def run_over_subjects(print_only=False, use_scaled_masks=False):
for subject_fol in utils.get_subfolders(SPM_ROOT):
subject = utils.namebase(subject_fol)
spm_brain_file = SPM_BRAIN_TEMPLATE.format(subject=subject.upper())
reg_file = '{}_register.lta'.format(subject)
reg_spm_brain = '{}_reg.mgz'.format(subject)
spm_map = os.path.basename(
glob.glob(os.path.join(subject_fol, 'spmT_*.nii'))[0])
spm_mask = SPM_MASK_TEMPLATE.format(subject=subject.upper())
if use_scaled_masks:
spm_mask_name, spm_mask_type = os.path.splitext(spm_mask)
spm_mask = '{}_scaled{}'.format(spm_mask_name, spm_mask_type)
spm_map_masked = '{}_masked.mgz'.format(os.path.splitext(spm_map)[0])
fs_hemi_map = FS_HEMI_MAP_TEMPLATE.format(subject=subject,
hemi='{hemi}')
run(subject_fol,
spm_brain_file,
FS_BRAIN_FILE,
reg_file,
reg_spm_brain,
spm_map,
spm_mask,
spm_map_masked,
fs_hemi_map,
fs_subject=FS_SUBJECT,
print_only=print_only)
def run_over_subjects(print_only=False, use_scaled_masks=False):
for subject_fol in utils.get_subfolders(SPM_ROOT):
subject = utils.namebase(subject_fol)
spm_brain_file = SPM_BRAIN_TEMPLATE.format(subject=subject.upper())
reg_file = "{}_register.lta".format(subject)
reg_spm_brain = "{}_reg.mgz".format(subject)
spm_map = os.path.basename(glob.glob(os.path.join(subject_fol, "spmT_*.nii"))[0])
spm_mask = SPM_MASK_TEMPLATE.format(subject=subject.upper())
if use_scaled_masks:
spm_mask_name, spm_mask_type = os.path.splitext(spm_mask)
spm_mask = "{}_scaled{}".format(spm_mask_name, spm_mask_type)
spm_map_masked = "{}_masked.mgz".format(os.path.splitext(spm_map)[0])
fs_hemi_map = FS_HEMI_MAP_TEMPLATE.format(subject=subject, hemi="{hemi}")
run(
subject_fol,
spm_brain_file,
FS_BRAIN_FILE,
reg_file,
reg_spm_brain,
spm_map,
spm_mask,
spm_map_masked,
fs_hemi_map,
fs_subject=FS_SUBJECT,
print_only=print_only,
)
def scale_masks(scale = 10):
for subject_fol in utils.get_subfolders(SPM_ROOT):
subject = utils.namebase(subject_fol)
spm_mask = SPM_MASK_TEMPLATE.format(subject=subject.upper())
spm_scaled_mask = os.path.join(subject_fol, '{}_scaled{}'.format(os.path.splitext(spm_mask)[0],os.path.splitext(spm_mask)[1]))
img = nib.load(os.path.join(subject_fol, spm_mask))
data = img.get_data()
affine = img.get_affine()
scaled_data = data * scale
new_img = nib.Nifti1Image(scaled_data, affine)
nib.save(new_img, spm_scaled_mask)
def check_colors():
subjects_folders = utils.get_subfolders(SPM_ROOT)
good_subjects = ['pp002', 'pp003', 'pp004', 'pp005', 'pp006']
subjects_folders = [os.path.join(SPM_ROOT, sub) for sub in good_subjects]
subjects_colors = utils.get_spaced_colors(len(subjects_folders))
# subjects_colors = utils.arr_to_colors(range(len(subjects_folders)), colors_map='Set1')
plt.figure()
for subject_fol, color in zip(subjects_folders, subjects_colors):
subject = utils.namebase(subject_fol)
plt.scatter([0], [0], label='{} {}'.format(subject, color), c=color)
plt.legend()
plt.show()
def check_colors():
subjects_folders = utils.get_subfolders(SPM_ROOT)
good_subjects = ["pp002", "pp003", "pp004", "pp005", "pp006"]
subjects_folders = [os.path.join(SPM_ROOT, sub) for sub in good_subjects]
subjects_colors = utils.get_spaced_colors(len(subjects_folders))
# subjects_colors = utils.arr_to_colors(range(len(subjects_folders)), colors_map='Set1')
plt.figure()
for subject_fol, color in zip(subjects_folders, subjects_colors):
subject = utils.namebase(subject_fol)
plt.scatter([0], [0], label="{} {}".format(subject, color), c=color)
plt.legend()
plt.show()
def scale_masks(scale=10):
for subject_fol in utils.get_subfolders(SPM_ROOT):
subject = utils.namebase(subject_fol)
spm_mask = SPM_MASK_TEMPLATE.format(subject=subject.upper())
spm_scaled_mask = os.path.join(
subject_fol, "{}_scaled{}".format(os.path.splitext(spm_mask)[0], os.path.splitext(spm_mask)[1])
)
img = nib.load(os.path.join(subject_fol, spm_mask))
data = img.get_data()
affine = img.get_affine()
scaled_data = data * scale
new_img = nib.Nifti1Image(scaled_data, affine)
nib.save(new_img, spm_scaled_mask)
horizontal_flip=args.h_flip,
vertical_flip=args.v_flip)
val_datagen = ImageDataGenerator(
preprocessing_function=preprocessing_function)
train_generator = train_datagen.flow_from_directory(TRAIN_DIR,
target_size=(HEIGHT,
WIDTH),
batch_size=BATCH_SIZE)
validation_generator = val_datagen.flow_from_directory(
VAL_DIR, target_size=(HEIGHT, WIDTH), batch_size=BATCH_SIZE)
# Save the list of classes for prediction mode later
class_list = utils.get_subfolders(TRAIN_DIR)
utils.save_class_list(class_list, model_name=args.model, dataset_name="")
finetune_model = utils.build_finetune_model(base_model,
dropout=args.dropout,
fc_layers=FC_LAYERS,
num_classes=len(class_list))
if args.continue_training:
finetune_model.load_weights("./checkpoints/" + args.model +
"_model_weights.h5")
print("load success!")
adam = Adam(lr=0.00001)
finetune_model.compile(adam,
loss='categorical_crossentropy',
val_datagen = ImageDataGenerator(
preprocessing_function=preprocessing_function)
train_generator = train_datagen.flow_from_directory(
BASE_IMG_DIR + TRAIN_DIR,
target_size=(HEIGHT, WIDTH),
batch_size=BATCH_SIZE)
validation_generator = val_datagen.flow_from_directory(
BASE_IMG_DIR + VAL_DIR,
target_size=(HEIGHT, WIDTH),
batch_size=BATCH_SIZE)
# Save the list of classes for prediction mode later
class_list = utils.get_subfolders(BASE_IMG_DIR + TRAIN_DIR)
utils.save_class_list(class_list, model_name=args.model)
finetune_model = utils.build_finetune_model(base_model,
dropout=args.dropout,
fc_layers=FC_LAYERS,
num_classes=len(class_list))
if args.continue_training:
finetune_model.load_weights("./checkpoints/" + args.model +
"_model_weights.h5")
adam = Adam(lr=0.00001)
finetune_model.compile(adam,
loss='categorical_crossentropy',
metrics=['accuracy'])
