def get_frames_for_sample(sample):
path = os.path.join(cg.local_dir, 'images', sample['Patient_Class'],
sample['Patient_ID'], sample['video_name_no_ext'])
files = ff.find_all_target_files(['*.jpg'], path)
images = sorted(files)
return images
print(excel_file.shape)
# build the excel sheet
result = []
for i in range(excel_file.shape[0]):
case = excel_file.iloc[i]
patient_id = case['Patient_ID']
patient_class = case['Patient_Class']
print(patient_class, patient_id)
# find all the movies belonged to this case
movie_list = ff.sort_timeframe(
ff.find_all_target_files(['*.avi'],
os.path.join(cg.fc_main_dir,
'avi_movie_collection',
patient_class, patient_id,
'Volume_Rendering_Movies')), 1,
'_')
assert len(movie_list) == 6
for m in movie_list:
# get video_name and video_name_no_ext
video_name = os.path.basename(m)
video_name_sep = video_name.split('.') # remove .avi
video_name_no_ext = video_name_sep[0]
if len(video_name_sep) > 2:
for ii in range(1, len(video_name_sep) - 1):
video_name_no_ext += '.'
video_name_no_ext += video_name_sep[ii]
# get angle
# get patient_list:
patient_id_list = np.unique(excel_file['Patient_ID'])
patient_list = []
for p in patient_id_list:
d = excel_file[excel_file['Patient_ID'] == p].iloc[0]
patient_list.append([d['Patient_Class'],p])
# get their segmentations and find the ES
timeframe_pick = []
for p in patient_list:
patient_class = p[0]
patient_id = p[1]
# find segmentations
segmentations = ff.sort_timeframe(ff.find_all_target_files(['pred_*.nii.gz'],os.path.join(cg.nas_main_dir,'predicted_seg',patient_class,patient_id,'seg-pred-0.625-4classes-connected-retouch-downsample')),2,'_')
lv_volume_list = []
for s in segmentations:
data = nb.load(s).get_fdata()
count,_ = ff.count_pixel(data,1)
lv_volume_list.append(count)
min_val = min(lv_volume_list)
lv_volume_list = np.asarray(lv_volume_list)
index = np.where(lv_volume_list == min_val)[0]
if index.shape[0] > 1:
ValueError('two mins')
ED = 0
ES = ff.find_timeframe(segmentations[index[0]],2,'_')
if (ES-ED) % 2 == 0:
mid_ES = (ES-ED) // 2
model_BASAL_t = 'model_batch2/model-U2_batch2_BASAL_t-032-*'
model_BASAL_r = 'model_batch1/model-U2_batch1_BASAL_r-018-*'
MODEL = [
model_s, model_2C_t, model_2C_r, model_3C_t, model_3C_r, model_4C_t,
model_4C_r, model_BASAL_t, model_BASAL_r
]
# prediction task list
task_list = [
's', '2C_t', '2C_r', '3C_t', '3C_r', '4C_t', '4C_r', 'BASAL_t', 'BASAL_r'
]
task_num_list = [0]
# define patient CT image list
patient_list = ff.find_all_target_files(['Abnormal/*', 'Normal/*'],
os.path.join(cg.main_dir,
'nii-images'))
patient_list = np.repeat(patient_list, 30)
print(len(patient_list))
print('finish loading all patient images')
# define save folder:
save_folder = os.path.join(cg.main_dir, 'predicted_seg')
# load time assignment sheet
two_timeframes = 1
################################
# build the model
shape = cg.dim + (1, )
model_inputs = [Input(shape)]
'_TO_', '_to_', 'CCTA', 'CTA', 'ccta', 'cta', 'ccta', 'HALF', 'Function',
'function'
]
dicom_parameter_list = [
'AccessionNumber', 'Manufacturer', 'ManufacturerModelName',
'StudyDescription', 'PatientSex', 'PatientAge', 'ProtocolName'
]
# put each year into one excel file
year_list = ['2020']
for Y in year_list:
#year = ff.find_all_target_files([Y],os.path.join('/Data/McVeighLabSuper/wip/zhennong/'))
#patients = ff.find_all_target_files(['AN*','CVC*','cvc*','Cvc*'],year[0])
patients = ff.find_all_target_files(
['0*/*', '1*/*'], os.path.join('/Data/McVeighLabSuper/dicom_images/',
Y))
patient_id = []
scan_year = []
function = []
directories = []
directories_sub = []
Dicom = []
AccessionNumber = []
Manufacturer = []
ModelName = []
StudyDescription = []
Sex = []
Age = []
##
# this script can do file copy and removal
##
import os
import numpy as np
import function_list_VR as ff
import shutil
import pandas as pd
#### copy files
main_folder = '/Data/McVeighLabSuper/wip/zhennong/'
patient_list = ff.find_all_target_files(['Abnormal/*/img-nii-0.625'],
os.path.join(main_folder,
'nii-images'))
save_folder = os.path.join(main_folder, 'upsample-nii-images')
for p in patient_list:
patient_id = os.path.basename(os.path.dirname(p))
patient_class = os.path.basename(os.path.dirname(os.path.dirname(p)))
print(patient_id, patient_class)
ff.make_folder([
os.path.join(save_folder, patient_class, patient_id),
os.path.join(save_folder, patient_class, patient_id, 'img-nii-0.625')
])
img_list = ff.find_all_target_files(['*.nii.gz'], p)
for i in img_list:
print(os.path.basename(i))
break
if (frameId % math.floor(frameRate) == 0):
if count < 10:
number = '000'+str(count)
if count >=10:
number = '00'+str(count)
dest = os.path.join(save_folder,case['video_name_no_ext']+'-'+number+'.jpg')
cv2.imwrite(dest,frame)
count += 1
cap.release()
# call(["ffmpeg", "-i", src, dest]) % this will cause some error (not extract exact 20 frames) in some avis
# Now get how many frames it is.
nb_frames = len(ff.find_all_target_files(['*.jpg'],save_folder))
data.append([case['video_name'],nb_frames])
print('done extraction')
data_df = pd.DataFrame(data,columns = ['video_name','nb_frames'])
data_file = pd.merge(excel_file,data_df,on = "video_name")
data_file.to_excel(os.path.join(cg.fc_main_dir,'Patient_List/movie_list_w_classes_w_picked_timeframes_test.xlsx'),index=False)
def main():
main_path = cg.local_dir
movie_path = os.path.join(main_path,'original_movie')
excel_file = os.path.join(cg.fc_main_dir,'Patient_List/movie_list_w_classes_w_picked_timeframes_test.xlsx')
extract_timeframes(main_path,movie_path,excel_file)
# this script can assign the time frame of each patient that pre-trained AI will predict on
# the main point here is that to collect a dataset with pre-trained AI predicted segmentation on some
# time frame, we should not only have time frame 0 but should half ED half ES
import os
import numpy as np
import function_list_VR as ff
import shutil
import pandas as pd
np.random.seed(0)
main_folder = '/Data/McVeighLabSuper/wip/zhennong/'
# find all patients in the training dataset
patient_list = ff.find_all_target_files(['Abnormal/*', 'Normal/*'],
os.path.join(main_folder,
'nii-images'))
total_num = len(patient_list)
print(total_num) # 375
# make ED = 0 and ES = 1 randoml shuffle list
a = [0] * int((total_num / 2)) # 187
a.extend([1] * int(total_num - int(total_num / 2)))
a = np.asarray(a)
np.random.shuffle(a)
print(a, a.shape)
# assign to each patient
result = []
for i in range(0, len(patient_list)):
p = patient_list[i]