def crop_ROI_data_by_label(imgs, labs, outputdir, lab_id,zscore=False):
# 裁剪数据
crop_img_dir=outputdir+"/crop/img"
mkdir_if_not_exist(crop_img_dir)
crop_lab_dir=outputdir+"/crop/lab"
mkdir_if_not_exist(crop_lab_dir)
crop_img_lab(imgs, labs, crop_img_dir, crop_lab_dir, lab_id)
#rot数据
imgs=sort_glob(crop_img_dir+"/*.nii.gz")
labs=sort_glob(crop_lab_dir+"/*.nii.gz")
rot_img_dir=outputdir+"/rot/img/"
mkdir_if_not_exist(rot_img_dir)
rot_lab_dir=outputdir+"/rot/lab/"
mkdir_if_not_exist(rot_lab_dir)
rotate_img_to_same_direction(imgs,labs,rot_img_dir,rot_lab_dir)
#resize数据
imgs=sort_glob(rot_img_dir+"/*.nii.gz")
labs=sort_glob(rot_lab_dir+"/*.nii.gz")
rez_img_dir=outputdir+"/rez/img/"
mkdir_if_not_exist(rez_img_dir)
rez_lab_dir=outputdir+"/rez/lab/"
mkdir_if_not_exist(rez_lab_dir)
resize_img_lab(imgs,labs,rez_img_dir,rez_lab_dir,zscore)
def prepare_unsupervised_reg_data(args):
if not os.path.exists(args.dataset_dir):
mk_or_cleardir(args.dataset_dir)
types=parse_arg_list(args.mode)
for t in types:
imgs=sort_glob("../../dataset/%s/%s-image/*.nii.gz" % (args.task,t))
labs=sort_glob("../../dataset/%s/%s-label/*.nii.gz" % (args.task,t))
crop_ROI_data_by_label(imgs, labs, args.dataset_dir + "/%s/"%(t), args.component)
def prepare_crossvalidation_reg_data(args):
if not os.path.exists(args.dataset_dir):
mk_or_cleardir(args.dataset_dir)
imgs=sort_glob("../../dataset/%s/%s-image/*.nii.gz" % (args.task,args.Tatlas))
labs=sort_glob("../../dataset/%s/%s-label/*.nii.gz" % (args.task,args.Tatlas))
crop_ROI_data_by_label(imgs, labs, args.dataset_dir + "/atlas/", args.component)
#####
imgs=sort_glob("../../dataset/%s/%s-image/*.nii.gz" % (args.task,args.Ttarget))
labs=sort_glob("../../dataset/%s/%s-label/*.nii.gz" % (args.task,args.Ttarget))
crop_ROI_data_by_label(imgs, labs, args.dataset_dir + "/target/", args.component)
def get_file(self):
if self.is_train==False:
sample_dir=self.data_samples[self.index]
self.index=(self.index+1)%self.len
else:
sample_dir = self.data_samples[np.random.randint(len(self.data_samples))]
atlases_imgs=(sort_glob(sample_dir + "/*mv_img*.nii.gz"))
atlases_labs=(sort_glob(sample_dir + "/*mv_lab*.nii.gz"))
fix_imgs=(sort_glob(sample_dir + "/*fix_img*.nii.gz"))
fix_labs=(sort_glob(sample_dir + "/*fix_lab*.nii.gz"))
return fix_imgs,fix_labs,atlases_imgs,atlases_labs
def __init__(self,args,type):
self.args = args
if type=='train':
self.data_samples = sort_glob(self.args.dataset_dir + "/train_fusion_atlas_target/*/")
self.is_train=True
elif type=='validate':
self.data_samples = sort_glob(self.args.dataset_dir + "/atlas_target/*/")
self.is_train=False
# elif type == 'test':
# self.data_samples = sort_glob(self.args.dataset_dir + "/test/*/")
# self.is_train = False
else:
print('unsupport type')
exit('-999')
self.index=0
self.len=len(self.data_samples)
def post_process(args):
base_ref_dir = args.dataset_dir + "/test_target/"
# target_dir=args.test_dir
#测试
# target_dir=args.dataset_dir+"/test_target/rez/lab/"
target_dirs = sort_glob(args.test_dir + "/atlas_wise/*")
for target_dir in target_dirs:
paste_lab_to_ori_space(args, base_ref_dir, target_dir)
def generator_ROI_data(args,type='MMWHS'):
####
imgs=sort_glob("../../dataset/%s/%s-image/*.nii.gz" % (type,args.Tatlas))
labs=sort_glob("../../dataset/%s/%s-label/*.nii.gz" % (type,args.Tatlas))
crop_ROI_data_by_label(imgs[:20], labs[:20], args.dataset_dir + "/train_atlas/", args.component)
# imgs=sort_glob("../../dataset/MMWHS/" + "/%s-image/*.nii.gz" % (args.Tatlas))
# labs=sort_glob("../../dataset/MMWHS/" + "/%s-label/*.nii.gz" % (args.Tatlas))
# prepare_data(imgs[12:16],labs[12:16],args.dataset_dir+"/train_fuse_atlas/",args.component)
# imgs=sort_glob("../../dataset/MMWHS/" + "/%s-image/*.nii.gz" % (args.Tatlas))
# labs=sort_glob("../../dataset/MMWHS/" + "/%s-label/*.nii.gz" % (args.Tatlas))
# prepare_data(imgs[16:],labs[16:],args.dataset_dir+"/test_fuse_atlas/",args.component)
#####
imgs=sort_glob("../../dataset/%s/%s-image/*.nii.gz" % (type,args.Ttarget))
