Update app.py

app.py

@@ -1,218 +1,240 @@
-# app.py
 import streamlit as st
-import numpy as np
-import os
 import tensorflow as tf
-import logging
+import numpy as np
+import joblib
+import json
 from PIL import Image
+import pandas as pd
+from lifelines import CoxPHFitter
 
-# Configure logging
-logging.basicConfig(level=logging.INFO)
-logger = logging.getLogger(__name__)
+# ---------------------------------------------------
+# CONFIG
+# ---------------------------------------------------
 
-# Set page configuration
 st.set_page_config(
-    page_title="Breast Cancer Prediction",
-    page_icon="🩺",
-    layout="wide",
-    initial_sidebar_state="expanded"
+    page_title="Breast Cancer Survival Prediction",
+    page_icon="🧬",
+    layout="wide"
 )
 
-# Disable GPU to save memory
-tf.config.set_visible_devices([], 'GPU')
-logger.info("TensorFlow configured for CPU-only")
-
-# ===== Model Loading =====
-MODEL_FILE = "final_combined_model.keras"
-
-@st.cache_resource(show_spinner=False)
-def load_model():
-    """Load TensorFlow model from local file with caching"""
-    try:
-        # Verify file exists
-        if not os.path.exists(MODEL_FILE):
-            logger.error(f"❌ Model file not found: {MODEL_FILE}")
-            return None
-        
-        logger.info(f"⏳ Loading model from local file: {MODEL_FILE}")
-        
-        # Load model with memory optimization
-        model = tf.keras.models.load_model(MODEL_FILE, compile=False)
-        
-        # Test prediction to verify loading
-        test_input = np.random.rand(1, 224, 224, 1).astype(np.float32)
-        test_pred = model.predict(test_input, verbose=0)
-        logger.info(f"🧪 Test prediction: {test_pred[0][0]:.4f}")
-        
-        logger.info("✅ Model loaded successfully")
-        return model
-    except Exception as e:
-        logger.error(f"❌ Error loading model: {e}")
-        # Print detailed traceback
-        import traceback
-        logger.error(traceback.format_exc())
-        return None
-
-# Load model at startup
-model = load_model()
-
-# ===== Image Preprocessing =====
+# CNN_MODEL_PATH = "best_breast_cancer_cnn.keras"
+CNN_MODEL_PATH = "final_combined_model.keras"
+DNN_MODEL_PATH = "survival_model.keras"
+
+SCALER_PATH = "scaler.pkl"
+FEATURES_PATH = "features.json"
+
+DATASET_PATH = 'processed_breast_cancer_data(1).csv'
+TIME_COL = "Overall_Survival_Months"
+EVENT_COL = "Event"
+ID_COL = "Patient_ID"
+
+# ---------------------------------------------------
+# LOAD MODELS
+# ---------------------------------------------------
+
+@st.cache_resource
+def load_cnn():
+    return tf.keras.models.load_model(CNN_MODEL_PATH, compile=False)
+
+@st.cache_resource
+def load_dnn():
+    return tf.keras.models.load_model(DNN_MODEL_PATH, compile=False)
+
+# ---------------------------------------------------
+# LOAD SURVIVAL ASSETS (COMPUTE BRESLOW BASELINE)
+# ---------------------------------------------------
+
+@st.cache_resource
+def load_survival_assets():
+
+    scaler = joblib.load(SCALER_PATH)
+    features = json.load(open(FEATURES_PATH))
+
+    df = pd.read_csv(DATASET_PATH)
+
+    feature_df = df[features].copy()
+    feature_df["duration"] = df[TIME_COL]
+    feature_df["event"] = df[EVENT_COL]
+
+    cox = CoxPHFitter()
+    cox.fit(feature_df, duration_col="duration", event_col="event")
+
+    baseline = cox.baseline_cumulative_hazard_
+
+    breslow_times = baseline.index.values
+    breslow_H0 = baseline.values.flatten()
+
+    return scaler, features, breslow_times, breslow_H0
+
+
+cnn_model = load_cnn()
+dnn_model = load_dnn()
+scaler, feature_cols, breslow_times, breslow_H0 = load_survival_assets()
+
+# ---------------------------------------------------
+# IMAGE PREPROCESSING
+# ---------------------------------------------------
+
 def preprocess_image(image):
-    """Preprocess image for model prediction"""
-    try:
-        # Convert to PIL Image
-        if isinstance(image, np.ndarray):
-            img = Image.fromarray(image.astype('uint8'))
-        else:
-            img = image
-        
-        # Processing pipeline
-        img = img.convert('L')        # Grayscale
-        img = img.resize((224, 224))  # Resize
