app.py

import gradio as gr
import numpy as np
import matplotlib.pyplot as plt

# Function to simulate blood glucose changes over time
def predict_glucose(current_glucose, meal_type, meal_time_hr, meal_time_min, galvus_dose, galvus_time_hr, galvus_time_min, exercise_duration, fast_carbs_ml, chocolate_time_min, prediction_time=3):
    # Constants for glucose reduction effects
    post_meal_reduction = 63.6  # mg/dL (avg reduction for Vildagliptin in 2 hours)
    fasting_reduction = 27.7  # mg/dL (avg reduction over 6-12 hours)
    
    # Convert meal time and Galvus time to minutes for easier calculations
    meal_time_total_min = meal_time_hr * 60 + meal_time_min
    galvus_time_total_min = galvus_time_hr * 60 + galvus_time_min

    # Adjust for fast carbs (milk, juice, etc.)
    carb_effect = fast_carbs_ml * 1.5  # Approximate glucose rise per mL of fast carbs

    # Calculate blood glucose over time considering meal type, Galvus dose, and exercise
    if meal_type == 'High-carb':
        glucose_after_meal = current_glucose + carb_effect + 60  # Higher glucose increase due to carbs
    elif meal_type == 'Protein-heavy':
        glucose_after_meal = current_glucose + 20  # Small glucose increase due to protein
    elif meal_type == 'Low-carb':
        glucose_after_meal = current_glucose - 10  # Small reduction due to low-carb meal
    elif meal_type == 'Fast carb':
        glucose_after_meal = current_glucose + carb_effect * 1.5  # Fast carbs like juice, milk increase glucose quickly
    else:
        glucose_after_meal = current_glucose  # Normal meal with moderate carbs

    # Calculate glucose levels over 1 hour and 3 hours
    glucose_1hr = glucose_after_meal - post_meal_reduction + carb_effect * 0.5  # Adjust for carb effect
    glucose_3hr = glucose_1hr - fasting_reduction

    # Adjust the effects of Galvus based on its administration time (medication effect starts at galvus_time + 1-2 hours)
    time_since_galvus = meal_time_total_min - galvus_time_total_min
    if time_since_galvus >= 60:  # After 1 hour, the effects of Galvus start kicking in
        glucose_3hr -= fasting_reduction  # Galvus effect after 3 hours

    # Exercise effect on glucose (hypothetical value, may vary based on intensity)
    glucose_3hr -= exercise_duration * 2  # Exercise reduces glucose by 2 mg/dL per minute

    # Chocolate bar effect
    if chocolate_time_min <= 60:  # If the chocolate bar is eaten within 1 hour
        glucose_1hr += 40  # Approximate glucose rise from chocolate
        glucose_3hr += 30  # This would still contribute to an increase at 3 hours

    # Plotting the graph of glucose prediction over time
    time_points = [0, 1, 3]  # Time: 0 hours, 1 hour, 3 hours
    glucose_values = [current_glucose, glucose_1hr, glucose_3hr]

    plt.plot(time_points, glucose_values, marker='o', color='b')
    plt.title("Blood Glucose Prediction Over Time")
    plt.xlabel("Time (Hours)")
    plt.ylabel("Blood Glucose (mg/dL)")
    plt.xticks([0, 1, 2, 3])
    plt.grid(True)
    plt.tight_layout()

    # Save the graph as a file to show it in Gradio
    plt.savefig('/tmp/blood_glucose_prediction.png')
    plt.close()

    # Return glucose predictions and the image file path
    return glucose_1hr, glucose_3hr, '/tmp/blood_glucose_prediction.png'


# Gradio Interface
def build_interface():
    with gr.Blocks() as iface:
        gr.Markdown("# Blood Glucose Prediction Model (With Vildagliptin Effects)")

        # Inputs for current glucose, meal info, medication dose, exercise, fast carbs, and Galvus time
        with gr.Row():
            current_glucose = gr.Number(label="Current Blood Glucose (mg/dL)", value=105)
            meal_type = gr.Radio(choices=["Normal", "High-carb", "Protein-heavy", "Low-carb", "Fast carb"], label="Meal Type", value="Low-carb")
            meal_time_hr = gr.Number(label="Last Meal Time (hours)", value=6)
            meal_time_min = gr.Number(label="Last Meal Time (minutes)", value=0)
            galvus_dose = gr.Number(label="Galvus Dose (mg)", value=50)
            galvus_time_hr = gr.Number(label="Galvus Time of Administration (hours)", value=2)
            galvus_time_min = gr.Number(label="Galvus Time of Administration (minutes)", value=0)
            exercise_duration = gr.Number(label="Exercise Duration (min)", value=60)
            fast_carbs_ml = gr.Number(label="Fast Carbs (mL)", value=0)
            chocolate_time_min = gr.Number(label="Time After Meal When Chocolate Bar Is Eaten (min)", value=60)

        # Output predictions and graph
        glucose_1hr_output = gr.Textbox(label="Predicted Glucose Level in 1 Hour (mg/dL)")
        glucose_3hr_output = gr.Textbox(label="Predicted Glucose Level in 3 Hours (mg/dL)")
        glucose_graph = gr.Image(label="Blood Glucose Prediction Graph")

        # Button to trigger prediction
        predict_button = gr.Button("Predict Blood Glucose")

        # Set button action
        predict_button.click(
            predict_glucose,
            inputs=[current_glucose, meal_type, meal_time_hr, meal_time_min, galvus_dose, galvus_time_hr, galvus_time_min, exercise_duration, fast_carbs_ml, chocolate_time_min],
            outputs=[glucose_1hr_output, glucose_3hr_output, glucose_graph]
        )

    return iface


# Build and launch the Gradio interface
iface = build_interface()
iface.launch()