-
Notifications
You must be signed in to change notification settings - Fork 0
/
GridSearch.py
439 lines (325 loc) · 14.3 KB
/
GridSearch.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
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
'''
responsável por realizar uma busca de melhores parâmetros
a ideia é ter uma fórmula matemática para explicar os melhores parâmetros
'''
# para fazer a gpu funcionar no pytorch
# siga este tutorial https://pub.towardsai.net/installing-pytorch-with-cuda-support-on-windows-10-a38b1134535e
# e finalmente instale pip3 install torch==1.8.1+cu111 torchvision==0.9.1+cu111 torchaudio===0.8.1 -f https://download.pytorch.org/whl/torch_stable.html
# from __future__ import print_function
import argparse
import os
from functools import partial
import pydicom
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.autograd import Variable
import cv2
import sys
import numpy as np
import torch.nn.init
import matplotlib.pyplot as plt
import random
import scipy.ndimage
from skimage import measure, morphology
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
import nibabel as nb
import ray
ray.init()
from ray import tune
from ray.tune import CLIReporter
from ray.tune.schedulers import ASHAScheduler
use_cuda = torch.cuda.is_available()
# parser = argparse.ArgumentParser(description='PyTorch Unsupervised Segmentation')
#
# parser.add_argument('--nChannel', metavar='N', default=100, type=int, help='number of channels')
# parser.add_argument('--maxIter', metavar='T', default=50, type=int, help='number of maximum iterations')
# parser.add_argument('--minLabels', metavar='minL', default=5, type=int, help='minimum number of labels')
# parser.add_argument('--lr', metavar='LR', default=0.1, type=float, help='learning rate')
# parser.add_argument('--nConv', metavar='M', default=3, type=int, help='number of convolutional layers')
# parser.add_argument('--visualize', metavar='1 or 0', default=1, type=int, help='visualization flag')
# # parser.add_argument('--input', metavar='FILENAME', default=r'D:\Users\paulo\PycharmProjects\pytorch-unsupervised-segmentation-tip\imagens\3.png', help='input image file name', required=False)
# parser.add_argument('--input', metavar='FILENAME',
# default=r'E:\PycharmProjects\pythonProject\imagens\Dsc32909.jpg',
# help='input image file name', required=False)
# parser.add_argument('--stepsize_sim', metavar='SIM', default=1, type=float, help='step size for similarity loss',
# required=False)
# parser.add_argument('--stepsize_con', metavar='CON', default=1, type=float, help='step size for continuity loss')
# parser.add_argument('--stepsize_scr', metavar='SCR', default=0.5, type=float, help='step size for scribble loss')
# args = parser.parse_args()
def get_pixels_hu_2(scans):
image = np.stack([s.pixel_array for s in scans])
# Convert to int16 (from sometimes int16),
# should be possible as values should always be low enough (<32k)
image = image.astype(np.int16)
# Set outside-of-scan pixels to 0
# The intercept is usually -1024, so air is approximately 0
image[image == -2000] = 0
# Convert to Hounsfield units (HU)
intercept = scans[0].RescaleIntercept
slope = scans[0].RescaleSlope
if slope != 1:
image = slope * image.astype(np.float64)
image = image.astype(np.int16)
image += np.int16(intercept)
return np.array(image, dtype=np.int16)
def get_pixels_hu(slices):
image = np.stack([s.pixel_array for s in slices])
# Convert to int16 (from sometimes int16),
# should be possible as values should always be low enough (<32k)
image = image.astype(np.int16)
# Set outside-of-scan pixels to 0
# The intercept is usually -1024, so air is approximately 0
# realiza o filtro de ar
# Ar −1000
# Pulmão −500
# Gordura −100 a −50
# Água 0
# Fluido cerebroespinhal 15
# Rim 30
# Sangue +30 a +45
# Músculo +10 a +40
# Massa cinzenta +37 a +45
# Massa branca +20 a +30
# Fígado +40 a +60
# Tecidos moles, Contraste +100 a +300
# Osso +700 (osso esponjoso) a +3000 (osso denso)
image[image <= -2000] = 0
image_original = image
plt.imshow(image_original[0, :, ])
plt.show()
intercept = slices[0].RescaleIntercept
slope = slices[0].RescaleSlope
# Convert to Hounsfield units (HU)
for slice_number in range(len(slices)):
if slope != 1:
image_original[slice_number] = slope * image_original[slice_number].astype(np.float64)
