filenames = [
x for x in os.listdir(PREPROCESSING_DIR)
if not os.path.isdir((os.path.join(PREPROCESSING_DIR, x)))
]
filenames.sort()
for filename in filenames:
# load nifti file data
nii_obj = nib.load(os.path.join(PREPROCESSING_DIR, filename))
nii_img = nii_obj.get_data()
header = nii_obj.header
affine = nii_obj.affine
# reshape to account for implicit "1" channel
nii_img = np.reshape(nii_img, nii_img.shape + (1, ))
nii_img = pad_image(nii_img)
# segment
segmented_img = apply_model_single_input(nii_img, model)
# save resultant image
segmented_filename = os.path.join(SEG_DIR, filename)
segmented_nii_obj = nib.Nifti1Image(segmented_img,
affine=affine,
header=header)
nib.save(segmented_nii_obj, segmented_filename)
# Reorient back to original before comparisons
print("Reorienting...")
utils.reorient(filename, DATA_DIR, SEG_DIR)
nii_img = nii_obj.get_data()
header = nii_obj.header
affine = nii_obj.affine
# load mask file data
mask_obj = nib.load(os.path.join(PREPROCESSING_DIR, mask))
mask_img = mask_obj.get_data()
# pad and reshape to account for implicit "1" channel
nii_img = np.reshape(nii_img, nii_img.shape + (1,))
orig_shape = nii_img.shape
# if the mask is larger, pad to hardcoded value
if mask_img.shape[0] > nii_img.shape[0] or mask_img.shape[1] > nii_img.shape[1]:
TARGET_DIMS = (656,656,96)
nii_img = pad_image(nii_img, target_dims=TARGET_DIMS)
mask_img = pad_image(mask_img, target_dims=TARGET_DIMS)
else: # otherwise pad normally
nii_img = pad_image(nii_img)
mask_img = pad_image(mask_img, target_dims=nii_img.shape[:3])
# segment
segmented_img = apply_model_single_input(nii_img, model)
pred_shape = segmented_img.shape
# create nii obj
segmented_nii_obj = nib.Nifti1Image(
segmented_img, affine=affine, header=header)
# used only for printing result
mean_dice = 0
pred_vols = []
gt_vols = []
for filename, mask in zip(filenames, masks):
# load nifti file data
nii_obj = nib.load(os.path.join(PREPROCESSING_DIR, filename))
nii_img = nii_obj.get_data()
header = nii_obj.header
affine = nii_obj.affine
# pad and reshape to account for implicit "1" channel
nii_img = np.reshape(nii_img, nii_img.shape + (1, ))
nii_img = pad_image(nii_img)
# segment
segmented_img = apply_model_single_input(nii_img, model)
# save resultant image
segmented_filename = os.path.join(SEG_DIR, filename)
segmented_nii_obj = nib.Nifti1Image(segmented_img,
affine=affine,
header=header)
nib.save(segmented_nii_obj, segmented_filename)
# load mask file data
mask_obj = nib.load(os.path.join(PREPROCESSING_DIR, mask))
mask_img = mask_obj.get_data()
mask_img = pad_image(mask_img)
src_dir=src_dir,
dst_dir=PREPROCESSING_DIR,
tmp_dir=TMPDIR,
verbose=0,
skullstrip_script_path=results.SKULLSTRIP,
remove_tmp_files=True)
######################## SEGMENT FILE ########################
# load nifti file data
nii_obj = nib.load(os.path.join(PREPROCESSING_DIR, filename))
nii_img = nii_obj.get_data()
# reshape to account for implicit "1" channel
nii_img = np.reshape(nii_img, nii_img.shape + (1,))
nii_img, pads = pad_image(nii_img)
# segment
segmented_img = apply_model_single_input(nii_img, model)
# Crop back to original dimensions to undo the padding
(
(left_pad, right_pad),
(top_pad, bottom_pad),
_,
) = pads
if left_pad != 0:
segmented_img = segmented_img[left_pad:, ...]
if right_pad != 0:
segmented_img = segmented_img[:-right_pad, ...]
if top_pad != 0: