MedVision / src /medvision_ds /utils /preprocess_utils.py
YongchengYAO's picture
minor update: move_folder()
853dbfd
import os
import glob
import shutil
import sys
import filecmp
import hashlib
import nibabel as nib
import numpy as np
import json
import math
from pathlib import Path
# NOTE:
# In a kubernetes environment, the number of CPUs available to the container may be limited by cgroups.
# This function retrieves the number of CPUs available to the container.
# DO NOT use os.cpu_count() directly, as it may return the total number of CPUs on the host machine,
def _get_cgroup_limited_cpus():
# cgroup v1
try:
base = Path("/sys/fs/cgroup/cpu")
q = base / "cpu.cfs_quota_us"
p = base / "cpu.cfs_period_us"
if q.exists() and p.exists():
quota = int(q.read_text().strip())
period = int(p.read_text().strip())
if quota > 0 and period > 0:
return math.floor(quota / period)
except (ValueError, OSError):
pass
# cgroup v2
try:
line = Path("/sys/fs/cgroup/cpu.max").read_text().strip()
quota, period = line.split()
if quota != "max":
return math.floor(int(quota) / int(period))
except (ValueError, OSError):
pass
# fallback to host-wide CPU count
return os.cpu_count()
def move_folder(source_folder, destination_folder, create_dest=True):
"""
Moves a folder from source to destination.
Args:
source_folder (str): Path to the source folder to move
destination_folder (str): Path to the destination location
create_dest (bool): Whether to create the destination parent directory if it doesn't exist
Returns:
bool: True if successful, False otherwise
Raises:
FileNotFoundError: If the source folder doesn't exist
"""
# Check if source folder exists
if not os.path.exists(source_folder):
raise FileNotFoundError(f"Source folder does not exist: {source_folder}")
# Create destination directory if it doesn't exist and create_dest is True
if create_dest and not os.path.exists(os.path.dirname(destination_folder)):
os.makedirs(os.path.dirname(destination_folder), exist_ok=True)
try:
# Check if destination folder exists
if os.path.exists(destination_folder):
# If destination exists, move contents
for item in os.listdir(source_folder):
s = os.path.join(source_folder, item)
d = os.path.join(destination_folder, item)
if os.path.isdir(s) and os.path.isdir(d):
# Merge contents instead of nesting directory
for sub_item in os.listdir(s):
ss = os.path.join(s, sub_item)
dd = os.path.join(d, sub_item)
shutil.move(ss, dd)
os.rmdir(s)
else:
shutil.move(s, d)
else:
# If destination doesn't exist, move the entire folder
shutil.move(source_folder, destination_folder)
print(f"Successfully moved '{source_folder}' to '{destination_folder}'")
return True
except Exception as e:
print(f"Failed to move folder: {e}")
return False
def check_nii_header_for_img_mask(image_path, mask_path):
# Inspect the mask files
mask_nii = nib.load(mask_path)
mask_data = mask_nii.get_fdata()
mask_file_info = {
"voxel_size": tuple(round(x, 3) for x in mask_nii.header.get_zooms()),
"affine": np.round(mask_nii.affine, 3),
"orientation": nib.orientations.aff2axcodes(mask_nii.affine),
"array_size": mask_data.shape,
}
# Inspect the image files
img_nii = nib.load(image_path)
img_data = img_nii.get_fdata()
image_file_info = {
"voxel_size": tuple(round(x, 3) for x in img_nii.header.get_zooms()),
"affine": np.round(img_nii.affine, 3),
"orientation": nib.orientations.aff2axcodes(img_nii.affine),
"array_size": img_data.shape,
}
# Check if mask and image properties match
print(
