src/app.py

import streamlit as st
from tensorflow.keras.models import load_model
import numpy as np
import joblib

def get_user_values():
    # Create a dictionary for the user_values
    user_values = {}

    # Add a title.
    st.title("Heart Disease Detector")

    # Add a slider for age.
    user_values["age"] = st.slider(label="What is your age?", min_value=0, max_value=100)

    # Add a radio button widget for sex.
    user_values["sex"] = st.radio(label="What is your sex?", options=["Male", "Female"])

    # Add a slider for chest pain type.
    user_values["cp"] = st.radio(label="Select a chest pain type:", options=["0 - Typical angina", "1 - Atypical angina", "2 - Non-anginal pain", "3 - Asymptomatic"]) 

    # Add a numeric input widget for resting blood pressure.
    user_values["trestbps"] = st.number_input(label="What is your resting blood pressure?")

    # Add a numeric input widget for total serum cholesterol.
    user_values["chol"] = st.number_input(label="What is your total serum cholesterol (in mg/dl)?")

    # Add a radio button widget for whether there is fasting blood sugar.
    user_values["fbs"] = st.radio(label="Is your fasting blood sugar greater than 120 mg/dl?", options=["Yes", "No"])

    # Add a radio button widget for restecg.
    user_values["restecg"] = st.radio(label="What is your resting electrocardiographic result?", options=["0 — Normal ECG", "1 — ST-T Wave Abnormality", "2 — Left Ventricular Hypertrophy (LVH) by Estes’ Criteria"])
    
    # Add a numeric input widget for maximum heart rate.
    user_values["thalach"] = st.number_input(label="What is your maximum heart rate?")

    # Add a radio button widget for exercise induced angina.
    user_values["exang"] = st.radio(label="Do you have exercise induced angina?", options=["Yes", "No"])

    # Add a numeric input widget for ST depression induced by exercise relative to rest.
    user_values["oldpeak"] = st.number_input(label="Type the ST depression induced by exercise relative to rest:")

    # Add a numeric input widget for the slope of the peak exercise ST segment.
    user_values["slope"] = st.number_input(label="Type the the slope of the peak exercise ST segment:")

    # Add a radio button widget for ca.
    user_values["ca"] = st.radio(label="How many major blood vessels were colored by flourosopy?", options=["0", "1", "2", "3"])

    # Add a radio button widget for thal.
    user_values["thal"] = st.radio(label="What is your thalassemia test result?", options=["1 - Normal", "2 - Fixed Defect", "3 - Reversible Defect"])

    # Return the user's values.
    return user_values

# Method generated by ChatGPT 5.1 on 12/10/25
def translate_user_values(user_values):
    translated = {}

    # Age (already numeric from slider, but cast to int for safety)
    translated["age"] = int(user_values["age"])

    # Sex: 1 = male; 0 = female
    translated["sex"] = 1 if user_values["sex"] == "Male" else 0

    # Chest pain type (cp): just the number before the " - "
    # e.g., "0 - Typical angina" -> 0
    translated["cp"] = int(user_values["cp"].split(" ")[0])

    # Resting blood pressure
    translated["trestbps"] = float(user_values["trestbps"])

    # Serum cholesterol
    translated["chol"] = float(user_values["chol"])

    # Fasting blood sugar > 120 mg/dl: 1 = yes; 0 = no
    translated["fbs"] = 1 if user_values["fbs"] == "Yes" else 0

    # Resting ECG: number before the " — "
    # e.g., "0 — Normal ECG" -> 0
    translated["restecg"] = int(user_values["restecg"].split(" ")[0])

    # Maximum heart rate
    translated["thalach"] = float(user_values["thalach"])

    # Exercise induced angina: 1 = yes; 0 = no
    translated["exang"] = 1 if user_values["exang"] == "Yes" else 0

    # ST depression
    translated["oldpeak"] = float(user_values["oldpeak"])

    # Slope of peak exercise ST segment
    translated["slope"] = float(user_values["slope"])

    # Number of major vessels (ca): radio options are "0", "1", "2", "3"
    translated["ca"] = int(user_values["ca"])

    # Thalassemia:
    thal_raw = int(user_values["thal"].split(" ")[0])
    translated["thal"] = thal_raw

    return translated

# Determine whether the given data indicates the presence of heart disease or not.
def report_heart_disease_verdict(tf_model, X, fitted_stand_scaler, fitted_rob_scaler):
    # Transform the given data with the fitted scalers.
    X = fitted_stand_scaler.transform(X)
    X = fitted_rob_scaler.transform(X)

    # Perform model inference to get a verdict (0 - No heart disease or 1 - Heart disease)
    verdict = tf_model.predict(X)

    # Report the verdict.
    if verdict == 1:
        st.write("Verdict: You likely have a heart disease.")
    else:
        st.write("Verdict: You likely do not have a heart disease.")


def main():
    # Load the custom TensorFlow model that will be used for heart disease detection.
    tf_model = load_model(filepath="models/tf_mlp.keras")

    # Load the fitted Standard Scaler and Robut Scaler from the custom TensorFlow model's training.
    
    # Following two lines of code generated by ChatGPT 5.1 on 12/10/25.
    fitted_stand_scaler = joblib.load("models/standard_scaler.pkl")
    fitted_rob_scaler = joblib.load("models/robust_scaler.pkl")

    # Get the user's values.
    user_values = get_user_values()

    # Get the translated values.
    translated_values = translate_user_values(user_values)

    # Perform feature engineering on the translated values.
    translated_values["age_squared"] = translated_values["age"] * translated_values["age"]
    translated_values["chol_times_trestbps"] = translated_values["chol"] * translated_values["trestbps"]
    translated_values["age_times_thalach"] = translated_values["age"] * translated_values["thalach"]
    translated_values["oldpeak_squared"] = translated_values["oldpeak"] * translated_values["oldpeak"]

    # Convert to the translated values dictionary into a NumPy ndarray.
    translated_values_array = np.array([list(translated_values.values())], dtype=float) # Line of code generated by ChatGPT 5.1 on 12/10/25


    # Predict whether the translated values indicate the presence of heart disease or not.
    want_prediction = st.radio(label="Predict?", options=["No", "Yes"])
    if want_prediction == "Yes":
        report_heart_disease_verdict(tf_model, translated_values_array, fitted_stand_scaler, fitted_rob_scaler)


if __name__ == '__main__':
    main()