labs=sort_glob("../../dataset/%s/%s-label/*.nii.gz" % (type,args.Ttarget))
crop_ROI_data_by_label(imgs[:20], labs[:20], args.dataset_dir + "/train_target/", args.component)
def __init__(self, args, is_train):
self.args = args
self.is_train = is_train
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/lab'))
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir +
'/test_target/rez/img'))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir +
'/test_target/rez/lab'))
if len(self.img_mv) != len(self.lab_mv):
print("error,number of image and lab not equal")
exit(-900)
self.num = len(self.img_mv)
self.nb_pairs = len(self.img_fix) * len(self.img_mv)
self.index = 0
self.len_mv = len(self.img_mv)
self.len_fix = len(self.img_fix)
def generator_ROI_mask(args):
labs=sort_glob("../../dataset/MMWHS/" + "/%s-test-label/*.nii.gz" % (args.Ttarget))
for p_lab in labs:
lab = sitk.ReadImage(p_lab)
bbox = get_bounding_box_by_id(sitk.GetArrayFromImage(lab), padding=10,id=None)
# sitk.RegionOfInterest(lab,)
# crop_lab = crop_by_bbox(lab, bbox)
array_lab=sitk.GetArrayFromImage(lab)
array_lab[0:,0:,0:]=0
array_lab[bbox[0].start:bbox[0].stop + 1, bbox[1].start:bbox[1].stop + 1,bbox[2].start:bbox[2].stop + 1 ]=1
sitk_write_lab(array_lab,parameter_img=lab,dir='../../dataset/MMWHS'+"/%s-label-test_ROI/"%(args.Ttarget),name=get_name_wo_suffix(p_lab))
def __init__(self, args, type):
self.args = args
validation_size=5
mode=parse_arg_list(args.mode)
if type == 'train':
self.is_train = True
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/img'%(mode[0])))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/lab'%(mode[0])))
# 所有的数据
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/img'%(mode[1])))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/lab'%(mode[1])))
if len(self.img_fix) > 0:
del self.img_fix[(args.fold - 1) * validation_size:(args.fold - 1) * validation_size + validation_size]
del self.lab_fix[(args.fold - 1) * validation_size:(args.fold - 1) * validation_size + validation_size]
if len(self.img_mv) > 0:
del self.img_mv[(args.fold - 1) * validation_size:(args.fold - 1) * validation_size + validation_size]
del self.lab_mv[(args.fold - 1) * validation_size:(args.fold - 1) * validation_size + validation_size]
elif type == 'validate':
self.is_train = False
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/img'%(mode[0])))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/lab'%(mode[0])))
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/img'%(mode[1])))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/%s/rez/lab'%(mode[1])))
if len(self.img_fix)>0:
self.img_fix=[self.img_fix[i] for i in range((args.fold-1)*validation_size,(args.fold-1)*validation_size+validation_size)]
self.lab_fix=[self.lab_fix[i] for i in range((args.fold-1)*validation_size,(args.fold-1)*validation_size+validation_size)]
if len(self.img_mv)>0:
self.img_mv=[self.img_mv[i] for i in range((args.fold-1)*validation_size,(args.fold-1)*validation_size+validation_size)]
self.lab_mv=[self.lab_mv[i] for i in range((args.fold-1)*validation_size,(args.fold-1)*validation_size+validation_size)]
else:
print("not support gen sampler type")
exit(-900)
if len(self.img_mv)!=len(self.lab_mv):
print("error,number of image and lab not equal")
exit(-900)
self.num=len(self.img_mv)
self.nb_pairs=len(self.img_fix)*len(self.img_mv)
self.len_fix=len(self.img_fix)
self.len_mv=len(self.img_mv)
self.index=0
def generator_ROI_data_for_3DUnet(args):
imgs=sort_glob("../../dataset/MMWHS/" + "/%s-image/*.nii.gz" % (args.Ttarget))
