stroke-classification / src /streamlit_app.py
bakhili's picture
Update src/streamlit_app.py
1ca6b73 verified
Raw
History Blame Contribute Delete
23.7 kB
import streamlit as st
import numpy as np
import os
import sys
from PIL import Image
from scipy import ndimage
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
# Set environment variables to fix permission issues
os.environ['MPLCONFIGDIR'] = '/tmp/matplotlib'
os.environ['STREAMLIT_SERVER_HEADLESS'] = 'true'
# Minimal imports to avoid conflicts
try:
import tensorflow as tf
TF_AVAILABLE = True
except ImportError:
TF_AVAILABLE = False
st.error("TensorFlow not available")
try:
import matplotlib
matplotlib.use('Agg') # Use non-interactive backend
import matplotlib.cm as cm
MPL_AVAILABLE = True
except ImportError:
MPL_AVAILABLE = False
# Page config
st.set_page_config(
page_title="Stroke Classifier",
page_icon="🧠",
layout="wide")
# Simple styling
st.markdown("""
<style>
.main-header {
font-size: 2.5rem;
color: #1f77b4;
text-align: center;
margin-bottom: 2rem;
}
.prediction-box {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
padding: 2rem;
border-radius: 1rem;
text-align: center;
margin: 1rem 0;
}
.status-box {
padding: 1rem;
border-radius: 0.5rem;
margin: 1rem 0;
}
.success { background-color: #d4edda; border: 1px solid #c3e6cb; color: #155724; }
.error { background-color: #f8d7da; border: 1px solid #f5c6cb; color: #721c24; }
.info { background-color: #d1ecf1; border: 1px solid #bee5eb; color: #0c5460; }
.warning { background-color: #fff3cd; border: 1px solid #ffeaa7; color: #856404; }
.debug { background-color: #f8f9fa; border: 1px solid #dee2e6; color: #495057; font-family: monospace; }
</style>""", unsafe_allow_html=True)
# Initialize session state
if 'model_loaded' not in st.session_state:
st.session_state.model_loaded = False
st.session_state.model = None
st.session_state.model_status = "Not loaded"
STROKE_LABELS = ["Hemorrhagic Stroke", "Ischemic Stroke", "No Stroke"]
def find_model_file():
"""Find the model file in various possible locations."""
possible_paths = [
"stroke_classification_model.h5",
"./stroke_classification_model.h5",
"/app/stroke_classification_model.h5",
"src/stroke_classification_model.h5",
os.path.join(os.getcwd(), "stroke_classification_model.h5")
]
# Also check all .h5 files in current directory and subdirectories
for root, dirs, files in os.walk('.'):
for file in files:
if file.endswith('.h5'):
possible_paths.append(os.path.join(root, file))
for path in possible_paths:
if os.path.exists(path):
return path
return None
@st.cache_resource
def load_stroke_model():
"""Load model with caching."""
if not TF_AVAILABLE:
return None, "❌ TensorFlow not available"
try:
# Find the model file
model_path = find_model_file()
if model_path is None:
# List all files to help debug
current_files = []
for root, dirs, files in os.walk('.'):
for file in files:
current_files.append(os.path.join(root, file))
return None, f"❌ Model file not found. Available files: {current_files[:10]}"
st.info(f"Found model at: {model_path}")
# Load model with minimal custom objects
model = tf.keras.models.load_model(model_path, compile=False)
return model, f"βœ… Model loaded successfully from: {model_path}"
except Exception as e:
return None, f"❌ Model loading failed: {str(e)}"
def analyze_heatmap_distribution(heatmap, name="Heatmap"):
"""Analyze the distribution of heatmap values."""
if heatmap is None:
return None
flat_values = heatmap.flatten()
analysis = {
'name': name,
'shape': heatmap.shape,
'total_pixels': heatmap.size,
'min': float(np.min(flat_values)),
'max': float(np.max(flat_values)),
'mean': float(np.mean(flat_values)),
'median': float(np.median(flat_values)),
'std': float(np.std(flat_values)),
'range': float(np.max(flat_values) - np.min(flat_values)),
'unique_values': len(np.unique(flat_values)),
'zero_pixels': int(np.sum(flat_values == 0)),
'non_zero_pixels': int(np.sum(flat_values > 0)),
'percentiles': {
'1%': float(np.percentile(flat_values, 1)),
'5%': float(np.percentile(flat_values, 5)),
'25%': float(np.percentile(flat_values, 25)),
'75%': float(np.percentile(flat_values, 75)),
'95%': float(np.percentile(flat_values, 95)),
'99%': float(np.percentile(flat_values, 99))
}
}
# Determine if heatmap has good contrast
if analysis['range'] < 0.1:
analysis['contrast_quality'] = 'Very Poor (range < 0.1)'
elif analysis['range'] < 0.3:
analysis['contrast_quality'] = 'Poor (range < 0.3)'
elif analysis['range'] < 0.7:
analysis['contrast_quality'] = 'Moderate (range < 0.7)'
else:
analysis['contrast_quality'] = 'Good (range >= 0.7)'
return analysis
def force_contrast_enhancement(heatmap, method='aggressive'):
"""Force better contrast in heatmap using various methods."""
