bright-eds-molecular

BRIGHT NER: EDS-NLP (CamemBERT + CRF) fine-tuned for molecular

Description

This is a EDS-NLP (CamemBERT + CRF) architecture fine-tuned to extract clinical neuro-oncology entities related to the molecular semantic group. It was trained on a synthetic dataset generated for the properly de-identified BRIGHT project dataset (see the generated_data folder in the primary repository).

This model repository was specifically designed to fit within the bright_db overarching namespace.

Fields

It extracts the following fields (described in French):

• mol_idh1: Statut mutation IDH1
• mol_idh2: Statut mutation IDH2
• mol_mgmt: Méthylation promoteur MGMT
• mol_h3f3a: Mutation H3F3A
• mol_hist1h3b: Mutation HIST1H3B
• mol_tert: Mutation promoteur TERT
• mol_CDKN2A: Délétion homozygote CDKN2A
• mol_atrx: Mutation ATRX
• mol_cic: Mutation CIC
• mol_fubp1: Mutation FUBP1
• mol_fgfr1: Mutation FGFR1
• mol_egfr_mut: Mutation EGFR
• mol_prkca: Mutation PRKCA
• mol_pten: Mutation PTEN
• mol_p53: Mutation p53
• mol_braf: Mutation BRAF

Performance on Validation Set

Aggregates:

• Macro F1: 0.5636 (Precision: 0.5866, Recall: 0.5486)
• Micro F1: 0.8612 (Precision: 0.8680, Recall: 0.8544)

Per-Label Breakdowns:

Usage

# Inference Code
import edsnlp

nlp = edsnlp.load("raphael-r/bright-eds-molecular")
doc = nlp("Patient presenting with epileptic seizures...")

for ent in doc.ents:
    print(ent.text, "=>", ent.label_)