app.py

import streamlit as st
import numpy as np
import pandas as pd
import joblib
import shap
import matplotlib.pyplot as plt
import seaborn as sns

st.set_page_config(page_title="Heart Hero", layout="wide")

model = joblib.load("heart_model.pkl")
scaler = joblib.load("scaler.pkl")
columns = joblib.load("columns.pkl")
raw_df = joblib.load("raw_data.pkl")

st.title("❤️ Heart-Hero: Heart Disease Predictor")
st.warning("⚠️ Disclaimer: This application is for educational purposes only and may not be fully accurate enough. Please consult a qualified healthcare professional for medical advice.")

df = pd.read_csv("dataset/dataset.csv")

left_col, right_col = st.columns([1, 2])

with left_col:
    st.subheader("🧾 Patient Input")

    age = st.number_input("Age", 1, 120, 50)
    sex = st.selectbox("Sex (0 = Female, 1 = Male)", [0, 1])
    cp = st.selectbox("Chest Pain Type (0–3)", [0, 1, 2, 3])
    trestbps = st.number_input("Resting Blood Pressure", 80, 200, 120)
    chol = st.number_input("Cholesterol", 100, 400, 200)
    fbs = st.selectbox("Fasting Blood Sugar > 120 mg/dl", [0, 1])
    restecg = st.selectbox("Resting ECG (0–2)", [0, 1, 2])
    thalach = st.number_input("Max Heart Rate", 60, 220, 150)
    exang = st.selectbox("Exercise Induced Angina", [0, 1])
    oldpeak = st.number_input("ST Depression", 0.0, 6.0, 1.0)
    slope = st.selectbox("Slope (0–2)", [0, 1, 2])
    ca = st.selectbox("Major Vessels (0–4)", [0, 1, 2, 3, 4])
    thal = st.selectbox("Thal (0–3)", [0, 1, 2, 3])

    predict_btn = st.button("🔍 Predict")

input_dict = {
    '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_df = pd.DataFrame([input_dict])
input_df = pd.get_dummies(input_df)
input_df = input_df.reindex(columns=columns, fill_value=0)

scale_cols = ['age','trestbps','chol','thalach','oldpeak']
input_df[scale_cols] = scaler.transform(input_df[scale_cols])

with right_col:
    st.subheader("📊 Prediction Result")

    if predict_btn:
        prediction = model.predict(input_df)[0]
        probability = model.predict_proba(input_df)[0][1]

        if probability >= 0.5:
            st.error(f"Heart Disease Detected (Probability: {probability:.2f})")
        else:
            st.success(f"No Heart Disease Detected (Probability: {probability:.2f})")

        st.subheader("🧠 Patient Feature Contributions")

        explainer = shap.TreeExplainer(model)
        shap_values = explainer.shap_values(input_df)
        
        if isinstance(shap_values, list):
            shap_val = shap_values[1][0]  
        else:
            shap_val = shap_values[0]

        feature_names = list(input_df.columns)
        shap_val_flat = np.asarray(shap_val).ravel()
        input_val_list = input_df.iloc[0].tolist()
        min_len = min(len(feature_names), len(shap_val_flat), len(input_val_list))
        shap_df = pd.DataFrame({
            'feature': feature_names[:min_len],
            'shap_value': shap_val_flat[:min_len],
            'input_value': input_val_list[:min_len]
        })

        def active_feature_name(feat):
            if "_" in feat:
                base, val = feat.split("_")
                val = int(val)
                if input_dict.get(base) == val:
                    return base
                else:
                    return None
            else:
                return feat

        shap_df['original'] = shap_df['feature'].apply(active_feature_name)
        shap_df_clean = shap_df.dropna(subset=['original'])

        agg_shap = shap_df_clean.groupby('original')['shap_value'].sum().sort_values(key=abs, ascending=False)

        st.write("**Top 5 Factors Contributing to Heart Disease Risk:**")
        for feat, val in agg_shap.head(5).items():
            direction = "increased" if val > 0 else "reduced"
            st.write(f"• **{feat}** contributed **{val:+.2f}** to {direction} heart disease probability")

