Spaces:
Sleeping
Sleeping
File size: 14,372 Bytes
8960670 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 | #!/usr/bin/env python3
"""
LUNA16 Dataset Exploration Script
Phase 1, Week 1: Data Understanding
This script analyzes the LUNA16 dataset structure and generates statistics.
"""
import os
import pandas as pd
import numpy as np
import SimpleITK as sitk
import matplotlib.pyplot as plt
import seaborn as sns
from pathlib import Path
import json
from collections import defaultdict
# Configure paths
DATA_DIR = Path("data")
REPORTS_DIR = Path("reports")
FIGURES_DIR = REPORTS_DIR / "figures"
# Create output directories
FIGURES_DIR.mkdir(parents=True, exist_ok=True)
def load_annotations():
"""Load and analyze annotations.csv (ground truth nodules)."""
annotations_path = DATA_DIR / "annotations.csv"
df = pd.read_csv(annotations_path)
print("\n" + "="*60)
print("ANNOTATIONS.CSV ANALYSIS (Ground Truth Nodules)")
print("="*60)
print(f"\nShape: {df.shape}")
print(f"\nColumns: {df.columns.tolist()}")
print(f"\nFirst 5 rows:\n{df.head()}")
print(f"\nBasic Statistics:\n{df.describe()}")
print(f"\nUnique series (scans): {df['seriesuid'].nunique()}")
print(f"Total nodules: {len(df)}")
return df
def load_candidates():
"""Load and analyze candidates.csv (all candidate nodules)."""
candidates_path = DATA_DIR / "candidates.csv"
df = pd.read_csv(candidates_path)
print("\n" + "="*60)
print("CANDIDATES.CSV ANALYSIS (All Candidates)")
print("="*60)
print(f"\nShape: {df.shape}")
print(f"\nColumns: {df.columns.tolist()}")
print(f"\nFirst 5 rows:\n{df.head()}")
print(f"\nClass distribution:")
print(df['class'].value_counts())
print(f"\nPositive/Negative ratio: 1:{len(df[df['class']==0]) / max(len(df[df['class']==1]), 1):.0f}")
print(f"\nUnique series (scans): {df['seriesuid'].nunique()}")
return df
def analyze_class_imbalance(candidates_df):
"""Analyze and visualize class imbalance."""
print("\n" + "="*60)
print("CLASS IMBALANCE ANALYSIS")
print("="*60)
pos = len(candidates_df[candidates_df['class'] == 1])
neg = len(candidates_df[candidates_df['class'] == 0])
print(f"\nPositive candidates (true nodules): {pos}")
print(f"Negative candidates (false positives): {neg}")
print(f"Total candidates: {pos + neg}")
print(f"\nClass imbalance ratio: 1:{neg/pos:.0f}")
print(f"Positive percentage: {100*pos/(pos+neg):.3f}%")
# Visualize
fig, ax = plt.subplots(figsize=(8, 5))
bars = ax.bar(['Positive\n(True Nodules)', 'Negative\n(Non-nodules)'],
[pos, neg], color=['#2ecc71', '#e74c3c'])
ax.set_ylabel('Count')
ax.set_title('LUNA16 Class Distribution\n(Severe Imbalance: ~1:1350)')
ax.set_yscale('log')
# Add value labels
for bar, val in zip(bars, [pos, neg]):
ax.text(bar.get_x() + bar.get_width()/2, bar.get_height(),
f'{val:,}', ha='center', va='bottom', fontweight='bold')
plt.tight_layout()
plt.savefig(FIGURES_DIR / 'class_imbalance.png', dpi=150)
plt.close()
print(f"\nSaved: {FIGURES_DIR / 'class_imbalance.png'}")
return {'positive': pos, 'negative': neg, 'ratio': neg/pos}
def analyze_nodule_sizes(annotations_df):
"""Analyze nodule size distribution."""