labs=sort_glob("../../dataset/MMWHS/" + "/%s-label/*.nii.gz" % (args.Ttarget))
crop_ROI_data_by_label(imgs[:12], labs[:12], args.dataset_dir + "/train_target/", args.component)
imgs=sort_glob("../../dataset/MMWHS/" + "/%s-image/*.nii.gz" % (args.Ttarget))
labs=sort_glob("../../dataset/MMWHS/" + "/%s-label/*.nii.gz" % (args.Ttarget))
crop_ROI_data_by_label(imgs[16:], labs[16:], args.dataset_dir + "/validate_target/", args.component)
imgs=sort_glob("../../dataset/MMWHS/" + "/%s-test-image/*.nii.gz" % (args.Ttarget))
labs=sort_glob("../../dataset/MMWHS/" + "/%s-test-label/*.nii.gz" % (args.Ttarget))
crop_ROI_data_by_label(imgs, labs, args.dataset_dir + "/test_target/", args.component)
def paste_lab_to_ori_space(args, base_ref_dir, target_dir):
print("processing:" + os.path.dirname(target_dir))
base_dir = (target_dir)
refs = sort_glob(base_ref_dir + "/rot/img/*.nii.gz")
tgts = sort_glob(target_dir + "/*label.nii.gz")
output_dir_de_rez = base_dir + "_deRez"
de_resize(refs, tgts, output_dir_de_rez)
refs = sort_glob(base_ref_dir + "/crop/img/*.nii.gz")
tgts = sort_glob(output_dir_de_rez + "/*.nii.gz")
output_dir_de_rot = base_dir + "_deRot"
de_rotate(refs, tgts, output_dir_de_rot)
refs = sort_glob("../../dataset/MMWHS/%s-test-image" % (args.Ttarget) +
"/*.nii.gz")
tgts = sort_glob(output_dir_de_rot + "/*.nii.gz")
output_dir_de_crop = base_dir + "_deCrop"
de_crop(refs, tgts, output_dir_de_crop, args.component)
def __init__(self,args,type):
self.args=args
if type=='train':
self.is_train=True
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +'/train_atlas/rez/lab' ))
#所有的数据
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/img'))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/lab'))
del self.img_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.lab_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.img_fix[(args.fold-1)*5:(args.fold-1)*5+5]
del self.lab_fix[(args.fold-1)*5:(args.fold-1)*5+5]
#for similarity
del self.img_mv[-4:]
del self.lab_mv[-4:]
del self.img_fix[-4:]
del self.lab_fix[-4:]
elif type == 'validate':
self.is_train = False
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +'/train_atlas/rez/lab' ))
#所有的数据
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/img'))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/lab'))
#validation
del self.img_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.lab_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.img_mv[-4:]
del self.lab_mv[-4:]
self.img_fix = [self.img_fix[i] for i in range((args.fold - 1) * 5, (args.fold - 1) * 5 + 5)]
self.lab_fix = [self.lab_fix[i] for i in range((args.fold - 1) * 5, (args.fold - 1) * 5 + 5)]
elif type == 'gen_fusion_train':
self.is_train = True
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +'/train_atlas/rez/lab' ))
#所有的数据
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/img'))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/lab'))
del self.img_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.lab_mv[(args.fold-1)*5:(args.fold-1)*5+5]
del self.img_fix[(args.fold-1)*5:(args.fold-1)*5+5]
del self.lab_fix[(args.fold-1)*5:(args.fold-1)*5+5]
else:
print("not support gen sampler type")
exit(-900)
if len(self.img_mv)!=len(self.lab_mv):
print("error,number of image and lab not equal")
exit(-900)
self.num=len(self.img_mv)
self.nb_pairs=len(self.img_fix)*len(self.img_mv)
self.len_fix=len(self.img_fix)
self.len_mv=len(self.img_mv)
self.index=0
image_array = cv2.cvtColor(image_array, cv2.COLOR_GRAY2BGR)
contours, hierarchy = cv2.findContours(gt_array, cv2.RETR_TREE,
cv2.CHAIN_APPROX_TC89_L1)
contours2, hierarchy2 = cv2.findContours(pred_array, cv2.RETR_TREE,
cv2.CHAIN_APPROX_TC89_L1)
cv2.drawContours(image_array, contours, -1, (0, 0, 255), 1)