if heatmap is None:
return None, "No heatmap provided"
original_analysis = analyze_heatmap_distribution(heatmap, "Original")
if method == 'aggressive':
# Method 1: Aggressive percentile stretching
p1, p99 = np.percentile(heatmap, [1, 99])
if p99 > p1:
enhanced = np.clip((heatmap - p1) / (p99 - p1), 0, 1)
else:
enhanced = heatmap
# Apply power transformation to spread values
enhanced = np.power(enhanced, 0.3) # Gamma < 1 spreads values
elif method == 'histogram_eq':
# Method 2: Histogram equalization
flat = heatmap.flatten()
hist, bins = np.histogram(flat, bins=256, range=(0, 1))
cdf = hist.cumsum()
cdf = cdf / cdf[-1] # Normalize
# Interpolate to get new values
enhanced = np.interp(flat, bins[:-1], cdf).reshape(heatmap.shape)
elif method == 'adaptive':
# Method 3: Adaptive enhancement based on local statistics
# Local mean and std
local_mean = ndimage.uniform_filter(heatmap, size=20)
local_std = ndimage.generic_filter(heatmap, np.std, size=20)
# Enhance based on local statistics
enhanced = (heatmap - local_mean) / (local_std + 1e-8)
enhanced = np.clip(enhanced, -3, 3) # Clip outliers
enhanced = (enhanced + 3) / 6 # Normalize to [0, 1]
elif method == 'artificial_peaks':
# Method 4: Create artificial peaks for visualization
enhanced = heatmap.copy()
# Find top 10% of values and enhance them
threshold = np.percentile(enhanced, 90)
mask = enhanced >= threshold
enhanced[mask] = enhanced[mask] * 2
# Find bottom 10% and suppress them
threshold_low = np.percentile(enhanced, 10)
mask_low = enhanced <= threshold_low
enhanced[mask_low] = enhanced[mask_low] * 0.1
# Normalize
enhanced = np.clip(enhanced, 0, 1)
else:
enhanced = heatmap
enhanced_analysis = analyze_heatmap_distribution(enhanced, f"Enhanced ({method})")
return enhanced, f"Enhanced using {method}", original_analysis, enhanced_analysis
def create_diagnostic_heatmap_visualization(heatmap, title="Heatmap Analysis"):
"""Create a comprehensive diagnostic visualization of the heatmap."""
