app.py

import pickle

import numpy as np
import pandas as pd
import streamlit as st
import tensorflow as tf

st.set_page_config(
    page_title="Heart Disease Risk Prediction",
    page_icon="heart",
    layout="wide",
)

RISK_LABELS = ["Low", "Medium", "High"]

SEX_OPTIONS = {"Female": 0, "Male": 1}
CP_OPTIONS = {
    "Typical angina": 0,
    "Atypical angina": 1,
    "Non-anginal pain": 2,
    "Asymptomatic": 3,
}
FBS_OPTIONS = {"False": 0, "True": 1}
RESTECG_OPTIONS = {
    "Normal": 0,
    "ST-T wave abnormality": 1,
    "Left ventricular hypertrophy": 2,
}
EXANG_OPTIONS = {"No": 0, "Yes": 1}
SLOPE_OPTIONS = {"Upsloping": 0, "Flat": 1, "Downsloping": 2}
CA_OPTIONS = {"0 vessels": 0, "1 vessel": 1, "2 vessels": 2, "3 vessels": 3, "4 vessels": 4}
THAL_OPTIONS = {"Fixed defect": 1, "Normal": 2, "Reversible defect": 3}


@st.cache_resource
def load_artifacts():
    model = tf.keras.models.load_model("heart_disease_ann.h5")
    with open("preprocessor.pkl", "rb") as f:
        preprocessor = pickle.load(f)
    return model, preprocessor


def pick_code(label, options, help_text=None):
    return options[st.selectbox(label, list(options.keys()), help=help_text)]


try:
    model, preprocessor = load_artifacts()
except Exception as exc:
    st.error(f"Could not load model files: {exc}")
    st.stop()

st.title("Heart Disease Risk Prediction")
st.caption(
    "Enter patient details, run prediction, and review risk probabilities. "
    "This tool supports screening and is not a medical diagnosis."
)

tab_predict, tab_guide = st.tabs(["Prediction", "Input Guide"])

with tab_predict:
    with st.form(key="heart_form"):
        col_a, col_b = st.columns(2)

        with col_a:
            st.subheader("Patient Profile")
            age = st.slider("Age", min_value=1, max_value=120, value=30)
            sex = pick_code("Sex", SEX_OPTIONS, "Male = 1, Female = 0")
            cp = pick_code("Chest Pain Type", CP_OPTIONS)
            exang = pick_code("Exercise Induced Angina", EXANG_OPTIONS, "Yes = 1, No = 0")
            slope = pick_code("Slope of ST Segment", SLOPE_OPTIONS)
            thal = pick_code("Thalassemia", THAL_OPTIONS)

        with col_b:
            st.subheader("Clinical Measurements")
            trestbps = st.slider("Resting Blood Pressure (mm Hg)", min_value=80, max_value=220, value=120)
            chol = st.slider("Cholesterol (mg/dl)", min_value=100, max_value=600, value=200)
            fbs = pick_code("Fasting Blood Sugar > 120 mg/dl?", FBS_OPTIONS, "True = 1, False = 0")
            restecg = pick_code("Resting ECG Result", RESTECG_OPTIONS)
            thalach = st.slider("Max Heart Rate Achieved", min_value=60, max_value=220, value=150)
            oldpeak = st.number_input("ST Depression (oldpeak)", min_value=0.0, max_value=10.0, value=1.0, step=0.1)
            ca = pick_code("Major Vessels Colored by Fluoroscopy", CA_OPTIONS)

        submit_button = st.form_submit_button("Predict Risk", use_container_width=True)

    if submit_button:
        input_df = pd.DataFrame(
            [
                {
                    "age": age,
                    "sex": sex,
                    "cp": cp,
                    "trestbps": trestbps,
                    "chol": chol,
                    "fbs": fbs,
                    "restecg": restecg,
                    "thalach": thalach,
                    "exang": exang,
                    "oldpeak": oldpeak,
                    "slope": slope,
                    "ca": ca,
                    "thal": thal,
                }
            ]
        )

        input_processed = preprocessor.transform(input_df)
        pred_probs = model.predict(input_processed, verbose=0)[0]
        pred_class = RISK_LABELS[int(np.argmax(pred_probs))]

        if pred_class == "Low":
            st.success(f"Predicted Risk Level: {pred_class}")
        elif pred_class == "Medium":
            st.warning(f"Predicted Risk Level: {pred_class}")
        else:
            st.error(f"Predicted Risk Level: {pred_class}")

        st.subheader("Probability Scores")
        metric_cols = st.columns(3)
        for idx, label in enumerate(RISK_LABELS):
            metric_cols[idx].metric(label, f"{pred_probs[idx] * 100:.2f}%")

        prob_df = pd.DataFrame(
            {"Risk": RISK_LABELS, "Probability": (pred_probs * 100).round(2)}
        ).set_index("Risk")
        st.bar_chart(prob_df)

        with st.expander("Show Submitted Values"):
            st.dataframe(input_df, hide_index=True, use_container_width=True)

with tab_guide:
    st.subheader("Category Encoding Reference")
    st.write("Sex: Female = 0, Male = 1")
    st.write("Chest Pain Type: Typical angina = 0, Atypical angina = 1, Non-anginal pain = 2, Asymptomatic = 3")
    st.write("Fasting Blood Sugar > 120 mg/dl: False = 0, True = 1")
    st.write(
        "Resting ECG Result: Normal = 0, ST-T wave abnormality = 1, "
        "Left ventricular hypertrophy = 2"
    )
    st.write("Exercise Induced Angina: No = 0, Yes = 1")
    st.write("Slope of ST Segment: Upsloping = 0, Flat = 1, Downsloping = 2")
    st.write("Major Vessels Colored by Fluoroscopy: Number of vessels from 0 to 4")
    st.write("Thalassemia: Fixed defect = 1, Normal = 2, Reversible defect = 3")