File size: 14,809 Bytes
7a986e7 f2c28c8 7a986e7 f2c28c8 b75380d c30c8d7 4234218 d5fb548 f2c28c8 c4921c6 7a986e7 c4921c6 7a986e7 c4921c6 7a986e7 c4921c6 c30c8d7 7a986e7 c30c8d7 7a986e7 131b493 7a986e7 0eb72c7 7a986e7 c30c8d7 7a986e7 0eb72c7 c30c8d7 0eb72c7 b75380d 7a986e7 0eb72c7 c30c8d7 0eb72c7 7a986e7 131b493 56a1577 131b493 c4921c6 131b493 c4921c6 131b493 c4921c6 131b493 c4921c6 131b493 c4921c6 131b493 4234218 7a986e7 f2c28c8 27efa08 c30c8d7 27efa08 d5fb548 c4921c6 27efa08 8a1d3d2 27efa08 8a1d3d2 27efa08 8a1d3d2 f192f0b 27efa08 c30c8d7 d5fb548 c4921c6 8a1d3d2 d5fb548 c4921c6 d5fb548 27efa08 56a1577 27efa08 56a1577 27efa08 c4921c6 56a1577 c4921c6 56a1577 27efa08 d5fb548 27efa08 d5fb548 8a1d3d2 27efa08 d5fb548 c73c262 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 |
import cv2
import torch
import numpy as np
import torch.nn.functional as F
from torch import nn
from transformers import AutoImageProcessor, Swinv2ForImageClassification, SegformerForSemanticSegmentation
import streamlit as st
from PIL import Image
import io
import zipfile
import pandas as pd
from datetime import datetime
import os
import tempfile
import base64
# Add at the top with other constants
MODEL_OPTIONS = {
"Default (ferferefer/segformer)": "ferferefer/segformer",
"Pamixsun": "pamixsun/segformer_for_optic_disc_cup_segmentation"
}
# --- GlaucomaModel Class ---
class GlaucomaModel(object):
def __init__(self,
cls_model_path="pamixsun/swinv2_tiny_for_glaucoma_classification",
seg_model_path=None, # Make this optional
device=torch.device('cpu')):
self.device = device
# Classification model for glaucoma (always the same)
self.cls_extractor = AutoImageProcessor.from_pretrained(cls_model_path)
self.cls_model = Swinv2ForImageClassification.from_pretrained(cls_model_path).to(device).eval()
# Segmentation model - use provided path or default
seg_path = seg_model_path or MODEL_OPTIONS["Pamixsun"] # Default to Pamixsun if none provided
self.seg_extractor = AutoImageProcessor.from_pretrained(seg_path)
self.seg_model = SegformerForSemanticSegmentation.from_pretrained(seg_path).to(device).eval()
# Mapping for class labels
self.cls_id2label = self.cls_model.config.id2label
def glaucoma_pred(self, image):
inputs = self.cls_extractor(images=image.copy(), return_tensors="pt")
with torch.no_grad():
inputs.to(self.device)
outputs = self.cls_model(**inputs).logits
probs = F.softmax(outputs, dim=-1)
disease_idx = probs.cpu()[0, :].numpy().argmax()
confidence = probs.cpu()[0, disease_idx].item() * 100
return disease_idx, confidence
def optic_disc_cup_pred(self, image):
inputs = self.seg_extractor(images=image.copy(), return_tensors="pt")
with torch.no_grad():
inputs.to(self.device)
outputs = self.seg_model(**inputs)
logits = outputs.logits.cpu()
upsampled_logits = nn.functional.interpolate(
logits, size=image.shape[:2], mode="bilinear", align_corners=False
)
seg_probs = F.softmax(upsampled_logits, dim=1)
pred_disc_cup = upsampled_logits.argmax(dim=1)[0]
# Calculate segmentation confidence based on probability distribution
# For each pixel classified as cup/disc, check how confident the model is
cup_mask = pred_disc_cup == 2
disc_mask = pred_disc_cup == 1
# Get confidence only for pixels predicted as cup/disc
cup_confidence = seg_probs[0, 2, cup_mask].mean().item() * 100 if cup_mask.any() else 0
disc_confidence = seg_probs[0, 1, disc_mask].mean().item() * 100 if disc_mask.any() else 0
return pred_disc_cup.numpy().astype(np.uint8), cup_confidence, disc_confidence
def process(self, image):
disease_idx, cls_confidence = self.glaucoma_pred(image)
disc_cup, cup_confidence, disc_confidence = self.optic_disc_cup_pred(image)
try:
vcdr = simple_vcdr(disc_cup)
except:
vcdr = np.nan
mask = (disc_cup > 0).astype(np.uint8)
x, y, w, h = cv2.boundingRect(mask)
padding = max(50, int(0.2 * max(w, h)))
x = max(x - padding, 0)
y = max(y - padding, 0)
w = min(w + 2 * padding, image.shape[1] - x)
h = min(h + 2 * padding, image.shape[0] - y)
cropped_image = image[y:y+h, x:x+w] if w >= 50 and h >= 50 else image.copy()
_, disc_cup_image = add_mask(image, disc_cup, [1, 2], [[0, 255, 0], [255, 0, 0]], 0.2)
