Update app.py

Author: VenkateshHJoshi

app.py

@@ -1,20 +1,19 @@
 # ============================================================
-# Imports & global config
+# Imports & configuration
 # ============================================================
 import streamlit as st
 import pandas as pd
-import joblib
 import numpy as np
+import joblib
 import plotly.graph_objects as go
-import warnings
 from sklearn import set_config
+import warnings
 
-# Keep pandas through pipeline (feature-name safety)
 set_config(transform_output="pandas")
 warnings.filterwarnings("ignore")
 
 # ============================================================
-# Load trained pipeline
+# Load trained pipeline bundle
 # ============================================================
 @st.cache_resource
 def load_pipeline():
@@ -25,7 +24,7 @@ def load_pipeline():
 feature_names = bundle["feature_names"]
 
 # ============================================================
-# Human-readable class labels
+# Class names (UNCHANGED – same as training)
 # ============================================================
 CLASS_MAP = {
     0: "Healthy Liver",
@@ -36,7 +35,7 @@ def load_pipeline():
 }
 
 # ============================================================
-# Medical-grade UI configuration
+# UI metadata for inputs
 # ============================================================
 FEATURE_UI = {
     "age": {"label": "Age (years)", "min": 1, "max": 100, "default": 32},
@@ -48,22 +47,21 @@ def load_pipeline():
     "cholinesterase": {"label": "Cholinesterase (U/L)", "min": 2000, "max": 12000, "default": 7000},
     "cholesterol": {"label": "Cholesterol (mg/dL)", "min": 80, "max": 400, "default": 170},
     "creatinina": {"label": "Creatinine (mg/dL)", "min": 0.3, "max": 5.0, "default": 0.9},
-    "gamma glutamyl transferase": {"label": "GGT – Gamma GT (U/L)", "min": 5, "max": 300, "default": 30},
+    "gamma glutamyl transferase": {"label": "Gamma GT (U/L)", "min": 5, "max": 300, "default": 30},
     "protein": {"label": "Total Protein (g/dL)", "min": 4.0, "max": 9.0, "default": 7.2},
 }
 
 # ============================================================
-# Page config
+# Page setup
 # ============================================================
 st.set_page_config(
-    page_title="Liver Health AI",
+    page_title="Liver Disease Pattern Analysis",
     page_icon="🩺",
     layout="centered"
 )
 
-st.title("🩺 Liver Health AI Dashboard")
-st.caption("Clinical inputs → ML pipeline → visual decision support")
-
+st.title("🩺 Liver Disease Pattern Analysis System")
+st.caption("ML-based pattern similarity & risk leaning (not a diagnosis)")
 st.divider()
 
 # ============================================================
@@ -78,151 +76,130 @@ def load_pipeline():
     key = feature.lower()
 
     with cols[i % 2]:
-
-        # Sex (human-friendly)
         if key in ["sex", "gender"]:
             sex = st.selectbox("Sex", ["Male", "Female"])
             input_data[feature] = 1 if sex == "Male" else 0
-
-        # Numeric features
         else:
             ui = FEATURE_UI.get(key, {})
             input_data[feature] = st.number_input(
                 label=ui.get("label", feature.replace("_", " ").title()),
                 min_value=ui.get("min", 0.0),
                 max_value=ui.get("max", 1000.0),
                 value=ui.get("default", 1.0),
-                help="Enter raw lab value (no transformation needed)"
             )
 
-# Preserve feature order
 input_df = pd.DataFrame([input_data], columns=feature_names)
 
 # ============================================================
-# PREDICTION
+# ANALYSIS
 # ============================================================
 st.divider()
 
-if st.button("🔍 Analyze Liver Health"):
+if st.button("🔍 Analyze Pattern"):
 
     probs = pipeline.predict_proba(input_df)[0]
     pred_class = int(np.argmax(probs))
     confidence = float(probs[pred_class])
-    disease = CLASS_MAP[pred_class]
-
-    # Risk bucket
-    if confidence < 0.45:
-        risk = "Low"
-        risk_color = "green"
-    elif confidence < 0.75:
-        risk = "Medium"
-        risk_color = "orange"
-    else:
-        risk = "High"
-        risk_color = "red"
+
+    st.success(f"### 🧾 Primary Model Output: **{CLASS_MAP[pred_class]}**")
+    st.info(f"Model Confidence: **{confidence:.2%}**")
+
+    st.divider()
+    st.subheader("📊 Disease Pattern Leaning")
 
     # ========================================================
-    # RESULT SUMMARY
+    # 1️⃣ Disease Leaning Chart
     # ========================================================
-    st.success(f"### 🧾 Diagnosis: **{disease}**")
-    st.info(f"Confidence: **{confidence:.2%}** | Risk Level: **{risk}**")
+    focus = {
+        "Healthy Liver": 0,
+        "Cirrhosis": 1,
+        "Hepatitis": 2,
+        "Fibrosis": 3,
+        "Suspected Liver Disorder": 4,
+    }
+
+    labels = list(focus.keys())
+    values = [probs[focus[l]] for l in labels]
+
+    fig = go.Figure(
+        data=[
+            go.Bar(
+                x=labels,
+                y=values,
+                marker_color=[
+                    "#2ecc71",  # Healthy
+                    "#e74c3c",  # Cirrhosis
+                    "#f39c12",  # Hepatitis
+                    "#d35400",  # Fibrosis
+                    "#9b59b6",  # Suspected
+                ]
+            )
+        ]
+    )
 