print("\n" + "="*60)
print("NODULE SIZE DISTRIBUTION")
print("="*60)
diameters = annotations_df['diameter_mm']
print(f"\nDiameter statistics (mm):")
print(f" Min: {diameters.min():.2f}")
print(f" Max: {diameters.max():.2f}")
print(f" Mean: {diameters.mean():.2f}")
print(f" Median: {diameters.median():.2f}")
print(f" Std: {diameters.std():.2f}")
# Size categories as per roadmap
tiny = len(diameters[diameters < 4])
small = len(diameters[(diameters >= 4) & (diameters < 6)])
medium = len(diameters[(diameters >= 6) & (diameters < 10)])
large = len(diameters[diameters >= 10])
print(f"\nSize categories:")
print(f" Tiny (<4mm): {tiny} ({100*tiny/len(diameters):.1f}%)")
print(f" Small (4-6mm): {small} ({100*small/len(diameters):.1f}%)")
print(f" Medium (6-10mm): {medium} ({100*medium/len(diameters):.1f}%)")
print(f" Large (>10mm): {large} ({100*large/len(diameters):.1f}%)")
# Visualizations
fig, axes = plt.subplots(1, 2, figsize=(14, 5))
# Histogram
axes[0].hist(diameters, bins=30, color='steelblue', edgecolor='white', alpha=0.8)
axes[0].axvline(4, color='red', linestyle='--', label='Easy/Hard boundary (4mm)')
axes[0].axvline(8, color='orange', linestyle='--', label='Medium/Large boundary (8mm)')
axes[0].set_xlabel('Diameter (mm)')
axes[0].set_ylabel('Count')
axes[0].set_title('Nodule Diameter Distribution')
axes[0].legend()
# Category pie chart
sizes = [tiny, small, medium, large]
labels = [f'Tiny\n(<4mm)\n{tiny}', f'Small\n(4-6mm)\n{small}',
f'Medium\n(6-10mm)\n{medium}', f'Large\n(>10mm)\n{large}']
colors = ['#e74c3c', '#f39c12', '#3498db', '#2ecc71']
axes[1].pie(sizes, labels=labels, colors=colors, autopct='%1.1f%%', startangle=90)
axes[1].set_title('Nodule Size Categories\n(For Curriculum Learning)')
plt.tight_layout()
plt.savefig(FIGURES_DIR / 'nodule_sizes.png', dpi=150)
plt.close()
print(f"\nSaved: {FIGURES_DIR / 'nodule_sizes.png'}")
return {
'tiny': tiny, 'small': small, 'medium': medium, 'large': large,
'min_mm': float(diameters.min()), 'max_mm': float(diameters.max()),
'mean_mm': float(diameters.mean()), 'median_mm': float(diameters.median())
}
def analyze_subsets():
"""Analyze the subset structure and scan counts."""
print("\n" + "="*60)
print("SUBSET ANALYSIS")
print("="*60)
subset_stats = {}
total_scans = 0
for i in range(5): # subset0 to subset4
subset_dir = DATA_DIR / f"subset{i}"
# Handle nested structure (subset0/subset0/)
nested_dir = subset_dir / f"subset{i}"
search_dir = nested_dir if nested_dir.exists() else subset_dir
if search_dir.exists():
# Count .mhd files (each represents a CT scan)
mhd_files = list(search_dir.glob("*.mhd"))
subset_stats[f"subset{i}"] = len(mhd_files)
total_scans += len(mhd_files)
print(f" subset{i}: {len(mhd_files)} scans")
print(f"\nTotal scans across all subsets: {total_scans}")
if total_scans == 0:
print(" WARNING: No scans found!")
return subset_stats
# Train/Val/Test split as per roadmap
train_scans = subset_stats.get('subset0', 0) + subset_stats.get('subset1', 0) + subset_stats.get('subset2', 0)
val_scans = subset_stats.get('subset3', 0)
test_scans = subset_stats.get('subset4', 0)
print(f"\nRecommended Split (per roadmap):")
print(f" Train (subset0-2): {train_scans} scans ({100*train_scans/total_scans:.0f}%)")
print(f" Validation (subset3): {val_scans} scans ({100*val_scans/total_scans:.0f}%)")
print(f" Test (subset4): {test_scans} scans ({100*test_scans/total_scans:.0f}%)")
return subset_stats
def analyze_sample_scan():
"""Load and analyze a sample CT scan to understand the data format."""