return disease_idx, disc_cup_image, vcdr, cls_confidence, cup_confidence, disc_confidence, cropped_image
# --- Utility Functions ---
def simple_vcdr(mask):
disc_area = np.sum(mask == 1)
cup_area = np.sum(mask == 2)
if disc_area == 0:
return np.nan
vcdr = cup_area / disc_area
return vcdr
def add_mask(image, mask, classes, colors, alpha=0.5):
overlay = image.copy()
for class_id, color in zip(classes, colors):
overlay[mask == class_id] = color
output = cv2.addWeighted(overlay, alpha, image, 1 - alpha, 0)
return output, overlay
def get_confidence_level(confidence):
"""Enhanced confidence descriptions for segmentation"""
if confidence >= 90:
return "Excellent (Model is very certain about the detected boundaries)"
elif confidence >= 75:
return "Good (Model is confident about most of the detected area)"
elif confidence >= 60:
return "Fair (Model has some uncertainty in parts of the detection)"
elif confidence >= 45:
return "Poor (Model is uncertain about many detected areas)"
else:
return "Very Poor (Model's detection is highly uncertain)"
def process_batch(model, images_data, progress_bar=None):
results = []
for idx, (file_name, image) in enumerate(images_data):
try:
disease_idx, disc_cup_image, vcdr, cls_conf, cup_conf, disc_conf, cropped_image = model.process(image)
results.append({
'file_name': file_name,
'diagnosis': model.cls_id2label[disease_idx],
'confidence': cls_conf,
'vcdr': vcdr,
'cup_conf': cup_conf,
'disc_conf': disc_conf,
'processed_image': disc_cup_image,
'cropped_image': cropped_image
})
if progress_bar:
progress_bar.progress((idx + 1) / len(images_data))
except Exception as e:
st.error(f"Error processing {file_name}: {str(e)}")
return results
def save_results(results, original_images):
# Create temporary directory for results
with tempfile.TemporaryDirectory() as temp_dir:
# Save report as CSV
df = pd.DataFrame([{
'File': r['file_name'],
'Diagnosis': r['diagnosis'],
'Confidence (%)': f"{r['confidence']:.1f}",
'VCDR': f"{r['vcdr']:.3f}",
'Cup Confidence (%)': f"{r['cup_conf']:.1f}",
'Disc Confidence (%)': f"{r['disc_conf']:.1f}"
} for r in results])
report_path = os.path.join(temp_dir, 'report.csv')
df.to_csv(report_path, index=False)
# Save processed images
for result, orig_img in zip(results, original_images):
img_name = result['file_name']
base_name = os.path.splitext(img_name)[0]
# Save original
orig_path = os.path.join(temp_dir, f"{base_name}_original.jpg")
Image.fromarray(orig_img).save(orig_path)
# Save segmentation
seg_path = os.path.join(temp_dir, f"{base_name}_segmentation.jpg")
Image.fromarray(result['processed_image']).save(seg_path)
# Save ROI
roi_path = os.path.join(temp_dir, f"{base_name}_roi.jpg")
Image.fromarray(result['cropped_image']).save(roi_path)
# Create ZIP file
zip_path = os.path.join(temp_dir, 'results.zip')
with zipfile.ZipFile(zip_path, 'w') as zipf:
for root, _, files in os.walk(temp_dir):
for file in files:
if file != 'results.zip':
file_path = os.path.join(root, file)
arcname = os.path.basename(file_path)
zipf.write(file_path, arcname)
with open(zip_path, 'rb') as f:
return f.read()
# --- Streamlit Interface ---
def main():
# Use the old layout setting method
st.set_page_config(layout="wide")
# Use simple title instead of markdown
st.title("Glaucoma Screening from Retinal Fundus Images")
st.write("Upload retinal images for automated glaucoma detection and optic disc/cup segmentation")
# Add model selection in sidebar before file upload
st.sidebar.title("Model Settings")
selected_model = st.sidebar.selectbox(
"Select Segmentation Model",
list(MODEL_OPTIONS.keys()),
index=1 # Default to Pamixsun
)
st.sidebar.title("Upload Images")
st.set_option('deprecation.showfileUploaderEncoding', False) # Important for old versions
uploaded_files = st.sidebar.file_uploader(
"Upload retinal images",
type=['png', 'jpeg', 'jpg'],
accept_multiple_files=True
)
# Simple explanation in sidebar
st.sidebar.markdown("""
### Understanding Results:
- Diagnosis Confidence: AI certainty level
- VCDR: Cup to disc ratio (>0.7 high risk)
- Segmentation: Accuracy of detection
""")
if uploaded_files:
try:
# Enhanced model loading feedback
st.write("🤖 Initializing AI models...")