if not MPL_AVAILABLE or heatmap is None:
return None
fig, axes = plt.subplots(2, 3, figsize=(18, 12))
# Original heatmap
im1 = axes[0, 0].imshow(heatmap, cmap='hot', vmin=0, vmax=1)
axes[0, 0].set_title(f"{title} - Hot Colormap")
plt.colorbar(im1, ax=axes[0, 0])
# Different colormap
im2 = axes[0, 1].imshow(heatmap, cmap='viridis', vmin=0, vmax=1)
axes[0, 1].set_title(f"{title} - Viridis Colormap")
plt.colorbar(im2, ax=axes[0, 1])
# High contrast version
im3 = axes[0, 2].imshow(heatmap, cmap='RdYlBu_r', vmin=np.min(heatmap), vmax=np.max(heatmap))
axes[0, 2].set_title(f"{title} - Auto-scaled")
plt.colorbar(im3, ax=axes[0, 2])
# Histogram
axes[1, 0].hist(heatmap.flatten(), bins=50, alpha=0.7, color='blue')
axes[1, 0].set_title("Value Distribution")
axes[1, 0].set_xlabel("Heatmap Value")
axes[1, 0].set_ylabel("Frequency")
# 3D surface plot
x = np.arange(heatmap.shape[1])
y = np.arange(heatmap.shape[0])
X, Y = np.meshgrid(x, y)
ax_3d = fig.add_subplot(2, 3, 5, projection='3d')
surf = ax_3d.plot_surface(X[::8, ::8], Y[::8, ::8], heatmap[::8, ::8],
cmap='hot', alpha=0.8)
ax_3d.set_title("3D Surface View")
# Statistics text
analysis = analyze_heatmap_distribution(heatmap)
stats_text = f"""
Shape: {analysis['shape']}
Range: {analysis['range']:.4f}
Mean: {analysis['mean']:.4f}
Std: {analysis['std']:.4f}
Unique values: {analysis['unique_values']}
Contrast: {analysis['contrast_quality']}
Percentiles:
1%: {analysis['percentiles']['1%']:.4f}
25%: {analysis['percentiles']['25%']:.4f}
75%: {analysis['percentiles']['75%']:.4f}
99%: {analysis['percentiles']['99%']:.4f}
"""
axes[1, 2].text(0.1, 0.9, stats_text, transform=axes[1, 2].transAxes,
fontsize=10, verticalalignment='top', fontfamily='monospace')
axes[1, 2].set_title("Statistics")
axes[1, 2].axis('off')
plt.tight_layout()
return fig
def create_multiple_enhancement_comparison(heatmap):
"""Compare different enhancement methods side by side."""
if not MPL_AVAILABLE or heatmap is None:
return None
methods = ['aggressive', 'histogram_eq', 'adaptive', 'artificial_peaks']
enhanced_maps = {}
for method in methods:
enhanced, _, _, _ = force_contrast_enhancement(heatmap, method)
enhanced_maps[method] = enhanced
fig, axes = plt.subplots(2, 3, figsize=(18, 12))
# Original
im0 = axes[0, 0].imshow(heatmap, cmap='hot', vmin=0, vmax=1)
axes[0, 0].set_title("Original Heatmap")
plt.colorbar(im0, ax=axes[0, 0])
# Enhanced versions
positions = [(0, 1), (0, 2), (1, 0), (1, 1)]
for i, (method, enhanced) in enumerate(enhanced_maps.items()):
row, col = positions[i]
im = axes[row, col].imshow(enhanced, cmap='hot', vmin=0, vmax=1)
axes[row, col].set_title(f"Enhanced: {method}")
plt.colorbar(im, ax=axes[row, col])
# Comparison histogram
axes[1, 2].hist(heatmap.flatten(), bins=30, alpha=0.5, label='Original', color='blue')
for method, enhanced in enhanced_maps.items():
axes[1, 2].hist(enhanced.flatten(), bins=30, alpha=0.3, label=method)
axes[1, 2].set_title("Value Distributions")
axes[1, 2].legend()
axes[1, 2].set_xlabel("Value")
axes[1, 2].set_ylabel("Frequency")
plt.tight_layout()
return fig
def predict_stroke(img, model):
"""Predict stroke type from image."""
if model is None:
return None, "Model not loaded"
try:
# Preprocess image
img_resized = img.resize((224, 224))
img_array = np.array(img_resized, dtype=np.float32)
# Handle grayscale
if len(img_array.shape) == 2:
img_array = np.stack([img_array] * 3, axis=-1)
# Normalize and add batch dimension
img_array = np.expand_dims(img_array, axis=0) / 255.0
# Predict
predictions = model.predict(img_array, verbose=0)
return predictions[0], None
except Exception as e:
return None, f"Prediction error: {str(e)}"
def create_test_heatmaps():
"""Create test heatmaps with known patterns for comparison."""
