usefull code in medical image processing and cluster usage

-how to generate a gif from all 2d images in a path:

from PIL import Image
import os
def create_gif(input_folder, output_gif_path, duration=200, loop=0):
# Get all image files from the folder
image_files = [f for f in os.listdir(input_folder) 
              if f.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp', '.gif'))]
# Sort the files if needed (optional)
image_files.sort()
# Open all images and append to a list
images = []
for image_file in image_files:
    image_path = os.path.join(input_folder, image_file)
    try:
        img = Image.open(image_path)
        images.append(img)
    except Exception as e:
        print(f"Could not open {image_file}: {e}")
# Save as GIF
if images:
    images[0].save(output_gif_path,save_all=True,append_images=images[1:],duration=duration,loop=loop)
    print(f"GIF created successfully at {output_gif_path}")
else:
    print("No valid images found to create GIF")
# Example usage
create_gif(input_folder='path/to/your/images',output_gif_path='output.gif',duration=60,  loop=0)

-how to find all ipynb files include "model.eval()" in a path:

grep -r --include="*.ipynb" "model.eval()" /mnt /home /media 2>/dev/null

-how to remove all files in a list.txt from a path:

while read -r file; do echo "Would remove: path/${file}_if_extension.mhd"; done < list.txt;

-how to copy all files from a path1 and subfolders into one path2:

find path1/ -type f \( -name "*.mhd" -o -name "*.zraw" \) | xargs -d '\n' cp -n -t path2/;

-how to create nnunet env in cluster (13/12/2024):

 if you need to train nnunet for segmentation you need to provide images and labels and a json file as:
 nnUNet_results, nnUNet_raw, nnUNet_preprocessed, 
 in folder nnUNet_raw/ you create a dataset with number and name such as:Dataset1001_M01 and put two folders inside along with dataset.jason as imagesTr, labelsTr. images and labels should have the same name but images need  _0000 at the end. in jason file:
 {
"channel_names": {
    "0": "CT"             % this refer to modalities
},
"labels": {
    "background": 0,           % normally where we do not want to segment
    "Emph": 1,                 % the first label which might be only this
    "GGO": 2                   % the secnod and so on with any name
},
"numTraining": 20,             % number of subjects for training, you may use a few to solve issue and then put maximum number 
"file_ending": ".nii.gz",      % refer to the extenstion of files, nii, mhd, ...
"name": "Emph_GGO",            % optional name for your model
"licence": ""
}
 Then, you need to create a nnUnet environment as:
 it might not work in future BTW:
      Creating a clean python 3.10 using venv:
      source /share/apps/examples/source_files/python/python-3.10.0.source
      python3 -m venv nnunetv2_env
      source nnunetv2_env/bin/activate
      pip install torch==2.3.0
      pip install nnunetv2
 if missing library  such as blosc2 … so installed that.
 finally downgrade numpy:
      pip install numpy==1.24
 Had missing path to libc, so added the following env variables declaration under the nnunet exports “export PATH=$PATH:/sbin:/usr/sbin”
 NOTE: Think there are gcc mismatches for the latest versions of torch and annoyingly if you install older versions of torch nnunetv2 updates to 2.1.2 all a bit of a dependency dance tbh.
 Finally, you run this code when source your environment in cluster:
#$ -S /bin/bash
#$ -l gpu=true
#$ -l gpu_type=rtx4090 ##!(gtx1080ti|titanxp|titanx|rtx2080ti)
#$ -l tmem=24G
#$ -l h_rt=100:00:00
#$ -N EmphGGO2
#$ -j y
#$ -wd /...n/code

date
source /share/a..../source_files/python/python-3.10.0.source
source /..../envs/nnUNETV2n124p310t212/bin/activate

export nnUNet_raw="..../GGO_Emph_segmentation/nnUNet_raw"
export nnUNet_preprocessed="..../GGO_Emph_segmentation/nnUNet_preprocessed"
export nnUNet_results=".../GGO_Emph_segmentation/nnUNet_results"
export PATH=$PATH:/sbin:/usr/sbin

nnUNetv2_plan_and_preprocess -d 1001 --verify_dataset_integrity

nnUNetv2_train 1001 3d_fullres all
date
## all= uses all images as training

-how to find a file in current folder and subfolders in cluster:

