appbackup.py

#file_path = "cleaned_bmd_medication_data.xlsx"

import streamlit as st
import pandas as pd
import plotly.graph_objs as go

# Constants from linear regression
REGRESSION_CONSTANTS = {
    'Femoral Neck': {'mu': 0.916852, 'sigma': 0.120754},
    'Total Hip': {'mu': 0.955439, 'sigma': 0.125406},
    'Lumbar spine (L1-L4)': {'mu': 1.131649, 'sigma': 0.139618},
}

# Load medication data
@st.cache_data
def load_medication_data():
    file_path = "cleaned_bmd_medication_data.xlsx"
    return pd.read_excel(file_path)

# Calculate predicted BMD after medication
def calculate_bmd(bmd, percentage_increase):
    return bmd * (1 + percentage_increase)

# Convert BMD to T-score
def calculate_tscore(bmd, mu, sigma):
    return (bmd - mu) / sigma

# Generate prediction table for all drugs
def generate_predictions(medication_data, site, bmd, mu, sigma):
    site_data = medication_data[medication_data['Site'] == site]
    all_results = []
    
    for _, row in site_data.iterrows():
        drug = row['Medication']
        predictions = {
            'Year': ['0'], 
            'Year Index': [0],  # Numeric x-axis for plotting
            'Predicted BMD': [round(bmd, 3)], 
            'Predicted T-score': [round(calculate_tscore(bmd, mu, sigma), 1)]
        }
        
        year_index = 1
        for year in row.index[1:-1]:  # Skip 'Medication' and 'Site' columns
            if not pd.isna(row[year]):
                percentage_increase = row[year]
                predicted_bmd = bmd * (1 + percentage_increase)
                predicted_tscore = calculate_tscore(predicted_bmd, mu, sigma)
                
                predictions['Year'].append(year.replace(" Year", ""))  # Simplify year label
                predictions['Year Index'].append(year_index)  # Numeric x-axis
                predictions['Predicted BMD'].append(round(predicted_bmd, 3))
                predictions['Predicted T-score'].append(round(predicted_tscore, 1))
                year_index += 1
        
        all_results.append({'Drug': drug, 'Predictions': predictions})
    return all_results


# Display results as table and plots
def display_results(predictions, site):
    st.subheader(f"Predictions for {site}")
    
    for result in predictions:
        drug = result['Drug']
        predictions = result['Predictions']
        
        # Display table
        st.write(f"### {drug}")
        st.dataframe(pd.DataFrame(predictions))
        
        # Plot BMD and T-score using Year Index
        bmd_plot = go.Scatter(
            x=predictions['Year Index'], y=predictions['Predicted BMD'], mode='lines+markers',
            name='Predicted BMD', line=dict(color='blue')
        )
        tscore_plot = go.Scatter(
            x=predictions['Year Index'], y=predictions['Predicted T-score'], mode='lines+markers',
            name='Predicted T-score', line=dict(color='green')
        )
        
        # Combine plots in a single row
        col1, col2 = st.columns(2)
        with col1:
            st.plotly_chart(go.Figure(data=[bmd_plot], layout=go.Layout(
                title=f"{drug} - Predicted BMD", xaxis_title="Year", yaxis_title="BMD (g/cm²)",
                xaxis=dict(tickmode='array', tickvals=predictions['Year Index'], ticktext=predictions['Year'])
            )))
        with col2:
            st.plotly_chart(go.Figure(data=[tscore_plot], layout=go.Layout(
                title=f"{drug} - Predicted T-score", xaxis_title="Year", yaxis_title="T-score",
                xaxis=dict(tickmode='array', tickvals=predictions['Year Index'], ticktext=predictions['Year'])
            )))

# Streamlit UI
def main():
    st.title("BMD and T-score Prediction Tool")
    
    # DEXA Machine Selection
    dexa_machine = st.selectbox("DEXA Machine", ["LUNAR"])
    
    # Gender Selection
    gender = st.selectbox("Gender", ["Female"])
    
    # Location (Site) Selection with Mapping
    site_mapping = {
        'Lumbar spine (L1-L4)': 'LS',
        'Femoral Neck': 'FN',
        'Total Hip': 'TH'
    }
    site_options = list(site_mapping.keys())
    selected_site = st.selectbox("Select Region (Site)", site_options)
    site = site_mapping[selected_site]  # Map to the actual value in the dataset
    
    # Input patient data
    bmd_patient = st.number_input(
        "Initial BMD",
        min_value=0.000, max_value=2.000,
        value=0.800, step=0.001,
        format="%.3f"
    )
    
    # Medication Selection with Checkboxes
    st.subheader("Select Medications to Display")
    
    # Add "Show All" Option
    show_all = st.checkbox("Show All Medications")
    
    # Define medications by rows
    medication_rows = [
        ["Alendronate", "Risedronate", "Ibandronate oral"],
        ["Zoledronate", "Ibandronate IV (3mg)"],
        ["Denosumab", "Denosumab + Teriparatide"],
        ["Teriparatide", "Teriparatide + Denosumab"],
        ["Romosozumab", "Romosozumab + Denosumab", "Romosozumab + Alendronate"],
        ["Romosozumab + Ibandronate", "Romosozumab + Zoledronate"]
    ]
    
    # Create checkboxes for each row
    selected_medications = []
    if not show_all:
        for row in medication_rows:
            cols = st.columns(len(row))
            for col, med in zip(cols, row):
                if col.checkbox(med):
                    selected_medications.append(med)
    else:
        # If "Show All" is checked, include all medications
        selected_medications = [med for row in medication_rows for med in row]
    
    # Load constants and medication data
    medication_data = load_medication_data()
    constants = REGRESSION_CONSTANTS.get(selected_site, {})

    # Generate and display predictions for selected medications
    if st.button("Predict"):
        all_predictions = generate_predictions(medication_data, site, bmd_patient, constants['mu'], constants['sigma'])
        filtered_predictions = [pred for pred in all_predictions if pred['Drug'] in selected_medications]
        
        if not filtered_predictions:
            st.warning("No medications selected. Please select at least one medication or use the 'Show All' option.")
        else:
            display_results(filtered_predictions, selected_site)

if __name__ == "__main__":
    main()