-
Notifications
You must be signed in to change notification settings - Fork 2
/
predict.py
199 lines (156 loc) · 7.48 KB
/
predict.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
import tensorflow as tf
import glob
import os
import pydicom
from dataset.thickness.factory import ThicknessFactory
from tools import resize_3d
import numpy as np
import nibabel as nib
import tools.craft_network as craft_network
import config as config
from tools.predict.factory import PredictFactory
from tools.predict.predict_no_tile import PredictNoTile
from tools.predict.predict_tile import PredictTile
class Predict:
def __read_dcm_slices(self, dcm_folder):
slices = [pydicom.dcmread(_) for _ in glob.glob(os.path.join(dcm_folder, "*.dcm"))]
slices.sort(key = lambda x: float(x.ImagePositionPatient[2]))
if len(slices) == 0:
print("ERROR: no dcim files found")
return slices
# Stored Values (SV) are the values stored in the image pixel data attribute.
# Representation value should be calculated as:
# Rescaled value = SV * Rescale Slope + Rescale Intercept
# https://dicom.innolitics.com/ciods/digital-x-ray-image/dx-image/00281052
def __dcim_slice_stored_value_to_rescaled_value(self, slice):
rescale_intercept = slice.RescaleIntercept if hasattr(slice, "RescaleIntercept") else 0
rescale_slope = slice.RescaleSlope if hasattr(slice, "RescaleSlope") else 1
return slice.pixel_array * rescale_slope + rescale_intercept
def __get_pixel_data(self, dcm_slices):
result = np.array([])
if len(dcm_slices):
result = np.stack([self.__dcim_slice_stored_value_to_rescaled_value(_) for _ in dcm_slices], axis = -1)
else:
print("ERROR: dcim list is empty")
return result
def __preprocess_data(self, data):
#
# keep CT HU in range [pancreas HU]
#
data = tf.minimum(data, config.PANCREAS_MAX_HU)
data = tf.maximum(data, config.PANCREAS_MIN_HU)
#
# scale CT to range [-1, 1]
#
data = (data - tf.reduce_min(data)) / (tf.reduce_max(data) - tf.reduce_min(data)) * 2 - 1
data = data[tf.newaxis, ..., tf.newaxis]
return data
# def __create_segmentation(self, data):
# mask = tf.argmax(data, axis = -1)
# mask = mask[..., tf.newaxis]
# return mask
def __get_patient_position_from_first_frame(self, dcm_slices):
min_number = dcm_slices[0][0x0020, 0x0013].value
min_idx = 0
for idx, _slice in enumerate(dcm_slices):
if _slice[0x0020, 0x0013].value < min_number:
min_number = _slice[0x0020, 0x0013].value
min_idx = idx
return dcm_slices[min_idx][0x0020, 0x0032].value
def __get_metadata(self, dcm_slices):
result = []
dcm_rows = dcm_slices[0][0x0028, 0x0010].value
dcm_columns = dcm_slices[0][0x0028, 0x0011].value
dcm_depth = len(dcm_slices)
dcm_pixel_spacing = dcm_slices[0][0x0028, 0x0030].value
thickness_func = ThicknessFactory(config.THICKNESS)
dcm_slice_thickness_0 = thickness_func(dcm_slices, 0)
dcm_slice_thickness_1 = thickness_func(dcm_slices, 1)
dcm_slice_thickness_2 = thickness_func(dcm_slices, 2)
dcm_patient_orientation = dcm_slices[0][0x0020, 0x0037].value
dcm_patient_position = self.__get_patient_position_from_first_frame(dcm_slices)
affine = np.zeros([4, 4])
affine[0, 0] = dcm_patient_orientation[0] * dcm_pixel_spacing[0]
affine[1, 0] = dcm_patient_orientation[1] * dcm_pixel_spacing[0]
affine[2, 0] = dcm_patient_orientation[2] * dcm_pixel_spacing[0]
affine[0, 1] = dcm_patient_orientation[3] * dcm_pixel_spacing[1]
affine[1, 1] = dcm_patient_orientation[4] * dcm_pixel_spacing[1]
affine[2, 1] = dcm_patient_orientation[5] * dcm_pixel_spacing[1]
affine[0, 3] = dcm_patient_position[0]
affine[1, 3] = dcm_patient_position[1]
affine[2, 3] = dcm_patient_position[2]
