README.md

metadata

language:
  - en
license: mit
library_name: transformers
tags:
  - medical
  - cardiology
  - emergency-medicine
  - classification
  - bert
datasets:
  - MIMIC-III
metrics:
  - accuracy
  - f1
  - auc
model-index:
  - name: OHCA-Classifier-V9
    results:
      - task:
          type: text-classification
          name: Out-of-Hospital Cardiac Arrest Detection
        metrics:
          - type: accuracy
            value: 0.86
          - type: f1
            value: 0.8
          - type: auc
            value: 0.905

OHCA Classifier V9

Model Description

This model classifies clinical notes to identify Out-of-Hospital Cardiac Arrest (OHCA) cases. It's based on BiomedNLP-PubMedBERT and trained on MIMIC-III data.

Model Performance

• AUC-ROC: 0.905
• Sensitivity: 80%
• Specificity: 95%
• F1-Score: 0.80
• Optimal Threshold: 0.120

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model
tokenizer = AutoTokenizer.from_pretrained("monajm36/ohca-classifier-v9")
model = AutoModelForSequenceClassification.from_pretrained("monajm36/ohca-classifier-v9")

# Predict
def predict_ohca(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.softmax(outputs.logits, dim=-1)
    return probs[0][1].item()  # OHCA probability

# Example
text = "Chief Complaint: Cardiac arrest. HPI: Patient found unresponsive..."
prob = predict_ohca(text)
print(f"OHCA Probability: {prob:.3f}")

Training Data

• Dataset: MIMIC-III
• Samples: 330 clinical notes
• Classes: Binary (OHCA vs Non-OHCA)

Disclaimer

This model is for research purposes only. Not intended for clinical decision-making.