fast-mri/scripts/4.frocs.py

213 lines
7.8 KiB
Python
Executable File

import SimpleITK as sitk
import tensorflow as tf
from tensorflow.keras.models import load_model
from focal_loss import BinaryFocalLoss
import numpy as np
import multiprocessing
from functools import partial
import os
from os import path
from sfransen.utils_quintin import *
from sfransen.DWI_exp.helpers import *
from sfransen.DWI_exp.preprocessing_function import preprocess
from sfransen.DWI_exp.callbacks import dice_coef
#from sfransen.FROC.blob_preprocess import *
from sfransen.FROC.cal_froc_from_np import *
from sfransen.load_images import load_images_parrallel
from sfransen.DWI_exp.losses import weighted_binary_cross_entropy
parser = argparse.ArgumentParser(
description='Calculate the froc metrics and store in froc_metrics.yml')
parser.add_argument('-experiment',
help='Title of experiment')
parser.add_argument('--series', '-s',
metavar='[series_name]', required=True, nargs='+',
help='List of series to include')
args = parser.parse_args()
# if __name__ = '__main__':
# bovenstaande nodig om fork probleem op te lossen (windows cs linux)
######## CUDA ################
os.environ["CUDA_VISIBLE_DEVICES"] = "2"
######## constants #############
SERIES = args.series
series_ = '_'.join(args.series)
EXPERIMENT = args.experiment
MODEL_PATH = f'./../train_output/{EXPERIMENT}_{series_}/models/{EXPERIMENT}_{series_}.h5'
YAML_DIR = f'./../train_output/{EXPERIMENT}_{series_}'
IMAGE_DIR = f'./../train_output/{EXPERIMENT}_{series_}'
DATA_DIR = "./../data/Nijmegen paths/"
TARGET_SPACING = (0.5, 0.5, 3)
INPUT_SHAPE = (192, 192, 24, len(SERIES))
IMAGE_SHAPE = INPUT_SHAPE[:3]
DATA_SPLIT_INDEX = read_yaml_to_dict('./../data/Nijmegen paths/train_val_test_idxs.yml')
TEST_INDEX = DATA_SPLIT_INDEX['val_set0']
N_CPUS = 12
########## load images in parrallel ##############
print_(f"> Loading images into RAM...")
# # read paths from txt
# image_paths = {}
# for s in SERIES:
# with open(path.join(DATA_DIR, f"{s}.txt"), 'r') as f:
# image_paths[s] = [l.strip() for l in f.readlines()]
# with open(path.join(DATA_DIR, f"seg.txt"), 'r') as f:
# seg_paths = [l.strip() for l in f.readlines()]
# num_images = len(seg_paths)
# # create pool of workers
# pool = multiprocessing.Pool(processes=N_CPUS)
# partial_images = partial(load_images_parrallel,
# seq = 'images',
# target_shape=IMAGE_SHAPE,
# target_space = TARGET_SPACING)
# partial_seg = partial(load_images_parrallel,
# seq = 'seg',
# target_shape=IMAGE_SHAPE,
# target_space = TARGET_SPACING)
# #load images
# images = []
# for s in SERIES:
# image_paths_seq = image_paths[s]
# image_paths_index = np.asarray(image_paths_seq)[TEST_INDEX][:5]
# data_list = pool.map(partial_images,image_paths_index)
# data = np.stack(data_list, axis=0)
# images.append(data)
# images_list = np.transpose(images, (1, 2, 3, 4, 0))
# print('>>>>> size image_list nmr 1:', np.shape(images_list))
# #load segmentations
# seg_paths_index = np.asarray(seg_paths)[TEST_INDEX][:5]
# data_list = pool.map(partial_seg,seg_paths_index)
# segmentations = np.stack(data_list, axis=0)
########## test with old method #############
image_paths = {}
for s in SERIES:
with open(path.join(DATA_DIR, f"{s}.txt"), 'r') as f:
image_paths[s] = [l.strip() for l in f.readlines()]
with open(path.join(DATA_DIR, f"seg.txt"), 'r') as f:
seg_paths = [l.strip() for l in f.readlines()]
num_images = len(seg_paths)
images = []
images_list = []
segmentations = []
# Read and preprocess each of the paths for each series, and the segmentations.
for img_idx in tqdm(range(len(TEST_INDEX))): #[:40]): #for less images
# print('images number',[TEST_INDEX[img_idx]])
img_s = {f'{s}': sitk.ReadImage(image_paths[s][TEST_INDEX[img_idx]], sitk.sitkFloat32) for s in SERIES}
seg_s = sitk.ReadImage(seg_paths[TEST_INDEX[img_idx]], sitk.sitkFloat32)
img_n, seg_n = preprocess(img_s, seg_s,
shape=IMAGE_SHAPE, spacing=TARGET_SPACING)
for seq in img_n:
images.append(img_n[f'{seq}'])
images_list.append(images)
images = []
segmentations.append(seg_n)
images_list = np.transpose(images_list, (0, 2, 3, 4, 1))
print('>>>>> size image_list nmr 2:', np.shape(images_list), '. equal to: (5, 192, 192, 24, 3)?')
