Neuroimaging-Example/src/eval.py

import numpy as np
from .const_eval import *
from .functions import *
from .model_architecture import build_model

t1 = [os.path.join(os.getcwd(),'data/raw/HGG/Brats18_2013_10_1/Brats18_2013_10_1_t1.nii.gz')]
t2 = [os.path.join(os.getcwd(),'data/raw/HGG/Brats18_2013_10_1/Brats18_2013_10_1_t2.nii.gz')]
flair = [os.path.join(os.getcwd(),'data/raw/HGG/Brats18_2013_10_1/Brats18_2013_10_1_flair.nii.gz')]
t1ce = [os.path.join(os.getcwd(),'data/raw/HGG/Brats18_2013_10_1/Brats18_2013_10_1_t1ce.nii.gz')]

eval_data: np.array = np.empty((1,) + input_shape, dtype=np.float32)
for index, item in enumerate(zip(t1, t2, flair, t1ce)):
    eval_data[index] = np.array(
        [process_img(load_Nift2np(modal_path), out_shape=input_shape[1:]) for modal_path in item])
    break

model = build_model(input_shape=input_shape, output_channels=output_channels)

model.load_weights(os.path.join(os.getcwd(),'model_weights/model_weights.h5'))

pred = model.predict(eval_data)

base_path = os.path.join(os.getcwd(),'data/eval/')

if not os.path.isdir(base_path):
    os.makedirs(base_path)

for index,modal in enumerate(['t1', 't2', 'flair', 't1ce']):
  ni_img = nib.Nifti1Image(pred[1][0,index,:,:,:], affine=np.eye(4))
  nib.save(ni_img, base_path+f'{modal}.nii.gz')