st.write(f"• Loading classification model: pamixsun/swinv2_tiny_for_glaucoma_classification")
st.write(f"• Loading segmentation model: {selected_model}")
# Initialize model with selected segmentation model
model = GlaucomaModel(
device=torch.device("cuda:0" if torch.cuda.is_available() else "cpu"),
seg_model_path=MODEL_OPTIONS[selected_model]
)
# Show model loading completion
st.write("✅ Models loaded successfully")
st.write(f"🖥️ Using: {'GPU' if torch.cuda.is_available() else 'CPU'} for processing")
st.write("---")
for file in uploaded_files:
try:
# Process each image with enhanced feedback
st.write(f"📸 Processing image: {file.name}")
image = Image.open(file).convert('RGB')
image_np = np.array(image)
# Get predictions
disease_idx, disc_cup_image, vcdr, cls_conf, cup_conf, disc_conf, cropped_image = model.process(image_np)
# Enhanced results display
st.write("---")
st.write(f"Results for {file.name}")
# Diagnosis section
st.write("📊 **Diagnosis Results:**")
st.write(f"• Finding: {model.cls_id2label[disease_idx]}")
st.write(f"• AI Confidence: {cls_conf:.1f}% ({get_confidence_level(cls_conf)})")
# Enhanced Segmentation confidence section with detailed explanations
st.write("\n🔍 **Understanding Segmentation Confidence:**")
st.write("""
Segmentation confidence shows how certain the AI is about each pixel it classified:
• For the Optic Cup (central depression):
- Measures the AI's certainty that the red-colored pixels are truly part of the cup
- Higher confidence means clearer cup boundaries and more reliable VCDR
• For the Optic Disc (entire circular area):
- Indicates how sure the AI is about the green-outlined disc boundary
- Higher confidence suggests better disc margin visibility
Confidence scores are calculated by averaging the model's certainty
for each pixel it identified as cup or disc. A score of 100% would mean
the model is absolutely certain about every pixel's classification.
""")
st.write("\n📊 **Current Segmentation Confidence Scores:**")
st.write(f"• Optic Cup Detection: {cup_conf:.1f}% - {get_confidence_level(cup_conf)}")
st.write(f"• Optic Disc Detection: {disc_conf:.1f}% - {get_confidence_level(disc_conf)}")
# Add interpretation guidance
if cup_conf >= 75 and disc_conf >= 75:
st.write("✅ High confidence scores indicate reliable measurements")
elif cup_conf < 60 or disc_conf < 60:
st.write("""
⚠️ Lower confidence scores might be due to:
• Image quality issues (blur, poor contrast)
• Unusual anatomical variations
• Pathological changes affecting visibility
• Poor image centering or focus
Consider retaking the image if possible.
""")
# Clinical metrics
st.write("\n📏 **Clinical Measurements:**")
st.write(f"• Cup-to-Disc Ratio (VCDR): {vcdr:.3f}")
if vcdr > 0.7:
st.write(" ⚠️ High VCDR - Potential risk indicator")
elif vcdr > 0.5:
st.write(" ℹ️ Borderline VCDR - Follow-up recommended")
else:
st.write(" ✅ Normal VCDR range")
# Image display with enhanced captions
st.write("\n🖼️ **Visual Analysis:**")
st.image(disc_cup_image, caption="""
Segmentation Overlay
• Green outline: Optic Disc boundary
• Red area: Optic Cup region
• Transparency shows underlying retina
""")
st.image(cropped_image, caption="Zoomed Region of Interest")
# Add quality note if needed
if cup_conf < 60 or disc_conf < 60:
st.write("\n⚠️ Note: Low segmentation confidence. Image quality might affect measurements.")
except Exception as e:
st.error(f"Error processing {file.name}: {str(e)}")
continue
# Simple summary at the end
st.write("---")
st.write("Processing complete!")
except Exception as e:
st.error(f"An error occurred: {str(e)}")
if __name__ == "__main__":
main()
|