print("\n" + "="*60)
print("SAMPLE SCAN ANALYSIS")
print("="*60)
# Find a sample .mhd file (handle nested directories)
sample_mhd = None
for i in range(5):
subset_dir = DATA_DIR / f"subset{i}"
nested_dir = subset_dir / f"subset{i}"
search_dir = nested_dir if nested_dir.exists() else subset_dir
if search_dir.exists():
mhd_files = list(search_dir.glob("*.mhd"))
if mhd_files:
sample_mhd = mhd_files[0]
break
if sample_mhd is None:
print("No .mhd files found!")
return None
print(f"\nLoading sample scan: {sample_mhd.name}")
# Load with SimpleITK
img = sitk.ReadImage(str(sample_mhd))
arr = sitk.GetArrayFromImage(img)
print(f"\nImage Properties:")
print(f" Size (SimpleITK): {img.GetSize()}") # (x, y, z)
print(f" Array shape (numpy): {arr.shape}") # (z, y, x)
print(f" Spacing (mm): {img.GetSpacing()}")
print(f" Origin: {img.GetOrigin()}")
print(f" Direction: {img.GetDirection()}")
print(f"\nHU Value Statistics:")
print(f" Min HU: {arr.min()}")
print(f" Max HU: {arr.max()}")
print(f" Mean HU: {arr.mean():.2f}")
print(f"\nMemory Usage:")
print(f" Data type: {arr.dtype}")
print(f" Size in MB: {arr.nbytes / (1024**2):.2f}")
# Visualize middle slices
fig, axes = plt.subplots(1, 3, figsize=(15, 5))
# Axial (z-axis)
z_mid = arr.shape[0] // 2
axes[0].imshow(arr[z_mid], cmap='gray', vmin=-1000, vmax=400)
axes[0].set_title(f'Axial View (Slice {z_mid}/{arr.shape[0]})')
axes[0].axis('off')
# Coronal (y-axis)
y_mid = arr.shape[1] // 2
axes[1].imshow(arr[:, y_mid, :], cmap='gray', vmin=-1000, vmax=400)
axes[1].set_title(f'Coronal View (Slice {y_mid}/{arr.shape[1]})')
axes[1].axis('off')
# Sagittal (x-axis)
x_mid = arr.shape[2] // 2
axes[2].imshow(arr[:, :, x_mid], cmap='gray', vmin=-1000, vmax=400)
axes[2].set_title(f'Sagittal View (Slice {x_mid}/{arr.shape[2]})')
axes[2].axis('off')
plt.suptitle(f'Sample CT Scan: {sample_mhd.stem}\n(HU windowed: -1000 to 400)')
plt.tight_layout()
plt.savefig(FIGURES_DIR / 'sample_scan_views.png', dpi=150)
plt.close()
print(f"\nSaved: {FIGURES_DIR / 'sample_scan_views.png'}")
return {
'shape': list(arr.shape),
'spacing': list(img.GetSpacing()),
'origin': list(img.GetOrigin()),
'hu_min': int(arr.min()),
'hu_max': int(arr.max()),
'size_mb': float(arr.nbytes / (1024**2))
}
def analyze_lung_segmentations():
"""Analyze the pre-computed lung segmentation masks."""
print("\n" + "="*60)
print("LUNG SEGMENTATION MASKS")
print("="*60)
seg_dir = DATA_DIR / "seg-lungs-LUNA16"
if not seg_dir.exists():
print(f"Segmentation directory not found: {seg_dir}")
return None
mhd_files = list(seg_dir.glob("*.mhd"))
print(f"\nTotal segmentation masks: {len(mhd_files)}")
if mhd_files:
# Load a sample segmentation
sample_seg = mhd_files[0]
print(f"\nSample segmentation: {sample_seg.name}")
img = sitk.ReadImage(str(sample_seg))
arr = sitk.GetArrayFromImage(img)
print(f" Shape: {arr.shape}")
print(f" Unique values: {np.unique(arr)}")
print(f" Lung voxels: {(arr > 0).sum()}")
return {'count': len(mhd_files)}
def generate_report(stats):
"""Generate a markdown exploration report."""
