app.py

import gradio as gr
from transformers import pipeline

# Load the model
classifier = pipeline(
    "text-classification", 
    model="ashishkgpian/biobert_icd9_classifier_ehr"
)

def classify_symptoms(text):
    """
    Classify medical symptoms and return top ICD9 codes
    
    Args:
        text (str): Input medical symptom description
    
    Returns:
        dict: Top classification results with ICD9 codes and probabilities
    """
    try:
        # Get classification results
        results = classifier(text, top_k=5)
        
        # Format results for more readable output
        formatted_results = []
        for result in results:
            formatted_results.append({
                "ICD9 Code": result['label'], 
                "Confidence": f"{result['score']:.2%}"
            })
        
        return formatted_results
    
    except Exception as e:
        return f"Error processing classification: {str(e)}"

# Create Gradio interface
demo = gr.Interface(
    fn=classify_symptoms,
    inputs=gr.Textbox(
        label="Enter Medical Symptoms", 
        placeholder="Describe patient symptoms here..."
    ),
    outputs=gr.JSON(label="Top 5 ICD9 Classifications"),
    title="BioBERT ICD9 Symptom Classifier",
    description="Classify medical symptoms into ICD9 diagnostic codes using a fine-tuned BioBERT model.",
    theme="huggingface",
    examples=[
        ["Patient experiencing chest pain and shortness of breath"],
        ["Recurring headaches with nausea"],
        ["Diabetic symptoms including frequent urination"]
    ]
)

if __name__ == "__main__":
    demo.launch()