f"Checking properties for the image and mask images:\nImage: {image_path}\nMask: {mask_path}"
)
for key in mask_file_info:
if isinstance(mask_file_info[key], np.ndarray):
if not np.allclose(
mask_file_info[key], image_file_info[key], atol=1e-5, rtol=1e-3
):
raise ValueError(
f"\n\nMismatch in {key} between image and mask:\n"
f"Image {key}:\n{image_file_info[key]}\n"
f"Mask {key}:\n{mask_file_info[key]}\n"
)
elif mask_file_info[key] != image_file_info[key]:
raise ValueError(
f"\n\nMismatch in {key} between image and mask:\n"
f"Image {key}:\n{image_file_info[key]}\n"
f"Mask {key}:\n{mask_file_info[key]}\n"
)
print(f"Properties (NIfTI file header) match!\n")
def check_nii_header_for_img_mask_batch(image_dir, mask_dir):
"""
Check NIfTI headers for all matching image and mask pairs in given directories
Args:
image_dir (str): Directory containing image files
mask_dir (str): Directory containing mask files
"""
# Get all nii.gz files in image directory
image_files = glob.glob(os.path.join(image_dir, "*.nii.gz"))
total_files = len(image_files)
print(f"Found {total_files} image files. Starting header check...\n")
for idx, image_path in enumerate(image_files, 1):
# Get corresponding mask file name
image_name = os.path.basename(image_path)
mask_path = os.path.join(mask_dir, image_name)
print(f"Processing file {idx}/{total_files}")
# Check if mask exists
if not os.path.exists(mask_path):
print(f"WARNING: No matching mask found for {image_name}\n")
continue
try:
check_nii_header_for_img_mask(image_path, mask_path)
except ValueError as e:
print(f"ERROR: {str(e)}")
continue
except Exception as e:
print(f"ERROR: Unexpected error processing {image_name}: {str(e)}\n")
continue
print("\nHeader check completed for all files!")
def compare_nifti_folders(folder1, folder2, check_content=False, recursive=False):
"""
Compare .nii.gz files in two folders, printing messages for files in folder1
that don't exist in folder2.
Args:
folder1 (str): Path to the first folder
folder2 (str): Path to the second folder
check_content (bool): If True, compare file contents, not just names
recursive (bool): If True, search subdirectories recursively
Returns:
list: List of missing files (relative paths)
"""
def files_are_identical(file1, file2):
"""
Check if two files have identical content using hash comparison.
Args:
file1 (Path): Path to first file
file2 (Path): Path to second file
Returns:
bool: True if files have identical content, False otherwise
"""
# For small files, use direct comparison
if file1.stat().st_size < 100 * 1024 * 1024: # Less than 100MB
return filecmp.cmp(file1, file2, shallow=False)
# For larger files, compare using hashing
return get_file_hash(file1) == get_file_hash(file2)
def get_file_hash(filepath, chunk_size=8192):
"""Calculate SHA-256 hash of a file in chunks to handle large files."""
sha256 = hashlib.sha256()
with open(filepath, "rb") as f:
while True:
data = f.read(chunk_size)
if not data:
break
sha256.update(data)
return sha256.hexdigest()
folder1_path = Path(folder1)
folder2_path = Path(folder2)
# Make sure both folders exist
if not folder1_path.exists():
raise ValueError(f"Source folder does not exist: {folder1}")
if not folder2_path.exists():
raise ValueError(f"Target folder does not exist: {folder2}")
# Get all .nii.gz files in folder1
pattern = "**/*.nii.gz" if recursive else "*.nii.gz"
files1 = list(folder1_path.glob(pattern))
missing_files = []
print(f"Comparing {len(files1)} .nii.gz files from {folder1} with {folder2}...")