image_original[slice_number] = image_original[slice_number].astype(np.int16)
image_original[slice_number] += np.int16(intercept)
return np.array(image_original, dtype=np.int16)
def resample3d(image):
# Determine current pixel spacing
# Set the desired depth
desired_depth = 64
desired_width = 128
desired_height = 128
# Get current depth
current_depth = image.shape[0]
current_width = image.shape[1]
current_height = image.shape[2]
# Compute depth factor
depth = current_depth / desired_depth
width = current_width / desired_width
height = current_height / desired_height
depth_factor = 1 / depth
width_factor = 1 / width
height_factor = 1 / height
image = scipy.ndimage.interpolation.zoom(image, (depth_factor, width_factor, height_factor), order=1)
image = np.transpose(image)
return image
def read_dicom_file(filepath):
slices = [pydicom.read_file(filepath + '/' + s) for s in os.listdir(filepath)]
slices.sort(key=lambda x: int(x.InstanceNumber))
try:
slice_thickness = np.abs(slices[0].ImagePositionPatient[2] - slices[1].ImagePositionPatient[2])
except:
slice_thickness = np.abs(slices[0].SliceLocation - slices[1].SliceLocation)
for s in slices:
s.SliceThickness = slice_thickness
patient1_hu_scans = get_pixels_hu_2(slices)
# images = resample3d(patient1_hu_scans)
#plot_3d(images)
# return images
return patient1_hu_scans
def plot_3d(image, threshold=-300):
# Position the scan upright,
# so the head of the patient would be at the top facing the camera
p = image.transpose(2, 1, 0)
p = p[:, :, ::-1]
verts, faces = measure.marching_cubes(p, threshold)
fig = plt.figure(figsize=(10, 10))
ax = fig.add_subplot(111, projection='3d')
# Fancy indexing: `verts[faces]` to generate a collection of triangles
mesh = Poly3DCollection(verts[faces], alpha=0.1)
face_color = [0.5, 0.5, 1]
mesh.set_facecolor(face_color)
ax.add_collection3d(mesh)
ax.set_xlim(0, p.shape[0])
ax.set_ylim(0, p.shape[1])
ax.set_zlim(0, p.shape[2])
plt.show()
def plotarHistograma(exame):
# plt.hist(exame.flatten(), bins=80, color='c')
plt.hist(exame.flatten(), color='c')
plt.xlabel("Hounsfield Units (HU)")
plt.ylabel("Frequency")
plt.show()
def train(config, checkpoint_dir=None, data_dir=None):
'''
# config['nChannel']
# config['nConv']
# config['lr']
# config['momentum'] = 0.9
# config['maxIter']
# config['stepsize_sim']
# config['stepsize_con']
# config['minLabels']
:param config:
:param checkpoint_dir:
:param data_dir:
:return:
'''
folder_dcm = r"E:\PycharmProjects\pythonProject\exame\CQ500CT257\Unknown Study\CT 0.625mm"
# files_dcm = [os.path.join(os.getcwd(), folder_dcm, x) for x in os.listdir(folder_dcm)]
# exame = np.array([read_dicom_file(path) for path in files_dcm])
exame = np.array([read_dicom_file(folder_dcm)])
exame1 = exame.reshape(256,512,512)
plotarHistograma(exame1)
# Our plot function takes a threshold argument which we can use to plot certain structures, such as all tissue or only the bones.
# 400 is a good threshold for showing the bones only (see Hounsfield unit table above). Let's do this!
# plot_3d(exame1, 400)
# CNN model
class MyNet(nn.Module):
def __init__(self, input_dim):
super(MyNet, self).__init__()
self.conv1 = nn.Conv2d(input_dim, config['nChannel'], kernel_size=3, stride=1, padding=1)
self.bn1 = nn.BatchNorm2d(config['nChannel'])
self.conv2 = nn.ModuleList()
self.bn2 = nn.ModuleList()
for i in range(config['nConv'] - 1):
self.conv2.append(nn.Conv2d(config['nChannel'], config['nChannel'], kernel_size=3, stride=1, padding=1))
self.bn2.append(nn.BatchNorm2d(config['nChannel']))
self.conv3 = nn.Conv2d(config['nChannel'], config['nChannel'], kernel_size=1, stride=1, padding=0)
self.bn3 = nn.BatchNorm2d(config['nChannel'])
def forward(self, x):
x = self.conv1(x)
x = F.relu(x)
x = self.bn1(x)
for i in range(config['nConv'] - 1):
x = self.conv2[i](x)
x = F.relu(x)
x = self.bn2[i](x)
x = self.conv3(x)
x = self.bn3(x)
return x
# train
# model = MyNet(data.size(1))
model = MyNet(1)
print(model)
if use_cuda:
model.cuda()
model.train()
# similarity loss definition
loss_fn = torch.nn.CrossEntropyLoss()
# scribble loss definition
loss_fn_scr = torch.nn.CrossEntropyLoss()
# continuity loss definition
loss_hpy = torch.nn.L1Loss(size_average=True)
loss_hpz = torch.nn.L1Loss(size_average=True)