affine[2, 0] = dcm_slice_thickness_0
affine[2, 1] = dcm_slice_thickness_1
affine[2, 2] = dcm_slice_thickness_2
# --- inverse axes X and Y. This was found experimental way
# --- could be wrong ...
affine[0, 0] = -affine[0, 0]
affine[1, 0] = -affine[1, 0]
affine[2, 0] = -affine[2, 0]
affine[0, 1] = -affine[0, 1]
affine[1, 1] = -affine[1, 1]
affine[2, 1] = -affine[2, 1]
affine[0, 2] = -affine[0, 2]
affine[1, 2] = -affine[1, 2]
affine[2, 2] = -affine[2, 2]
affine[0, 3] = -affine[0, 3]
affine[1, 3] = -affine[1, 3]
return {"affine": affine, "spacing": dcm_pixel_spacing, "dim": [dcm_rows, dcm_columns, dcm_depth]}
# def __scale_up(self, mask, dcm_slices):
# dcm_rows = dcm_slices[0][0x0028, 0x0010].value
# dcm_columns = dcm_slices[0][0x0028, 0x0011].value
# dcm_depth = len(dcm_slices)
# result = resize_3d.resize_3d_image(tf.squeeze(mask), tf.constant([dcm_rows, dcm_columns, dcm_depth]))
# return result
# def __scale_down(self, data):
# if config.IS_TILE == True:
# data = tf.cast(data, tf.float32)
# elif config.IS_TILE == False:
# data = resize_3d.resize_3d_image(data, tf.constant(
# [config.IMAGE_DIMENSION_X, config.IMAGE_DIMENSION_Y, config.IMAGE_DIMENSION_Z]))
# data = tf.cast(data, tf.float32)
# else:
# raise ValueError("Unknown IS_TILE value: " + config.IS_TILE)
# return data
def __save_img_to_nifti(self, data, meta, result_file_name):
# TODO: add meta information
# affine = meta['affine'][0].cpu().numpy()
# pixdim = meta['pixdim'][0].cpu().numpy()
# dim = meta['dim'][0].cpu().numpy()
# img = nib.Nifti1Image(input_nii_array, affine=affine)
# img.header['dim'] = dim
# img.header['pixdim'] = pixdim
img = nib.Nifti1Image(data, meta["affine"])
# img.header['dim'] = meta["dim"]
# img.header['pixdim'] = meta["spacing"]
nib.save(img, result_file_name)
def __print_stat(self, data, title=""):
if len(title):
print('-' * 25, title, '-' * 25)
print("shape", data.shape)
print("min/mean/max/sum {}/{:.5f}/{}/{}".format(tf.reduce_min(data),
tf.reduce_mean(tf.cast(data, dtype = tf.float32)),
tf.reduce_max(data), tf.reduce_sum(data)))
def main(self, dcm_folder, result_file_name):
model = craft_network.craft_network(config.MODEL_CHECKPOINT)
# model.summary()
predict_class = PredictFactory()("tile" if config.IS_TILE else "no_tile")
predict_obj = predict_class(model)
dcm_slices = self.__read_dcm_slices(dcm_folder)
raw_pixel_data = self.__get_pixel_data(dcm_slices)
scaled_data = predict_obj.scale_down(raw_pixel_data)
# scaled_data = self.__scale_down(raw_pixel_data)
src_data = self.__preprocess_data(scaled_data)
mask = predict_obj.predict(src_data)
# mask = self.__create_segmentation(mask)
# mask = tf.squeeze(mask)
mask = predict_obj.scale_up(mask)
# mask = self.__scale_up(mask, dcm_slices)
metadata = self.__get_metadata(dcm_slices)
self.__save_img_to_nifti(np.asarray(mask.numpy(), dtype = np.uint8), metadata, result_file_name)
self.__print_stat(src_data, "src CT data")
self.__print_stat(mask, "mask")
print(metadata)
if __name__ == "__main__":
pred = Predict()
pred.main("predict", "prediction.nii")