-        img_array = np.array(img) / 255.0  # Normalize
-        
-        # Add batch and channel dimensions
-        return img_array[np.newaxis, ..., np.newaxis]
-    except Exception as e:
-        logger.error(f"🖼️ Image preprocessing error: {e}")
-        return None
-
-# ===== Prediction Function =====
-def predict(image):
-    """Make prediction using the loaded model"""
-    if model is None:
-        return "Model failed to load", "Check logs", None
-    
-    try:
-        # Preprocess image
-        processed_image = preprocess_image(image)
-        if processed_image is None:
-            return "Invalid image", "Try another", image
-        
-        # Make prediction
-        prediction = model.predict(processed_image, verbose=0)[0][0]
-        
-        # Format results
-        confidence = abs(prediction - 0.5) + 0.5  # Convert to 0.5-1.0 scale
-        result = "Malignant" if prediction > 0.5 else "Benign"
-        
-        return result, f"{confidence*100:.2f}%", image
-    except Exception as e:
-        error_msg = f"Prediction error: {str(e)}"
-        logger.error(error_msg)
-        return error_msg, "Try again", image
-
-# ===== Streamlit UI =====
-
-# Custom CSS for styling
-st.markdown("""
-<style>
-    .stApp {
-        background-color: #f0f2f6;
-    }
-    .header {
-        color: #2c3e50;
-        text-align: center;
-        padding: 1rem;
-    }
-    .result-box {
-        border-radius: 10px;
-        padding: 1.5rem;
-        margin: 1rem 0;
-        box-shadow: 0 4px 6px rgba(0,0,0,0.1);
-    }
-    .malignant {
-        background-color: #ffcccc;
-        border-left: 5px solid #e74c3c;
-    }
-    .benign {
-        background-color: #ccffcc;
-        border-left: 5px solid #2ecc71;
-    }
-    .stButton>button {
-        background-color: #3498db;
-        color: white;
-        border-radius: 5px;
-        padding: 0.5rem 1rem;
-        width: 100%;
-    }
-    .stButton>button:hover {
-        background-color: #2980b9;
-    }
-</style>
-""", unsafe_allow_html=True)
-
-# Header
-st.markdown("<h1 class='header'>🩺 Breast Cancer Prediction</h1>", unsafe_allow_html=True)
-st.markdown("Upload a breast medical image for cancer prediction")
-
-# Status indicator
-status = "✅ Model loaded successfully" if model else "❌ Model failed to load"
-st.info(status)
-
-# Create two columns for layout
-col1, col2 = st.columns([1, 1])
-
-# Input column
-with col1:
-    st.subheader("Patient Information")
-    
-    # Input fields
-    age = st.number_input("Patient Age", min_value=18, max_value=100, value=45)
-    tumor_size = st.number_input("Tumor Size (mm)", min_value=0.1, value=15.0)
-    
-    # Image upload
-    uploaded_file = st.file_uploader(
-        "Upload Medical Image", 
-        type=["jpg", "jpeg", "png"],
-        help="Supported formats: JPG, JPEG, PNG"
+
+    if image.mode != "L":
+        image = image.convert("L")
+
+    image = image.resize((224, 224))
+
+    img = np.array(image) / 255.0
+    img = img[np.newaxis, ..., np.newaxis]
+
+    return img
+
+# ---------------------------------------------------
+# CNN PREDICTION
+# ---------------------------------------------------
+
+def predict_cancer(image):
+
+    img = preprocess_image(image)
+
+    pred = cnn_model.predict(img, verbose=0)[0][0]
+
+    result = "Malignant" if pred > 0.5 else "Benign"
+
+    confidence = pred if pred > 0.5 else 1 - pred
+
+    return result, confidence, pred
+
+# ---------------------------------------------------
+# SURVIVAL FUNCTION
+# ---------------------------------------------------
+
+def survival_prob(risk, t):
+
+    idx = np.searchsorted(breslow_times, t, side="right") - 1
+
+    if idx < 0:
+        return 1.0
+
+    h0 = breslow_H0[idx]
+
+    return float(np.exp(-h0 * np.exp(risk)))
+
+# ---------------------------------------------------
+# SURVIVAL PREDICTION
+# ---------------------------------------------------
+
+def predict_survival(feature_values):
+
+    row = np.array([feature_values], dtype=np.float32)
+
+    row = scaler.transform(row)
+
+    risk = float(dnn_model.predict(row, verbose=0)[0][0])
+
+    s1 = survival_prob(risk, 12) * 100
+    s3 = survival_prob(risk, 36) * 100
+    s5 = survival_prob(risk, 60) * 100