# HPy_target = torch.zeros(im.shape[0] - 1, im.shape[1], args.nChannel)
# HPz_target = torch.zeros(im.shape[0], im.shape[1] - 1, args.nChannel)
HPy_target = torch.zeros(512 - 1, 512, config['nChannel'])
HPz_target = torch.zeros(512, 512 - 1, config['nChannel'])
if use_cuda:
HPy_target = HPy_target.cuda()
HPz_target = HPz_target.cuda()
optimizer = optim.SGD(model.parameters(), lr=config['lr'], momentum=config['momentum'])
label_colours = np.random.randint(255, size=(100, 3))
# label_colours = np.random.randint(255, size=(100, 1))
nifti_teste = np.ones((256, 512, 512, 3), dtype=np.uint8) # dummy data in numpy matrix
for batch_idx in range(config['maxIter']):
for slice in range(256):
data1 = exame1[slice, :, :]
data = torch.from_numpy(data1.reshape(1,1,512,512).astype('float32'))
if use_cuda:
data = data.cuda()
data = Variable(data)
# forwarding
optimizer.zero_grad()
output1 = model(data)[0]
output = output1.permute(1, 2, 0).contiguous().view(-1, config['nChannel'])
# plt.imshow(output.data.cpu().numpy())
# plt.show()
outputHP = output.reshape((data.shape[2], data.shape[3], config['nChannel']))
HPy = outputHP[1:, :, :] - outputHP[0:-1, :, :]
HPz = outputHP[:, 1:, :] - outputHP[:, 0:-1, :]
lhpy = loss_hpy(HPy, HPy_target)
lhpz = loss_hpz(HPz, HPz_target)
ignore, target = torch.max(output, 1)
im_target = target.data.cpu().numpy()
# plt.imshow(im_target.reshape(191, 194))
# plt.show()
nLabels = len(np.unique(im_target))
# if args.visualize:
im_target_rgb = np.array([label_colours[c % config['nChannel']] for c in im_target])
im_target_rgb = im_target_rgb.reshape(512,512,3).astype(np.uint8)
nifti_teste[slice, :, :] = im_target_rgb
im_target_rgb = cv2.resize(im_target_rgb, (600, 600))
data2 = cv2.resize(data1, (600, 600))
cv2.imshow("output", im_target_rgb)
cv2.imshow("original", data2)
cv2.waitKey(10)
# loss
loss = config['stepsize_sim'] * loss_fn(output, target) + config['stepsize_con'] * (lhpy + lhpz)
loss.backward()
optimizer.step()
torch.save(model.state_dict(), 'results/model.pth')
torch.save(optimizer.state_dict(), 'results/optimizer.pth')
print(batch_idx, '/', config['maxIter'], '|', ' label num :', nLabels, ' | loss :', loss.item())
# tune.report(batch_idx=batch_idx, maxIter=config['maxIter'], nLabels=nLabels, loss=(loss), stepsize_sim=config['stepsize_sim'], stepsize_con=config['stepsize_con'], nChannel=config['nChannel'])
if nLabels <= config['minLabels']:
print("nLabels", nLabels, "reached minLabels", config['minLabels'], ".")
break
tune.report(nChannel=config['nChannel'], nConv=config['nConv'], lr=config['lr'], momentum=config['momentum'], maxIter=config['maxIter'], stepsize_sim= config['stepsize_sim'], stepsize_con=config['stepsize_con'], minLabels=config['minLabels'])
print("Finished Training")
# config['nChannel']
# config['nConv']
# config['lr']
# config['momentum'] = 0.9
# config['maxIter']
# config['stepsize_sim']
# config['stepsize_con']
# config['minLabels']
def main(num_samples=10, max_num_epochs=10, gpus_per_trial=1):
config = {
"nChannel": tune.choice([10, 30, 50, 10, 150]),
"nConv": tune.choice([1, 3, 5, 7]),
"lr": tune.choice([0.1, 0.01, 0.001, 0.0001]),
"momentum": tune.choice([0.9, 0.99, 0.999]),
"maxIter": tune.choice([5, 10, 15]),
"stepsize_sim": tune.choice([0.5, 1, 1.5]),
"stepsize_con": tune.choice([0.5, 1, 1.5]),
"minLabels": tune.choice([6, 5, 4, 3])
}
# numero maximo de iterações
max_num_epochs = 15
num_samples = 10
data_dir = os.path.abspath("./data")
scheduler = ASHAScheduler(
metric="loss",
mode="min",
max_t=max_num_epochs,
grace_period=1,
reduction_factor=2)
reporter = CLIReporter(
# parameter_columns=["l1", "l2", "lr", "batch_size"],
metric_columns=["nChannel", "nConv", "lr", "momentum", "maxIter", "stepsize_sim", "stepsize_con", "minLabels"])
result = tune.run(
partial(train, data_dir=data_dir),
resources_per_trial={"cpu": 8, "gpu": 1},
config=config,
num_samples=num_samples,
scheduler=scheduler,
progress_reporter=reporter,
checkpoint_at_end=True)
a = 45
b = 90
best_trial = result.get_best_trial("loss", "min", "last")
print("Best trial config: {}".format(best_trial.config))
print("Best trial final validation loss: {}".format(best_trial.last_result["loss"]))
print("Best trial final validation accuracy: {}".format(best_trial.last_result["accuracy"]))
if __name__ == "__main__":
# You can change the number of GPUs per trial here:
main(num_samples=10, max_num_epochs=10, gpus_per_trial=0)