        st.subheader("📈 Feature Impact")
        top5 = agg_shap.head(5)
        fig, ax = plt.subplots(figsize=(10, 5))
        top5.plot(kind="barh", ax=ax, color=['red' if x > 0 else 'green' for x in top5.values])
        ax.set_xlabel("SHAP Contribution (Impact on Prediction)")
        ax.set_title("Top 5 Features Influencing Heart Disease Prediction")
        st.pyplot(fig)

        st.subheader("Clinical Interpretation")

        for feat, val in agg_shap.head(5).items():
            user_val = input_dict.get(feat, "N/A")
            
            if feat == "chol":
                msg = "High cholesterol is associated with increased heart disease risk."
            elif feat == "thalach":
                msg = "Lower maximum heart rate during exercise may indicate cardiac issues."
            elif feat == "trestbps":
                msg = "High resting blood pressure increases strain on the heart."
            elif feat == "age":
                msg = "Older age is associated with increased cardiovascular risk."
            elif feat == "sex":
                msg = "Males historically show higher heart disease prevalence in the dataset."
            elif feat == "cp":
                msg = "Certain chest pain types are linked with higher cardiac risk."
            elif feat == "oldpeak":
                msg = "ST depression induced by exercise relative to rest (higher = more cardiac stress)."
            elif feat == "exang":
                msg = "Exercise-induced angina is a strong indicator of heart disease."
            elif feat == "fbs":
                msg = "Elevated fasting blood sugar indicates metabolic risk."
            elif feat == "restecg":
                msg = "Abnormal resting ECG findings suggest cardiac abnormalities."
            elif feat == "slope":
                msg = "The slope of ST segment during exercise is clinically significant."
            elif feat == "ca":
                msg = "Number of major vessels with calcification indicates atherosclerosis severity."
            elif feat == "thal":
                msg = "Thalassemia type affects oxygen transport and heart function."
            else:
                msg = "This feature significantly influenced the prediction."

            st.write(f"**{feat} = {user_val}** → {msg}")

        st.subheader("📥 Export Report")

        report_text = "Heart Disease Prediction Report\n\n"
        report_text += "=" * 50 + "\n"
        report_text += f"Prediction: {'Heart Disease Detected' if prediction == 1 else 'No Heart Disease'}\n"
        report_text += f"Confidence: {probability:.2%}\n"
        report_text += "=" * 50 + "\n\n"
        report_text += "Patient Input Values:\n"
        report_text += "-" * 50 + "\n"
        for key, value in input_dict.items():
            report_text += f"{key}: {value}\n"
        report_text += f"\nPredicted Probability of Heart Disease: {probability:.2%}\n\n"
        report_text += "Top 5 Contributing Factors:\n"
        report_text += "-" * 50 + "\n"
        for feat, val in agg_shap.head(5).items():
            direction = "increased" if val > 0 else "reduced"
            report_text += f"• {feat}: contributed {val:+.2f} to {direction} heart disease probability\n"
        report_text += "\n" + "=" * 50 + "\n"
        report_text += "DISCLAIMER:\n"
        report_text += "This result is generated by a machine learning model and is for\n"
        report_text += "educational purposes only. It is NOT a medical diagnosis.\n"
        report_text += "Please consult with a qualified healthcare professional for\n"
        report_text += "accurate medical evaluation and diagnosis.\n"
        report_text += "=" * 50 + "\n"

        st.download_button(
            label="Download Prediction Report",
            data=report_text,
            file_name="heart_disease_report.txt",
            mime="text/plain"
        )

st.markdown("---")
st.subheader("Dataset Preview")
st.write(df.head())

st.subheader("Target Distribution")
fig1 = plt.figure()
sns.countplot(x='target', data=df)
st.pyplot(fig1)

st.subheader("Correlation Heatmap")
fig2 = plt.figure(figsize=(8,6))
sns.heatmap(df.corr(), cmap="coolwarm", annot=True)
st.pyplot(fig2)