report_path = REPORTS_DIR / "data_exploration_report.md"
with open(report_path, 'w') as f:
f.write("# LUNA16 Data Exploration Report\n\n")
f.write("## Dataset Overview\n\n")
f.write(f"- **Total CT Scans**: {sum(stats['subsets'].values())}\n")
f.write(f"- **Total True Nodules**: {stats['class_balance']['positive']}\n")
f.write(f"- **Total Candidates**: {stats['class_balance']['positive'] + stats['class_balance']['negative']}\n")
f.write(f"- **Class Imbalance**: 1:{stats['class_balance']['ratio']:.0f}\n\n")
f.write("## Subset Distribution\n\n")
f.write("| Subset | Scans | Purpose |\n")
f.write("|--------|-------|--------|\n")
for i in range(5):
purpose = "Train" if i < 3 else ("Validation" if i == 3 else "Test")
f.write(f"| subset{i} | {stats['subsets'].get(f'subset{i}', 0)} | {purpose} |\n")
f.write("\n## Nodule Size Distribution\n\n")
f.write(f"- **Tiny (<4mm)**: {stats['nodule_sizes']['tiny']}\n")
f.write(f"- **Small (4-6mm)**: {stats['nodule_sizes']['small']}\n")
f.write(f"- **Medium (6-10mm)**: {stats['nodule_sizes']['medium']}\n")
f.write(f"- **Large (>10mm)**: {stats['nodule_sizes']['large']}\n\n")
f.write("## Key Figures\n\n")
f.write("\n\n")
f.write("\n\n")
f.write("\n\n")
f.write("## Hardware Considerations\n\n")
if stats.get('sample_scan'):
f.write(f"- **Single scan size**: ~{stats['sample_scan']['size_mb']:.0f} MB\n")
f.write(f"- **Scan shape**: {stats['sample_scan']['shape']}\n")
f.write("- **GPU VRAM**: 3.5GB (RTX 3050)\n")
f.write("- **Strategy**: Extract patches, don't load full scans\n")
print(f"\nSaved report: {report_path}")
def main():
"""Main exploration pipeline."""
print("\n" + "="*60)
print("LUNA16 DATASET EXPLORATION")
print("="*60)
stats = {}
# 1. Load and analyze annotations
annotations_df = load_annotations()
# 2. Load and analyze candidates
candidates_df = load_candidates()
# 3. Analyze class imbalance
stats['class_balance'] = analyze_class_imbalance(candidates_df)
# 4. Analyze nodule sizes
stats['nodule_sizes'] = analyze_nodule_sizes(annotations_df)
# 5. Analyze subsets
stats['subsets'] = analyze_subsets()
# 6. Analyze sample scan
stats['sample_scan'] = analyze_sample_scan()
# 7. Analyze lung segmentations
stats['segmentations'] = analyze_lung_segmentations()
# Save statistics as JSON
stats_path = REPORTS_DIR / "statistics.json"
with open(stats_path, 'w') as f:
json.dump(stats, f, indent=2)
print(f"\nSaved statistics: {stats_path}")
# Generate markdown report
generate_report(stats)
print("\n" + "="*60)
print("EXPLORATION COMPLETE!")
print("="*60)
print(f"\nOutput files:")
print(f" - {REPORTS_DIR / 'statistics.json'}")
print(f" - {REPORTS_DIR / 'data_exploration_report.md'}")
print(f" - {FIGURES_DIR / 'class_imbalance.png'}")
print(f" - {FIGURES_DIR / 'nodule_sizes.png'}")
print(f" - {FIGURES_DIR / 'sample_scan_views.png'}")
if __name__ == "__main__":
main()
|