########### load module ##################
print(' >>>>>>> LOAD MODEL <<<<<<<<<')
dependencies = {
'dice_coef': dice_coef,
'weighted_cross_entropy_fn':weighted_binary_cross_entropy
}
reconstructed_model = load_model(MODEL_PATH, custom_objects=dependencies)
# reconstructed_model.summary(line_length=120)
# make predictions on all TEST_INDEX
print(' >>>>>>> START prediction <<<<<<<<<')
predictions_blur = reconstructed_model.predict(images_list, batch_size=1)
############# preprocess #################
# preprocess predictions by removing the blur and making individual blobs
print('>>>>>>>> START preprocess')
def move_dims(arr):
# UMCG numpy dimensions convention: dims = (batch, width, heigth, depth)
# Joeran numpy dimensions convention: dims = (batch, depth, heigth, width)
arr = np.moveaxis(arr, 3, 1)
arr = np.moveaxis(arr, 3, 2)
return arr
# Joeran has his numpy arrays ordered differently.
predictions_blur = move_dims(np.squeeze(predictions_blur))
segmentations = move_dims(np.squeeze(segmentations))
predictions = [preprocess_softmax(pred, threshold="dynamic")[0] for pred in predictions_blur]
# Remove outer edges
zeros = np.zeros(np.shape(predictions))
test = np.squeeze(predictions)[:,2:-2,2:190,2:190]
zeros[:,2:-2,2:190,2:190] = test
predictions = zeros
# perform Froc
metrics = evaluate(y_true=segmentations, y_pred=predictions)
dump_dict_to_yaml(metrics, YAML_DIR, "froc_metrics_focal_10", verbose=True)
############## save image as example #################
# save image nmr 2
img_s = sitk.GetImageFromArray(predictions_blur[2].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_blur_002_old.nii.gz")
img_s = sitk.GetImageFromArray(predictions[2].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_002_old.nii.gz")
img_s = sitk.GetImageFromArray(segmentations[2].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/segmentations_002_old.nii.gz")
img_s = sitk.GetImageFromArray(np.transpose(images_list[2,:,:,:,0].squeeze()))
sitk.WriteImage(img_s, f"{IMAGE_DIR}/t2_002_old.nii.gz")
# save image nmr 3
img_s = sitk.GetImageFromArray(predictions_blur[3].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_blur_003_old.nii.gz")
img_s = sitk.GetImageFromArray(predictions[3].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_003_old.nii.gz")
img_s = sitk.GetImageFromArray(segmentations[3].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/segmentations_003_old.nii.gz")
img_s = sitk.GetImageFromArray(np.transpose(images_list[3,:,:,:,0].squeeze()))
sitk.WriteImage(img_s, f"{IMAGE_DIR}/t2_003_old.nii.gz")
img_s = sitk.GetImageFromArray(np.transpose(images_list[3,:,:,:,1].squeeze()))
sitk.WriteImage(img_s, f"{IMAGE_DIR}/highb_003_old.nii.gz")
# save image nmr 3
img_s = sitk.GetImageFromArray(predictions_blur[4].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_blur_004_old.nii.gz")
img_s = sitk.GetImageFromArray(predictions[4].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/predictions_004_old.nii.gz")
img_s = sitk.GetImageFromArray(segmentations[4].squeeze())
sitk.WriteImage(img_s, f"{IMAGE_DIR}/segmentations_004_old.nii.gz")
img_s = sitk.GetImageFromArray(np.transpose(images_list[4,:,:,:,0].squeeze()))
sitk.WriteImage(img_s, f"{IMAGE_DIR}/t2_004_old.nii.gz")
img_s = sitk.GetImageFromArray(np.transpose(images_list[4,:,:,:,1].squeeze()))
sitk.WriteImage(img_s, f"{IMAGE_DIR}/highb_004_old.nii.gz")