for file1 in files1:
# Get relative path if recursive
rel_path = file1.relative_to(folder1_path) if recursive else file1.name
file2 = folder2_path / rel_path
if not file2.exists():
print(f"Missing file: {rel_path}")
missing_files.append(str(rel_path))
elif check_content and not files_are_identical(file1, file2):
print(f"Different content: {rel_path}")
missing_files.append(str(rel_path))
if not missing_files:
print("All files from folder1 exist in folder2")
else:
print(f"Found {len(missing_files)} missing or different files")
return missing_files
def print_unique_values(nii_path, verbose=True):
"""
Print the unique values in a NIfTI (.nii.gz) file
Args:
nii_path (str): Path to the NIfTI file
verbose (bool): If True, print additional statistics
max_display (int): Maximum number of values to display
Returns:
numpy.ndarray: Array of unique values
"""
# Load the NIfTI file
img = nib.load(nii_path)
data = img.get_fdata()
# Get unique values
unique_vals = np.unique(data)
# Print results
filename = os.path.basename(nii_path)
print(f"\nUnique values in {filename}:")
print(f"Total unique values: {len(unique_vals)}")
if verbose:
print(f"Data type: {data.dtype}")
print(f"Min value: {np.min(data)}")
print(f"Max value: {np.max(data)}")
print(f"Data shape: {data.shape}")
# Display all unique values, regardless of the number
print(f"Values: {unique_vals}")
return unique_vals
# Example for processing a directory of files
def print_unique_values_batch(directory):
"""Process all .nii.gz files in a directory"""
for nii_file in Path(directory).glob("*.nii.gz"):
print_unique_values(str(nii_file))
def check_noninteger_labels(folder_path, log_out_dir):
# List to store filenames with non-integer values
non_integer_files = []
# Get total number of files for progress tracking
total_files = sum(
1
for _, _, files in os.walk(folder_path)
for file in files
if file.endswith(".nii.gz")
)
processed = 0
# Walk through the directory
print(f"Checking {total_files} files for non-integer labels...")
for root, dirs, files in os.walk(folder_path):
for file in files:
if file.endswith(".nii.gz"):
processed += 1
file_path = os.path.join(root, file)
print(f" - Checking {processed}/{total_files}: {file}")
try:
img = nib.load(file_path)
data = img.get_fdata()
unique_vals = np.unique(data)
is_all_integer = np.all(np.equal(np.mod(unique_vals, 1), 0))
if not is_all_integer:
non_integer_files.append(
{
"filename": file,
"unique_values": unique_vals.tolist(), # Convert to list for JSON serialization
}
)
except Exception as e:
print(f"\n\nError checking {file}: {str(e)}\n\n")
# Print results and save to file if non-integer files found
if non_integer_files:
print(f"\nMasks with non-integer values in this folder: {folder_path}:\n")
for item in non_integer_files:
print(f"Filename: {item['filename']}")
print("Unique values found:", item["unique_values"], "\n")
# Save to file
with open(f"{log_out_dir}/non_integer_mask_files.json", "w") as f:
json.dump(non_integer_files, f, indent=2)
sys.exit("\n\nError: Non-integer values found in segmentation masks\n\n")
else:
print("\nAll mask files contain integer values only!\n")
def split_4d_nifti(input_dir, out_dir):
"""
Split 4D NIfTI files in the input directory into separate 3D files.
Automatically detects the length of the 4th dimension.
"""
# Get all .nii.gz files in the Images directory
nifti_files = glob.glob(os.path.join(input_dir, "*.nii.gz"))
for file_path in nifti_files:
# Load the NIfTI file
img = nib.load(file_path)
data = img.get_fdata()
# Check if it's a 4D image
if len(data.shape) != 4:
print(f"Skipping {file_path} - not a 4D image")
continue
# Get the length of the 4th dimension
time_points = data.shape[3]
# Create output directories if they don't exist
for i in range(1, time_points + 1):
os.makedirs(f"{out_dir}/Images-{i}", exist_ok=True)
# Get the base filename without extension
base_name = os.path.basename(file_path).replace(".nii.gz", "")
# Split and save each volume
for i in range(time_points):
volume = data[:, :, :, i]
new_img = nib.Nifti1Image(volume, img.affine)
output_path = os.path.join(f"{out_dir}/Images-{i+1}", f"{base_name}.nii.gz")
