app.py

import gradio as gr
import pandas as pd
import numpy as np
import joblib

# Load the model and scaler
model = joblib.load('catboost_model.pkl')
scaler = joblib.load('scaler.pkl')

# Feature list（顺序要与训练时一致）
cat_features = ['Gender', 'Race', 'Education', 'Marital status', 'Smoke']
num_features = ['Age', 'Family income', 'BMI', 'MPAH', 'PFDE', 'PFHxS', 'PFNA', 'PFOA', 'PFOS', 'PFUA']
all_features = cat_features + num_features


# Define the prediction function
def predict_copd(age, gender, race, education, marital_status, family_income, bmi, smoke,
                 mpah, pfde, pfhxs, pfna, pfoa, pfos, pfua):
    # Step 1: Create input DataFrame
    input_data = pd.DataFrame([{
        'Gender': gender,
        'Race': race,
        'Education': education,
        'Marital status': marital_status,
        'Smoke': smoke,
        'Age': age,
        'Family income': family_income,
        'BMI': bmi,
        'MPAH': mpah,
        'PFDE': pfde,
        'PFHxS': pfhxs,
        'PFNA': pfna,
        'PFOA': pfoa,
        'PFOS': pfos,
        'PFUA': pfua
    }])

    # Step 2: Standardize numeric features only
    input_data[num_features] = scaler.transform(input_data[num_features])

    # Step 3: Ensure column order matches model training
    input_data = input_data[all_features]

    # Step 4: Make prediction
    prediction = model.predict(input_data)[0]
    probability = model.predict_proba(input_data)[0][1]

    # Step 5: Format output
    result_text = "Prediction: COPD" if prediction == 1 else "Prediction: Healthy"
    risk_level = "Low Risk" if probability < 0.3 else "Medium Risk" if probability < 0.7 else "High Risk"

    return f"{result_text}\nRisk Probability: {probability:.2%}\nRisk Level: {risk_level}"


# Gradio UI
with gr.Blocks() as iface:
    gr.Markdown("# COPD Risk Prediction Calculator")
    gr.Markdown("Enter patient information to predict COPD risk.")

    with gr.Row():
        with gr.Column():
            gr.Markdown("## Demographic Information")
            age = gr.Number(label="Age", value=40)
            gender = gr.Radio([1, 2], label="Gender", info="1: Male, 2: Female", value=1)
            race = gr.Dropdown([1, 2, 3, 4, 5], label="Race",
                               info="1: Mexican American, 2: Other Hispanic, 3: Non-Hispanic White, 4: Non-Hispanic Black, 5: Other Race",
                               value=3)
            education = gr.Radio([0, 1], label="Education Level",
                                 info="0: High school or below, 1: Above high school",
                                 value=0)
            marital_status = gr.Dropdown([1, 2, 3, 4, 5, 6], label="Marital Status",
                                         info="1: Married, 2: Widowed, 3: Divorced, 4: Separated, 5: Never married, 6: Living with partner",
                                         value=1)
            family_income = gr.Number(label="Family Income (PIR)", value=3.0,
                                      info="Poverty Income Ratio")

        with gr.Column():
            gr.Markdown("## Health Indicators")
            bmi = gr.Number(label="BMI", value=25.0)
            smoke = gr.Radio([0, 1], label="Smoking", info="0: No, 1: Yes", value=0)

            gr.Markdown("## Biomarkers (ng/mL)")
            mpah = gr.Number(label="MPAH", value=0.5)
            pfde = gr.Number(label="PFDE", value=0.5)
            pfhxs = gr.Number(label="PFHxS", value=1.0)
            pfna = gr.Number(label="PFNA", value=0.5)
            pfoa = gr.Number(label="PFOA", value=2.0)
            pfos = gr.Number(label="PFOS", value=5.0)
            pfua = gr.Number(label="PFUA", value=0.5)

    predict_button = gr.Button("Predict")
    output_textbox = gr.Textbox(label="Prediction Result")

    predict_button.click(
        predict_copd,
        inputs=[age, gender, race, education, marital_status, family_income, bmi, smoke,
                mpah, pfde, pfhxs, pfna, pfoa, pfos, pfua],
        outputs=output_textbox
    )

    gr.Markdown("### Variable Information")
    gr.Markdown("""
    **Demographic Variables:**
    - **Gender**: 1=Male, 2=Female  
    - **Race**: 1=Mexican American, 2=Other Hispanic, 3=Non-Hispanic White, 4=Non-Hispanic Black, 5=Other Race  
    - **Education**: 0=High school or below, 1=Above high school  
    - **Marital Status**: 1=Married, 2=Widowed, 3=Divorced, 4=Separated, 5=Never married, 6=Living with partner  

    **Health Variables:**
    - **BMI**: Body Mass Index (kg/m²)  
    - **Smoke**: 0=No, 1=Yes  

    **Biomarkers:**
    - **MPAH, PFDE, PFHxS, PFNA, PFOA, PFOS, PFUA**: Per- and polyfluoroalkyl substances (PFAS) in ng/mL
    """)

# Launch app
iface.launch()