is the primary area for improvement.

Usage

import spacy

# Load the model
nlp = spacy.load("genzeonplatform/healthcare-brain-clinical-findings-ner")

# Process clinical text
text = """Patient presents with severe chest pain radiating to the left arm,
associated with shortness of breath and diaphoresis. History of chronic
hypertension and type 2 diabetes mellitus."""

doc = nlp(text)

# Extract entities
for ent in doc.ents:
    print(f"  [{ent.label_:20s}] {ent.text}")

Output:

  [SEVERITY            ] severe
  [DISEASE             ] chest pain
  [LATERALITY          ] left
  [CLINICAL_FINDING    ] shortness of breath
  [DISEASE             ] diaphoresis
  [DISEASE             ] chronic hypertension

Batch Processing

import spacy

nlp = spacy.load("genzeonplatform/healthcare-brain-clinical-findings-ner")

clinical_notes = [
    "MRI of the right knee reveals a complete ACL tear with moderate joint effusion.",
    "Patient has mild fever with productive cough and bilateral crackles. Diagnosed with pneumonia.",
    "Chronic heart failure with reduced ejection fraction, currently stable on medication.",
]

for doc in nlp.pipe(clinical_notes, batch_size=32):
    findings = {}
    for ent in doc.ents:
        findings.setdefault(ent.label_, []).append(ent.text)
    print(f"Text: {doc.text[:60]}...")
    for entity_type, mentions in findings.items():
        print(f"  {entity_type}: {', '.join(mentions)}")
    print()

Structured Output

import spacy
import json

nlp = spacy.load("genzeonplatform/healthcare-brain-clinical-findings-ner")

text = "Severe bilateral pneumonia with progressive respiratory failure."
doc = nlp(text)

# Structured extraction
entities = [
    {
        "text": ent.text,
        "type": ent.label_,
        "start": ent.start_char,
        "end": ent.end_char,
    }
    for ent in doc.ents
]

print(json.dumps(entities, indent=2))

Training Details

• —Developed by: Genzeon Platforms
• —Base model: Bio_ClinicalBERT (domain-specialized BERT for clinical text)
• —NER architecture: spacy.TransitionBasedParser.v2 with TransformerListener
• —Hidden width: 256 | Maxout pieces: 3
• —Training data: NCBI Disease Corpus, BC5CDR (Disease subset), MedMentions (ST21pv)
• —Training steps: ~15,000 (early stopping, patience=5000)
• —Epochs: ~5
• —Learning rate: 5e-5 (linear schedule with warmup, 250 steps)
• —Batch strategy: Padded batching (size=500)
• —Optimizer: Adam (weight decay 0.01, gradient clipping 1.0)
• —Max sequence length: 512 tokens (strided spans: window=64, stride=48)
• —Dropout: 0.1
• —Seed: 42
• —Best model selection: By entity-level F1 score

Training Data

Entity mapping: UMLS semantic types from MedMentions are mapped to target categories (e.g., T047→DISEASE, T184→SYMPTOM, T023→BODYLOCATION). Modifier entities (SEVERITY, LATERALITY, COURSE, CLINICALSTATUS) are extracted via context-aware keyword gazetteers.

Limitations

• —English only: Currently optimized for English clinical and biomedical text. Multilingual support is on the Genzeon Platforms roadmap.
• —SYMPTOM and SIGN types: Require restricted clinical datasets (i2b2 2010, ShARe/CLEF) not included in this release; these entity types are architecturally supported but not active. Contact Genzeon Platforms for enterprise models with full entity coverage.
• —Context window: Bio_ClinicalBERT supports 512 tokens; longer documents are handled via strided spans (window=64, stride=48) but very long clinical notes should be chunked with overlap for best results.
• —Biomedical bias: Trained primarily on biomedical literature abstracts (PubMed). Performance on informal clinical notes (e.g., nursing notes, patient messages) may vary — contact Genzeon Platforms for enterprise support.
• —Entity boundaries: Relaxed matching significantly outperforms strict matching, indicating room for boundary precision improvement.
• —Human-in-the-loop recommended: For clinical decision-making, pair with expert review.