nib.save(new_img, output_path)
if i == time_points - 1:
print(f"Saved {output_path} (volume {i+1}/{time_points})\n")
else:
print(f"Saved {output_path} (volume {i+1}/{time_points})")
def process_dataset_mm(data_dirs, seg_pattern, modalities, base_suffix, replace=False):
"""Generic function to process multi-modality datasets with different patterns"""
for data_dir in data_dirs:
for seg_file in glob.glob(f"{data_dir}/**/{seg_pattern}", recursive=True):
# Extract base ID
base_id = os.path.basename(seg_file).replace(base_suffix, "")
dir_name = os.path.dirname(seg_file)
# Move segmentation file
mv_cmd = (
shutil.move
if not replace
else lambda src, dst: shutil.move(src, dst, copy_function=shutil.copy2)
)
os.makedirs("Masks", exist_ok=True)
mv_cmd(seg_file, f"Masks/{os.path.basename(seg_file)}")
# Move modality files
for modality in modalities:
if "_" in base_suffix:
img_file = f"{dir_name}/{base_id}_{modality}.nii.gz"
else:
img_file = f"{dir_name}/{base_id}-{modality}.nii.gz"
if os.path.exists(img_file):
os.makedirs(f"Images-{modality}", exist_ok=True)
mv_cmd(img_file, f"Images-{modality}/{os.path.basename(img_file)}")
else:
print(f"Warning: Missing {modality} file for {base_id}")
def process_dataset(
data_dirs,
seg_pattern,
base_suffix,
img_suffix=".nii.gz",
out_dir=None,
replace=False,
masks_fname="Masks",
images_fname="Images",
):
"""
Generic function to process datasets with optional arguments: output directory, file replacement
Logic:
1. Within the <data_dirs> folder, find segmentation files with a given pattern: <seg_pattern>
2. Extract base ID from the segmentation file name by removing the <base_suffix>
3. Move the segmentation file to the <out_dir>/Masks folder (if provided)
4. Find the corresponding image file by appending the <img_suffix> to the base ID
5. Move the image file to the <out_dir>/Images folder (if provided)
"""
for data_dir in data_dirs:
for seg_file in glob.glob(f"{data_dir}/**/{seg_pattern}", recursive=True):
# Extract base ID
base_id = os.path.basename(seg_file).replace(base_suffix, "")
dir_name = os.path.dirname(seg_file)
# Move segmentation file
mv_cmd = (
shutil.move
if not replace
else lambda src, dst: shutil.move(src, dst, copy_function=shutil.copy2)
)
masks_dir = f"{out_dir}/{masks_fname}" if out_dir else masks_fname
os.makedirs(masks_dir, exist_ok=True)
mv_cmd(seg_file, f"{masks_dir}/{os.path.basename(seg_file)}")
# Move image files
img_file = f"{dir_name}/{base_id}{img_suffix}"
if os.path.exists(img_file):
images_dir = f"{out_dir}/{images_fname}" if out_dir else images_fname
os.makedirs(images_dir, exist_ok=True)
mv_cmd(img_file, f"{images_dir}/{os.path.basename(img_file)}")
else:
print(f"Warning: Missing image file for {base_id}")
def match_and_clean_files(images_dir, masks_dir):
print(
f"Checking for matching image and mask files in {images_dir} and {masks_dir}...\n"
)
print("Removing image files without corresponding mask files...\n")
# Get list of image files
image_files = glob.glob(os.path.join(images_dir, "*_0000.nii.gz"))
total_images = len(image_files)
print(f"Found {total_images} image files")
removed_count = 0
# Process each image file
for i, image_path in enumerate(image_files, 1):
# Extract ID from image filename
image_name = os.path.basename(image_path)
image_id = image_name.replace("_0000.nii.gz", "")
# Construct expected mask filename
mask_path = os.path.join(masks_dir, f"{image_id}.nii.gz")
print(f"Checking {i}/{total_images}: {image_name}")
# Check if corresponding mask exists
if not os.path.exists(mask_path):
print(f" - Removing {image_name} - No corresponding mask found")
os.remove(image_path)
removed_count += 1
else:
print(f" - Found matching mask for {image_name}")
print("\nSummary:")
print(f"Total images processed: {total_images}")
print(f"Images removed: {removed_count}")
print(f"Images remaining: {total_images - removed_count}")
def convert_to_serializable(obj):
"""Convert numpy types to Python native types for JSON serialization"""
if isinstance(obj, np.floating):
return float(obj)
elif isinstance(obj, np.integer):
return int(obj)
elif isinstance(obj, np.ndarray):
return obj.tolist()
return obj