 $find ./ -type f -name "*part_of_file*"$

-How to automatically set lung window for CT images in Slicer (the parameters inside window level can be modified as you require):

in slicer python interactor window:

all_nodes = slicer.mrmlScene.GetNodesByClass('vtkMRMLScalarVolumeNode')
for ct in all_nodes:
    if isinstance(ct, slicer.vtkMRMLLabelMapVolumeNode):
        continue
    else:
        ct.GetDisplayNode().SetAutoWindowLevel(False)
        ct.GetDisplayNode().SetWindowLevel(1500,-500) 

-How to automatically set blue-red colors for each segment of a segmentation node(10 colors) in Slicer:

in slicer python interactor window:

all_nodes = slicer.mrmlScene.GetNodesByClass('vtkMRMLSegmentationNode')
# Define the color palette
colors = [[0, 0, 255], [0, 85, 255], [0, 170, 255], [0, 255, 255], [170, 255, 255],
     [255, 255, 0], [255, 170, 0], [255, 85, 0], [255, 0, 0], [255, 0, 255]]
# Iterate over all segmentation nodes
for segm in all_nodes:
    # Normalize the colors and assign to each segment
    for ind, color in enumerate(colors):
        # Normalize RGB values to 0-1 range
        normalized_color = [c / 255 for c in color]
        # Ensure the index is within the range of available segments
        if ind < segm.GetSegmentation().GetNumberOfSegments():
            segm.GetSegmentation().GetNthSegment(ind).SetColor(*normalized_color)

-How to stop repeating password on ssh cluster:

$ssh-keygen -t rsa -b 4096 -C "your_email@ucl.ac.uk"
$ssh-copy-id -i ~/.ssh/id_rsa.pub cluster_ID@comic

three times Enter->Enter->Enter

then each time you only need to do:

$ ssh cluster_ID@comic

-How to pull files from cluster to your local machine:

$method 1: rsync
method 2: in linux files→ "Other Locations" → Enter server addresses : ssh://cluster_ID@... and copy-paste
method 3: create list files (e.g., remote_listing.txt) and run a bash file:
REMOTE_USER="youruser"
REMOTE_HOST="node in cluster" # Use little as it is the most stable cluster node
REMOTE_PATH="/cluster/project.../"
LOCAL_PATH="/path_to_local/"
RETRY_LIMIT=10
SLEEP_TIME=5
FILE_LIST="remote_listing.txt" # Text file containing the list of files to be transferred
# Function to perform scp
perform_scp() {
local file=$1
scp "$REMOTE_USER@$REMOTE_HOST:$REMOTE_PATH/$file" "$LOCAL_PATH"
}
# Main loop to iterate over files and retry if necessary
while IFS= read -r file; do
# Check if file already exists locally
if [[ -f "$LOCAL_PATH/$file" ]]; then
echo "File $file already exists locally. Skipping transfer."
continue
fi
 RETRY_COUNT=0
while (( RETRY_COUNT < RETRY_LIMIT )); do
perform_scp "$file"
 # Check the exit status of scp
if [[ $? -eq 0 ]]; then
echo "Successfully transferred $file."
break
else
(( RETRY_COUNT++ ))
echo "Failed to transfer $file. Attempt $RETRY_COUNT of $RETRY_LIMIT. Retrying in $SLEEP_TIME seconds..."
sleep $SLEEP_TIME
fi
done
 if (( RETRY_COUNT == RETRY_LIMIT )); then
echo "Failed to transfer $file after $RETRY_LIMIT attempts."
fi
done < "$FILE_LIST"
echo "File transfer process completed."

-how to hide current running code in terminal:

$install screen: sudo apt install screen

hide current processing:

$screen

bring back current procesing:

$screen -r

-how to find a file in directory or subdirectory in terminal:

$find /path_to_directory/ -type f -name "COPDGene_A68731*"

-How to generate 3D rendered label for QC in Slicer:

in slicer python interactor window:

 import csv
 import DICOMScalarVolumePlugin
 import os
 import pandas as pd
 input_path1='/media/mehran/SSD Storage 2/AAVS/COPDGene_AAVS'
 output_path='/media/mehran/SSD Storage 2/AAVS/COPDGene_AAVS_2d'
 list_files_xlsx=pd.read_excel('/media/mehran/SSD Storage 2/AAVS/List_QC_COPDGene_half.xlsx')
 print('number of summit subjects:',len(list_files_xlsx))
 list_files=[f for f in os.listdir(input_path1) if f.endswith('.nii.gz')]
 for i in range(len(list_files_xlsx)):
     ID=list_files_xlsx.at[i,'ID']
     if ID+'.nii.gz' in list_files:
         slicer.util.loadLabelVolume(input_path1+'/'+ID+'.nii.gz')
         labelMapNode = getNode(ID)
         thresholdValue = 1
         segmentationNode = slicer.mrmlScene.AddNewNodeByClass("vtkMRMLSegmentationNode",'Segm1')
         slicer.modules.segmentations.logic().ImportLabelmapToSegmentationNode(labelMapNode, segmentationNode)
         segmentIDs = vtk.vtkStringArray()
         segmentationNode.GetSegmentation().GetSegmentIDs(segmentIDs)
         displayNode = slicer.mrmlScene.AddNewNodeByClass("vtkMRMLSegmentationDisplayNode",'disp_segm')
         segmentationNode.SetAndObserveDisplayNodeID(displayNode.GetID())
         segmentationDisplayNode = segmentationNode.GetDisplayNode()
         segmentationDisplayNode.SetSegmentVisibility(segmentIDs.GetValue(0), False)
         segmentationDisplayNode.SetSegmentVisibility(segmentIDs.GetValue(2), False)
         slicer.modules.segmentations.logic().ExportVisibleSegmentsToModels(segmentationNode, 0)
         segmentationDisplayNode.SetVisibility3D(True)
         view = slicer.app.layoutManager().threeDWidget(0).threeDView()
         view.resetFocalPoint()
         view.mrmlViewNode().SetBackgroundColor(0,0,0)
         view.mrmlViewNode().SetBackgroundColor2(0,0,0)
         view.mrmlViewNode().SetBoxVisible(False)
         view.mrmlViewNode().SetAxisLabelsVisible(False)
         view.resetCamera()
         view.forceRender()
         renderWindow = view.renderWindow()
         renderWindow.SetAlphaBitPlanes(1)
         wti = vtk.vtkWindowToImageFilter()
         wti.SetInputBufferTypeToRGBA()
         wti.SetInput(renderWindow)
         writer = vtk.vtkJPEGWriter()
         writer.SetFileName(output_path+'/'+ID+'.jpg')
         writer.SetInputConnection(wti.GetOutputPort())
         writer.Write()
         slicer.mrmlScene.Clear(0)
          >>> 

-how to estimate thoracic volume:

 for coronal_idx in range(lung_mask.shape[1]):
     coronal_slice = lung_mask[:, coronal_idx, :]    
     for row_idx in range(coronal_slice.shape[0]):
         row = coronal_slice[row_idx, :]        
         non_zero_indices = np.nonzero(row)[0]        
         if non_zero_indices.size > 0:
             first_idx = non_zero_indices[0]
             last_idx = non_zero_indices[-1]            
             row[first_idx:last_idx+1] = 1       
         coronal_slice[row_idx, :] = row    
     lung_mask[:, coronal_idx, :] = coronal_slice
 filled_lung_mask_sitk = sitk.GetImageFromArray(lung_mask)
 filled_lung_mask_sitk.CopyInformation(img)
 sitk.WriteImage(filled_lung_mask_sitk, 'filled_lung_mask.nii.gz')

-how to combine a folder of xlsx files into one:

 import pandas as pd
 import glob
 import argparse
 import os
 import numpy as np
 
 def merge_excel_files(input_folder, output_file):
     # Construct the file path pattern for all .xlsx files in the input folder
     file_paths = glob.glob(os.path.join(input_folder, "*.xlsx"))
     
     dataframes = []
     for file_path in file_paths:
         df = pd.read_excel(file_path)
         df = df.replace(0, np.nan)
         dataframes.append(df)
     
     # Merge DataFrames by taking the first occurrence of each ID
     merged_df = pd.concat(dataframes, axis=0, join='outer', ignore_index=True).groupby('ID', as_index=False).first()
     
     # Remove any duplicate rows based on the 'ID' column
     merged_df = merged_df.drop_duplicates(subset=['ID'])
     
     # Save the merged DataFrame to the specified output Excel file
     merged_df.to_excel(output_file, index=False)
     print(f"Data merged and saved to {output_file}")
 
 if __name__ == "__main__":
     # Set up argument parsing
     parser = argparse.ArgumentParser(description="Merge .xlsx files in a folder and save the result.")
     parser.add_argument("input_folder", type=str, help="Path to the folder containing .xlsx files to merge")
     parser.add_argument("output_file", type=str, help="Path to save the merged output Excel file")
     
     # Parse the command-line arguments
     args = parser.parse_args()
     
     # Call the merge function with provided arguments
     merge_excel_files(args.input_folder, args.output_file)