NSCLC-Radiomics-NIFTI / convert.py
farrell236's picture
Update convert.py
5809257 verified
raw
history blame
1.07 kB
import os
import pydicom
import pydicom_seg
import dicom2nifti
import pandas as pd
import SimpleITK as sitk
from tqdm import tqdm
data = pd.read_csv('NSCLC-Radiomics/metadata.csv')
patient_ids = data['Subject ID'].unique()
for pid in tqdm(patient_ids):
row = data[data['Subject ID'] == pid]
out_fn = f'NSCLC-Radiomics-NIFTI/{pid}'
os.makedirs(out_fn)
inp_fn_img = row[row['SOP Class Name'] == 'CT Image Storage']['File Location'].values[0]
dicom2nifti.convert_directory(inp_fn_img, out_fn)
if pid == 'LUNG1-128': continue # LUNG1-128 missing segmentation
inp_fn_seg = row[row['SOP Class Name'] == 'Segmentation Storage']['File Location'].values[0] + '/1-1.dcm'
dcm = pydicom.dcmread(inp_fn_seg)
reader = pydicom_seg.SegmentReader()
result = reader.read(dcm)
for segment_number in result.available_segments:
image = result.segment_image(segment_number) # lazy construction
sitk.WriteImage(image, os.path.join(out_fn, f'seg-{result.segment_infos[segment_number].SegmentDescription}.nii.gz'), True)