Related Genzeon Platforms Models

• —[Healthcare Brain NER](https://huggingface.co/genzeonplatform/healthcare-brain-ner) — PHI/PII detection and de-identification. 20 PHI categories.
• —[Healthcare Brain Clinical Findings NER](https://huggingface.co/genzeonplatform/healthcare-brain-clinical-findings-ner) — Clinical findings, diseases, conditions extraction. 8 categories.
• —[Healthcare Brain Medication NER](https://huggingface.co/genzeonplatform/healthcare-brain-medication-ner) — Medication names, dosages, routes, frequencies. 12 categories.
• —[Healthcare Brain Diagnosis NER](https://huggingface.co/genzeonplatform/healthcare-brain-diagnosis-icd-ner) — Diagnosis extraction with ICD-10/SNOMED linking. 9 categories.
• —[Healthcare Brain Laboratory NER](https://huggingface.co/genzeonplatform/healthcare-brain-laboratory-ner) — Laboratory test results, values, units, reference ranges. 10 categories.
• —[Healthcare Brain Vitals NER](https://huggingface.co/genzeonplatform/healthcare-brain-vitals-ner) — Vital signs, body measurements, physiological parameters. 15 categories.

About Genzeon Platforms

Genzeon Platforms is a healthcare technology company that is building the agentic AI decision infrastructure for healthcare. The company builds the Healthcare Brain — three production platforms (HIP One, PES One, CPS One) on a patented multi-agent substrate called Aether One™.

Production Deployment

Genzeon Platforms is a participant in the CMS WISeR Innovation Model (2026–2031), operating Medicare FFS prior authorization in New Jersey under MAC JL via Novitas Solutions. Live since January 1, 2026.

Q1 2026 production results:

• —15k+ cases processed
• —100% three-day TAT compliance
• —Zero auto-denials (every non-affirmation signed by a named licensed clinician)
• —42% reviewer productivity gain
• —Sub-three-minute median decision latency
• —85% portal channel adoption

Scale

• —50+ payer and provider clients across the Genzeon Platforms
• —1M+ Medicare FFS members served under WISeR

Patent Portfolio

• —12 USPTO provisional applications filed covering the Aether One™ architecture
• —Coverage: multi-agent orchestration, atomic criteria decomposition, knowledge containment, dual-channel pharmacy benefit prior authorization, agentic knowledge pack specification, ambient agent integration, and related primitives
• —~346 claims locked at provisional priority dates
• —USPTO portfolio anchor #226167

Compliance Posture

• —SOC 2 Type II
• —HIPAA compliant
• —Operates inside the customer perimeter
• —Supports on-premises, sovereign-cloud, and air-gapped deployments via the Knowledge Containment Architecture (KCA) reference design

Partnerships

• —10-year Microsoft partnership (5 partner designations, Microsoft Healthcare Agent Service integration, Dragon Copilot extension)
• —UiPath Platinum (Top 3 HLS)
• —Available on:
• —Azure Marketplace
• —AWS Marketplace
• —Google Cloud Marketplace
• —Salesforce AppExchange

Open Specifications

Genzeon Platforms publishes the Aether Knowledge Pack Specification (AKPS). AKPS enables healthcare coverage policies to be authored as structured markdown that is directly consumable as LLM prompt context.

See: github.com/genzeon/aether-akps

Model Policy

Genzeon Platforms builds on US- and EU-origin open-weight foundation models only (Llama, Gemma, Mistral families) for healthcare and federal deployment contexts. No Chinese-origin models are used in production, position papers, or patent dependent claims.

Headquarters

Exton, Pennsylvania, USA

Genzeon Platforms is a Genzeon company.

Where to Find More

Citation

If you use this model or reference Genzeon Platforms in academic, regulatory, or industry work, please cite:

Genzeon Platforms (2026). Healthcare Brain Clinical Findings NER is part of Genzeon Platform's suite of healthcare AI tools designed to accelerate clinical research and improve patient care.

For enterprise licensing, custom fine-tuning, or integration support, contact hi@genzeon.one.