---

language:

• en library_name: transformers tags:
• medical
• symptoms
• diagnosis
• text-classification
• healthcare
• clinical
• symptom-to-disease
• experimental
• test-version pipeline_tag: text-classification metrics:
• accuracy
• f1
• precision
• recall license: mit model-index:
• name: medical-diagnosis-en results:
  • task: type: text-classification name: Medical Diagnosis Classification (English) — TEST VERSION dataset: name: NeuronZero/Symptom2Disease type: NeuronZero/Symptom2Disease metrics:
    • name: Accuracy type: accuracy value: 0.8167
    • name: F1 (weighted) type: f1 value: 0.7966
    • name: F1 (macro) type: f1 value: 0.7966
    • name: F1 (micro) type: f1 value: 0.8167
    • name: Precision (weighted) type: precision value: 0.8441
    • name: Recall (weighted) type: recall value: 0.8167

---

🏥 Medical Diagnosis Classifier (EN) — TEST VERSION

⚠️ TEST VERSION / EXPERIMENTAL MODEL

This is an experimental test version of a medical diagnosis classifier. NOT intended for clinical use, self-diagnosis, or medical decision-making. Provided AS-IS for research and educational purposes only.

---

🚨 Important Notice

THIS IS A TEST VERSION of a fine-tuned Bio_ClinicalBERT model for classifying patient symptom descriptions into medical diagnoses in English.

⚠️ Test Version Limitations

• 🧪 Experimental quality: Trained on limited synthetic data (1,200 examples)
• 🔬 Not clinically validated: Results may be inaccurate or misleading
• 📉 ~18% error rate: Significant chance of wrong predictions
• 🌍 Limited scope: Only 24 medical conditions
• 🚫 No medical authority: Not approved for use in any healthcare setting

---

📋 Model Description

The model takes a natural-language description of patient symptoms as input and returns the most likely diagnosis (or top-N diagnoses with confidence scores).

• Base model: emilyalsentzer/Bio_ClinicalBERT
• Dataset: NeuronZero/Symptom2Disease
• Number of classes: 24
• Max input length: 128 tokens
• Framework: PyTorch / Hugging Face Transformers
• Status: 🧪 TEST VERSION

Supported Diagnoses (Test Version)

---

📊 Test Version Metrics

Per-Class Performance (Test Version)

Known Issues (Test Version)

• ❌ Peptic ulcer disease: Model failed to learn this class (0% recall)
• ⚠️ Drug reaction: Very low recall (20%) — model misses 80% of cases
• ⚠️ Allergy: Low recall (40%) — model misses 60% of cases
• ⚠️ GERD: Low precision (42%) — many false positives

---

🚀 Usage (For Research Only)

⚠️ WARNING: This is a TEST VERSION. Do not use for real medical decisions.

Using pipeline (recommended)

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="ilyali034/bert-medical-diagnosis-en",
    top_k=3
)

# FOR RESEARCH ONLY — NOT FOR MEDICAL USE
result = classifier("I have a severe headache, nausea, and sensitivity to light for two days")
print(result)
# [[{'label': 'Migraine', 'score': 0.85}, {'label': 'Common Cold', 'score': 0.08}, ...]]