+
+    return risk, s1, s3, s5
+
+# ---------------------------------------------------
+# UI
+# ---------------------------------------------------
+
+st.title("🧬 Breast Cancer AI Diagnosis & Survival System")
+
+st.markdown(
+"""
+This system integrates two AI models:
+
+• **CNN model** → detects tumor malignancy from medical images  
+• **Survival DNN** → predicts patient survival probabilities
+"""
+)
+
+tab1, tab2 = st.tabs(["🔬 Image Diagnosis", "📈 Survival Analysis"])
+
+# ---------------------------------------------------
+# TAB 1 : IMAGE PREDICTION
+# ---------------------------------------------------
+
+with tab1:
+
+    st.header("Tumor Image Classification")
+
+    uploaded = st.file_uploader(
+        "Upload Histopathology Image",
+        type=["png", "jpg", "jpeg"]
     )
-    
-    # Predict button
-    predict_btn = st.button("Analyze Image")
-
-# Results column
-with col2:
-    st.subheader("Prediction Results")
-    
-    # Initialize session state for results
-    if 'result' not in st.session_state:
-        st.session_state.result = None
-        st.session_state.confidence = None
-        st.session_state.image = None
-    
-    # Process image when button is clicked
-    if predict_btn and uploaded_file is not None:
-        try:
-            image = Image.open(uploaded_file)
-            st.session_state.result, st.session_state.confidence, st.session_state.image = predict(image)
-        except Exception as e:
-            st.error(f"Error processing image: {str(e)}")
-    
-    # Display results if available
-    if st.session_state.result:
-        # Result box with color coding
-        result_class = "malignant" if st.session_state.result == "Malignant" else "benign"
-        st.markdown(
-            f"<div class='result-box {result_class}'>"
-            f"<h3>Diagnosis: {st.session_state.result}</h3>"
-            f"<p>Confidence: {st.session_state.confidence}</p>"
-            "</div>", 
-            unsafe_allow_html=True
+
+    if uploaded:
+
+        image = Image.open(uploaded)
+
+        st.image(image, width=300)
+
+        if st.button("Analyze Image"):
+
+            result, conf, score = predict_cancer(image)
+
+            st.subheader("Prediction Result")
+
+            col1, col2 = st.columns(2)
+
+            col1.metric("Diagnosis", result)
+
+            col2.metric("Confidence", f"{conf*100:.2f}%")
+
+            st.write("Prediction Score:", round(score, 4))
+
+# ---------------------------------------------------
+# TAB 2 : SURVIVAL ANALYSIS
+# ---------------------------------------------------
+
+with tab2:
+
+    st.header("Patient Survival Prediction")
+
+    st.write("Enter patient clinical features")
+
+    inputs = []
+
+    cols = st.columns(3)
+
+    for i, f in enumerate(feature_cols):
+
+        value = cols[i % 3].number_input(
+            f,
+            value=0.0,
+            step=0.1
         )
-        
-        # Display image
-        if st.session_state.image:
-            st.image(
-                st.session_state.image, 
-                caption="Uploaded Image",
-                use_container_width=True
-            )
-    
-    # Show placeholder if no results
-    elif not predict_btn:
-        st.info("Upload an image and click 'Analyze Image' to get prediction")
-
-# Footer
+
+        inputs.append(value)
+
+    if st.button("Predict Survival"):
+
+        risk, s1, s3, s5 = predict_survival(inputs)
+
+        st.subheader("Risk Score")
+
+        st.metric("Risk Score", round(risk, 4))
+
+        st.subheader("Survival Probability")
+
+        c1, c2, c3 = st.columns(3)
+
+        c1.metric("1-Year Survival", f"{s1:.1f}%")
+        c2.metric("3-Year Survival", f"{s3:.1f}%")
+        c3.metric("5-Year Survival", f"{s5:.1f}%")
+
+        if risk > 0:
+            st.error("High Risk Category")
+        else:
+            st.success("Low Risk Category")
+
+# ---------------------------------------------------
+# FOOTER
+# ---------------------------------------------------
+
 st.markdown("---")
-st.caption("This tool is for research purposes only. Consult a medical professional for clinical diagnosis.")
+st.caption("AI-assisted clinical decision support system")