cv2.drawContours(image_array, contours2, -1, (255, 0, 0), 1)
image_array = np.flip(image_array, 0)
# image_array=cv2.resize(image_array,(256,256),cv2.INTER_LANCZOS4)
cv2.imshow("liver_contour", image_array)
cv2.waitKey()
out_path = os.path.join(
os.path.dirname(pred_label),
"%s-%d.png" % (os.path.basename(pred_label), slice))
cv2.imwrite(out_path, image_array)
'''
'''
if __name__ == '__main__':
pred_labels = sort_glob(
r'E:\homework\cmmas-tmi\img\registration\ct-mr-1\*' +
"\*label*.nii.gz")
target_label = r"E:\homework\cmmas-tmi\img\registration\ct-mr-1\8_mr_target_label.nii.gz"
target_img = r"E:\homework\cmmas-tmi\img\registration\ct-mr-1\8_mr_target_image.nii.gz"
for lab in pred_labels:
drawContour(target_img, target_label, lab)
def __init__(self, args, type='train_sim'):
self.args = args
if type == 'train_sim':
self.is_train = False
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/lab'))
self.img_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir +
'/train_fuse_target/rez/img'))
self.lab_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir +
'/train_fuse_target/rez/lab'))
elif type == 'fusion':
self.is_train = True
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/lab'))
self.img_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir +
'/train_fuse_target/rez/img'))
self.lab_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir +
'/train_fuse_target/rez/lab'))
#训练融合的代码的时候，把fuse和train_target一起放入进去
# self.img_fix_add = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/img'))
# self.lab_fix_add = sort_glob('{}/*.*'.format(args.dataset_dir + '/train_target/rez/lab'))
# self.img_fix=self.img_fix+self.img_fix_add
# self.lab_fix=self.lab_fix+self.lab_fix_add
elif type == 'test':
self.is_train = False
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/lab'))
self.img_fix = sort_glob('{}/*.*'.format(args.dataset_dir +
'/test_target/rez/img'))
self.lab_fix = sort_glob('{}/*.*'.format(args.dataset_dir +
'/test_target/rez/lab'))
elif type == 'validate':
self.is_train = False
self.img_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/img'))
self.lab_mv = sort_glob('{}/*.*'.format(args.dataset_dir +
'/train_atlas/rez/lab'))
self.img_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir + '/validate_target/rez/img'))
self.lab_fix = sort_glob(
'{}/*.*'.format(args.dataset_dir + '/validate_target/rez/lab'))
else:
print("not support gen sampler type")
exit(-900)
if len(self.img_mv) != len(self.lab_mv):
print("error,number of image and lab not equal")
exit(-900)
self.num = len(self.img_mv)
self.nb_pairs = len(self.img_fix) * len(self.img_mv)
self.index = 0
self.len_mv = len(self.img_mv)
self.len_fix = len(self.img_fix)
for img_dir, lab_dir in zip(image_dirs, lab_dirs):
# files=sort_glob(img_dir+"\\*")
img = sitk_read_dico_series(img_dir)
lab = read_png_seriesV2(lab_dir)
sitk_write_image(img,
dir=output + "//%s-image//" % (type),
name=img_dir.split('\\')[-2] + "_%s_image" % (type))
sitk_write_lab(lab,
parameter_img=img,
dir=output + "//%s-label//" % (type),
name=lab_dir.split('\\')[-2] + "_%s_label" % (type))
if __name__ == "__main__":
image_dirs = sort_glob(
'E:\\consistent_workspace\\dataset\\Ori_CHOAS\\CHAOS_Train_Sets\\Train_Sets\\CT\\*\\DICOM_anon'
)
lab_dirs = sort_glob(
'E:\\consistent_workspace\\dataset\\Ori_CHOAS\\CHAOS_Train_Sets\\Train_Sets\\CT\\*\\Ground'
)
generate_3dCT(image_dirs, lab_dirs, 'ct')
image_dirs = sort_glob(
'E:\\consistent_workspace\\dataset\\Ori_CHOAS\\CHAOS_Train_Sets\\Train_Sets\\MR\\*\\T1DUAL\\DICOM_anon\\InPhase'
)
lab_dirs = sort_glob(
'E:\\consistent_workspace\\dataset\\Ori_CHOAS\\CHAOS_Train_Sets\\Train_Sets\\MR\\*\\T1DUAL\\Ground'
)
generate_3dMR(image_dirs, lab_dirs, 't1_in_DUAL_mr')
image_dirs = sort_glob(