test_maps = {}
# Test 1: High contrast pattern
test_maps['high_contrast'] = np.zeros((224, 224))
test_maps['high_contrast'][50:150, 50:150] = 1.0
test_maps['high_contrast'][75:125, 75:125] = 0.0
# Test 2: Gradient pattern
x = np.linspace(0, 1, 224)
y = np.linspace(0, 1, 224)
X, Y = np.meshgrid(x, y)
test_maps['gradient'] = X * Y
# Test 3: Gaussian blobs
test_maps['gaussian'] = np.zeros((224, 224))
centers = [(60, 60), (160, 160), (60, 160)]
for cx, cy in centers:
y, x = np.ogrid[:224, :224]
mask = np.exp(-((x - cx)**2 + (y - cy)**2) / (2 * 30**2))
test_maps['gaussian'] += mask
test_maps['gaussian'] = test_maps['gaussian'] / np.max(test_maps['gaussian'])
# Test 4: Low contrast (similar to your issue)
test_maps['low_contrast'] = np.random.normal(0.5, 0.05, (224, 224))
test_maps['low_contrast'] = np.clip(test_maps['low_contrast'], 0, 1)
return test_maps
# Main App
def main():
# Header
st.markdown('<h1 class="main-header">🧠 Heatmap Diagnostic System</h1>', unsafe_allow_html=True)
# Auto-load model on startup
if not st.session_state.model_loaded:
with st.spinner("Loading AI model..."):
st.session_state.model, st.session_state.model_status = load_stroke_model()
st.session_state.model_loaded = True
# System status
st.markdown("### πŸ”§ System Status")
col1, col2, col3 = st.columns(3)
with col1:
if TF_AVAILABLE:
st.markdown('<div class="status-box success">βœ… TensorFlow Ready</div>', unsafe_allow_html=True)
st.write(f"TF Version: {tf.__version__}")
else:
st.markdown('<div class="status-box error">❌ TensorFlow Error</div>', unsafe_allow_html=True)
with col2:
if MPL_AVAILABLE:
st.markdown('<div class="status-box success">βœ… Matplotlib Ready</div>', unsafe_allow_html=True)
else:
st.markdown('<div class="status-box error">❌ Matplotlib Error</div>', unsafe_allow_html=True)
with col3:
if "βœ…" in st.session_state.model_status:
st.markdown('<div class="status-box success">βœ… Model Loaded</div>', unsafe_allow_html=True)
else:
st.markdown('<div class="status-box error">❌ Model Error</div>', unsafe_allow_html=True)
# Test heatmaps section
st.markdown("### πŸ§ͺ Test Heatmap Patterns")
test_maps = create_test_heatmaps()
col1, col2 = st.columns(2)
with col1:
st.write("**Test Pattern:**")
test_pattern = st.selectbox(
"Choose a test pattern",
list(test_maps.keys()),
help="Test different heatmap patterns to see how they display"
)
if test_pattern:
test_heatmap = test_maps[test_pattern]
# Show diagnostic visualization
diagnostic_fig = create_diagnostic_heatmap_visualization(test_heatmap, f"Test: {test_pattern}")
if diagnostic_fig:
st.pyplot(diagnostic_fig)
plt.close()
with col2:
st.write("**Enhancement Comparison:**")
if test_pattern:
test_heatmap = test_maps[test_pattern]
# Show enhancement comparison
comparison_fig = create_multiple_enhancement_comparison(test_heatmap)
if comparison_fig:
st.pyplot(comparison_fig)
plt.close()
# Sidebar
with st.sidebar:
st.header("πŸ“€ Upload Brain Scan")
uploaded_file = st.file_uploader(
"Choose a brain scan image...",
type=['png', 'jpg', 'jpeg', 'bmp', 'tiff'],
help="Upload a brain scan image for stroke classification"
)
st.markdown("---")
st.header("🎨 Enhancement Options")
enhancement_method = st.selectbox(
"Enhancement Method",
['none', 'aggressive', 'histogram_eq', 'adaptive', 'artificial_peaks'],
index=1,
help="Choose how to enhance heatmap contrast"
)
show_diagnostics = st.checkbox("Show Diagnostic Analysis", value=True)
show_comparisons = st.checkbox("Show Enhancement Comparisons", value=True)
if uploaded_file is not None:
# Load image
image = Image.open(uploaded_file)
st.subheader("πŸ“‹ Classification Results")
if st.session_state.model is not None:
# Predict
with st.spinner("πŸ” Analyzing brain scan..."):
predictions, error = predict_stroke(image, st.session_state.model)
if error:
st.error(error)
else:
# Get top prediction
class_idx = np.argmax(predictions)
confidence = predictions[class_idx] * 100
predicted_class = STROKE_LABELS[class_idx]
# Display main result
st.markdown(f"""
<div class="prediction-box">
<h2>{predicted_class}</h2>
<h3>Confidence: {confidence:.1f}%</h3>