-    st.divider()
-    st.subheader("📊 Visual Health Analysis")
+    fig.update_layout(
+        title="Disease Pattern Similarity (Probability Distribution)",
+        yaxis=dict(range=[0, 1], title="Probability"),
+        xaxis_title="Disease Pattern",
+    )
+
+    st.plotly_chart(fig, use_container_width=True)
 
     # ========================================================
-    # DASHBOARD (2 × 2)
+    # 2️⃣ Lean Strength toward Fibrosis / Suspected
     # ========================================================
-    col1, col2 = st.columns(2)
-
-    # ---- Chart 1: Risk Gauge ----
-    with col1:
-        fig1 = go.Figure(go.Indicator(
-            mode="gauge+number",
-            value=confidence * 100,
-            number={"suffix": "%"},
-            title={"text": "Overall Risk Confidence"},
-            gauge={
-                "axis": {"range": [0, 100]},
-                "bar": {"color": risk_color},
-                "steps": [
-                    {"range": [0, 45], "color": "#b6f2c2"},
-                    {"range": [45, 75], "color": "#ffeaa7"},
-                    {"range": [75, 100], "color": "#fab1a0"},
-                ],
-            },
-        ))
-        st.plotly_chart(fig1, use_container_width=True)
-
-    # ---- Chart 2: Top-2 Probabilities ----
-    with col2:
-        top_idx = np.argsort(probs)[-2:][::-1]
-        fig2 = go.Figure(
-            data=[
-                go.Bar(
-                    x=[CLASS_MAP[i] for i in top_idx],
-                    y=probs[top_idx],
-                    marker_color=["#0984e3", "#6c5ce7"]
-                )
-            ]
-        )
-        fig2.update_layout(
-            title="Most Likely Conditions",
-            yaxis=dict(range=[0, 1], title="Probability"),
-        )
-        st.plotly_chart(fig2, use_container_width=True)
-
-    col3, col4 = st.columns(2)
-
-    # ---- Chart 3: Feature Profile ----
-    with col3:
-        values = np.array([float(input_data[f]) for f in feature_names])
-        normalized = values / (np.mean(values) + 1e-6)
-
-        fig3 = go.Figure(
-            data=[go.Bar(x=list(range(len(normalized))), y=normalized)]
-        )
-        fig3.add_hline(y=1, line_dash="dash")
-        fig3.update_layout(
-            title="Overall Feature Profile (Relative Scale)",
-            xaxis_title="Feature Index",
-            yaxis_title="Relative Magnitude",
-        )
-        st.plotly_chart(fig3, use_container_width=True)
-
-    # ---- Chart 4: Top Feature Deviations ----
-    with col4:
-        deviations = np.abs(normalized - 1.0)
-        top_k = np.argsort(deviations)[-5:][::-1]
-
-        fig4 = go.Figure(
-            data=[
-                go.Bar(
-                    x=[feature_names[i].replace("_", " ") for i in top_k],
-                    y=deviations[top_k],
-                    marker_color="#d63031"
-                )
-            ]
-        )
-        fig4.update_layout(
-            title="Top Influential Feature Deviations",
-            yaxis_title="Deviation Strength",
-        )
-        st.plotly_chart(fig4, use_container_width=True)
+    fibrosis_prob = probs[3]
+    suspected_prob = probs[4]
+    lean_strength = fibrosis_prob + suspected_prob
+
+    st.subheader("📌 Chronic Pattern Lean Strength")
+
+    st.metric(
+        label="Fibrosis + Suspected Similarity",
+        value=f"{lean_strength:.2%}"
+    )
 
     # ========================================================
-    # AI INTERPRETATION
+    # 3️⃣ Recommendation Engine (Probability-based)
     # ========================================================
-    st.subheader("💬 AI Interpretation")
+    st.subheader("🩺 Clinical Recommendation (Decision Support)")
+
     with st.chat_message("assistant"):
-        st.write(
-            f"The model predicts **{disease}** with **{confidence:.2%} confidence**. "
-            f"The risk level is **{risk.lower()}**, based on learned interactions across multiple biomarkers. "
-            "This enables early detection even when individual lab values appear normal."
-        )
+
+        if lean_strength >= 0.80:
+            st.write(
+                "🔴 **Strong leaning toward chronic liver disease patterns detected.**\n\n"
+                "The patient's biomarker profile closely resembles historical patterns "
+                "associated with **fibrosis or early chronic liver disorders**.\n\n"
+                "✅ **Recommendation:** Prompt clinical evaluation is advised, including "
+                "advanced liver assessment (e.g., FibroScan, imaging, specialist referral)."
+            )
+
+        elif lean_strength >= 0.50:
+            st.write(
+                "🟠 **Moderate leaning toward fibrotic or suspected patterns.**\n\n"
+                "Some biomarkers show early deviations commonly seen in progressive liver conditions.\n\n"
+                "✅ **Recommendation:** Regular monitoring, repeat liver function tests, "
+                "and lifestyle risk assessment are recommended."
+            )
+
+        else:
+            st.write(
+                "🟢 **Low similarity to fibrotic or suspected disease patterns.**\n\n"
+                "The current biomarker profile does not strongly align with chronic liver disease patterns.\n\n"
+                "✅ **Recommendation:** Continue routine health check-ups and preventive care."
+            )
+
+    # ========================================================
+    # Transparency note
+    # ========================================================
+    st.caption(
+        "⚠️ This system performs statistical pattern analysis based on historical data. "
+        "It does not provide a medical diagnosis and should be used only as decision-support."
+    )
 
 # ============================================================
-# FOOTER
+# Footer
 # ============================================================
 st.divider()
-st.caption("Pipeline-based ML • Plotly visuals • Clinical decision-support tool")
+st.caption("LightGBM Pipeline • Probability-driven interpretation • Ethical ML design")