</div>
""", unsafe_allow_html=True)
# Create a simple test heatmap based on prediction
st.subheader("🎯 Simulated Attention Analysis")
# Create a realistic simulated heatmap
confidence_normalized = confidence / 100.0
predicted_class_idx = np.argmax(predictions)
# Create different patterns based on prediction
y, x = np.ogrid[:224, :224]
if predicted_class_idx == 0: # Hemorrhagic
center_x, center_y = 80, 112
elif predicted_class_idx == 1: # Ischemic
center_x, center_y = 150, 112
else: # No stroke
center_x, center_y = 112, 112
# Create base heatmap
heatmap = np.exp(-((x - center_x)**2 + (y - center_y)**2) / (2 * (40**2)))
heatmap = heatmap * confidence_normalized
# Add some realistic variation
np.random.seed(42)
noise = np.random.normal(0, 0.02, heatmap.shape)
heatmap = np.maximum(heatmap + noise, 0)
# Normalize
if np.max(heatmap) > 0:
heatmap = heatmap / np.max(heatmap)
# Show diagnostic analysis
if show_diagnostics:
st.write("**πŸ“Š Heatmap Diagnostic Analysis:**")
diagnostic_fig = create_diagnostic_heatmap_visualization(heatmap, "Your Model's Attention")
if diagnostic_fig:
st.pyplot(diagnostic_fig)
plt.close()
# Show enhancement comparisons
if show_comparisons:
st.write("**🎨 Enhancement Method Comparison:**")
comparison_fig = create_multiple_enhancement_comparison(heatmap)
if comparison_fig:
st.pyplot(comparison_fig)
plt.close()
# Apply selected enhancement
if enhancement_method != 'none':
enhanced_heatmap, enhancement_msg, orig_analysis, enh_analysis = force_contrast_enhancement(heatmap, enhancement_method)
st.write(f"**πŸ”§ Applied Enhancement: {enhancement_method}**")
# Show before/after comparison
fig, axes = plt.subplots(1, 3, figsize=(15, 5))
# Original
im1 = axes[0].imshow(heatmap, cmap='hot', vmin=0, vmax=1)
axes[0].set_title("Original Heatmap")
axes[0].axis('off')
plt.colorbar(im1, ax=axes[0])
# Enhanced
im2 = axes[1].imshow(enhanced_heatmap, cmap='hot', vmin=0, vmax=1)
axes[1].set_title(f"Enhanced ({enhancement_method})")
axes[1].axis('off')
plt.colorbar(im2, ax=axes[1])
# Overlay on image
img_resized = image.resize((224, 224))
img_array = np.array(img_resized)
axes[2].imshow(img_array)
im3 = axes[2].imshow(enhanced_heatmap, cmap='hot', alpha=0.6, vmin=0, vmax=1)
axes[2].set_title("Enhanced Overlay")
axes[2].axis('off')
plt.colorbar(im3, ax=axes[2])
plt.tight_layout()
st.pyplot(fig)
plt.close()
# Show improvement statistics
col1, col2 = st.columns(2)
with col1:
st.write("**Original Stats:**")
st.write(f"Range: {orig_analysis['range']:.4f}")
st.write(f"Std: {orig_analysis['std']:.4f}")
st.write(f"Contrast: {orig_analysis['contrast_quality']}")
with col2:
st.write("**Enhanced Stats:**")
st.write(f"Range: {enh_analysis['range']:.4f}")
st.write(f"Std: {enh_analysis['std']:.4f}")
st.write(f"Contrast: {enh_analysis['contrast_quality']}")
else:
st.error("❌ Model not loaded.")
else:
# Welcome message
st.markdown("""
## πŸ‘‹ Welcome to the Heatmap Diagnostic System
This system helps you understand **why your heatmaps appear as one color** and how to fix it.
### πŸ” What This Shows You:
- **Value distribution analysis** - See if your heatmap has variation
- **Multiple visualization methods** - Different ways to display the same data
- **Enhancement techniques** - Force better contrast and visibility
- **Test patterns** - Compare with known good patterns
### 🎯 Common Issues:
- **Low variance** - All values are nearly the same
- **Poor normalization** - Values compressed into narrow range
- **Uniform attention** - Model doesn't focus on specific areas
### πŸ› οΈ Solutions:
- **Aggressive enhancement** - Force contrast stretching
- **Histogram equalization** - Spread values evenly
- **Artificial peaks** - Enhance high-attention areas
**Try the test patterns above, then upload your image! πŸ‘†**
""")
# Medical disclaimer
st.markdown("---")
st.warning("⚠️ **Medical Disclaimer:** This AI system is for educational and research purposes only. It should not be used for actual medical diagnosis. Always consult qualified healthcare professionals for medical decisions.")
if __name__ == "__main__":
main()