-
Notifications
You must be signed in to change notification settings - Fork 0
/
VGG_AD_NET_3d_new_with_MCI_small_net.py
348 lines (275 loc) · 12.4 KB
/
VGG_AD_NET_3d_new_with_MCI_small_net.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
import os
import numpy as np
import struct
import tensorflow as tf
import pandas as pd
import sys
import math
from PIL import Image
FLAGS = None
def weight_variable(shape, weight_name):
# generates random values for initial weights
#initial = tf.truncated_normal(shape, stddev=0.1)
initializer = tf.get_variable(weight_name, shape,
initializer=tf.contrib.layers.xavier_initializer())
#return tf.Variable(initializer)
return initializer;
def bias_variable(shape, bias_name):
initial = tf.constant(0.0, shape=shape)
return tf.Variable(initial)
def conv3d(x, W, layername):
return tf.nn.conv3d(x, W, strides=[1, 1, 1, 1, 1], padding='SAME')
def max_pool_2x2(x, layer_name):
return tf.nn.max_pool3d(x, ksize=[1, 2, 2, 2, 1], strides=[1, 2, 2, 2, 1], padding='SAME')
def next_batch(num, data, labels): ## fix next batch for 3D
# Return a total of `num` random samples and labels.
idx = np.arange(0 , len(data))
np.random.shuffle(idx)
idx = idx[:num]
data_shuffle = [data[ i] for i in idx]
labels_shuffle = [labels[ i] for i in idx]
return np.asarray(data_shuffle), np.asarray(labels_shuffle)
def next_batch_valid(num, data, labels, batch_num): ## fix next batch for 3D
# Return a total of `num` random samples and labels.
idx = np.arange(0 , len(data))
#np.random.shuffle(idx)
#idx = idx[:num]
idx = idx[batch_num*j:num*(batch_num+1)]
data_not_shuffled = [data[ i] for i in idx]
labels_not_shuffled = [labels[ i] for i in idx]
return np.asarray(data_not_shuffled), np.asarray(labels_not_shuffled)
# #######################################
#Loading dataset#
# #######################################
#
#DIR = os.path.dirname(os.path.abspath(__file__))
DIR = os.getcwd() + "/"
print("DIR folder is ", DIR)
train_tmp = np.load('../img_array_train_6k_1.npy')#[:5952]
# use the commands train_all.size or train_all.shape to get info
for i in range(2,18):
train_cur = np.load('../img_array_train_6k_%d.npy' %i)
#stacking all the current train data in a vector - before filtering brain "sides"
train_tmp = np.vstack((train_tmp, train_cur))
train_cur = np.load('../img_array_train_6k_%d.npy' %(i+1))[:52]#[:32]
train_tmp = np.vstack((train_tmp, train_cur))
print("train_tmp.shape is ",train_tmp.shape)
#train_all = []
#train_full_scan_tmp = []
#for i in range(train_tmp.shape[0]): #range(21*6000-1):
# train_full_scan_tmp.append(train_tmp[i])
# if ((i+1) % 62 == 0):
# train_all.append(train_full_scan_tmp)
# train_full_scan_tmp.clear()
train_all_list = []
for i in range(train_tmp.shape[0]): #range(21*6000-1):
train_all_list.append(train_tmp[i])
train_all = np.asarray(train_all_list)
train_all = train_all.reshape(-1, 96, 96, 62) ####fix it to be 62
demo = pd.read_csv('../adni_demographic_master_kaggle.csv')
# gets all the indices of 0 = train and puts all of their
# data in trX_subjs
trX_subjs_train = demo[(demo['train_valid_test']==0)]
# puts the diagnosis in trY 0/1/2
trY = np.asarray(trX_subjs_train.diagnosis)
trY_all = []
for n in range(len(trY)): # len(trY) = 2109
#print("n is " ,n)
#for i in range(62): # duplicating diagnosis for each slice
trY_all.append(trY[n])
print("Sanity check:")
#print("len(train_all) is " ,len(train_all))
#print("trY_all.shape is " ,trY_all.shape)
print("trY_all is " ,len(trY_all))
print ("-----loaded .npy files-----")
# #######################################
#Building network#
# #######################################
x = tf.placeholder(tf.float32, shape=[None, 96 , 96, 62, 1])
y_ = tf.placeholder(tf.float32, shape = [None, 3])
x_image=x
##### first convolutional later #####
W_conv1 = weight_variable([5 ,5 ,5 , 1, 32], "W_conv1")
# 5X5X5 receptive field ,1 input channel, 32 feature maps
b_conv1 = bias_variable([32], "b_conv1")
#32 feature maps - bias
h_conv1 = tf.nn.relu(conv3d(x_image, W_conv1,"first layer") + b_conv1)
# 96X96X62 -> 0-padding -> 100X100X66 -> conv -> 96X96X33X32
h_pool1 = max_pool_2x2(h_conv1,"first layer")
# input: 96X96X32 -> max_pool -> 48X48X16?X32
##### second convolutional layer #####
W_conv2 = weight_variable([5, 5, 5, 32, 64], "W_conv2")
# 5X5 receptive field ,32 input channel, 64 feature maps
b_conv2 = bias_variable([64],"b_conv2")
#64 feature maps - bias
h_conv2 = tf.nn.relu(conv3d(h_pool1, W_conv2,"second layer") + b_conv2)
# input is 48X48X16X32 ----> 48X48X16X64
h_pool2 = max_pool_2x2(h_conv2,"second layer")
# input is 48X48X16X64 ----> 24X24X8X64
'''
##### third convolutional layer #####
W_conv3 = weight_variable([3, 3, 3, 64, 64], "W_conv3")
# 3X3 receptive field ,64 input channel, 64 feature maps
b_conv3 = bias_variable([64],"b_conv3")
#64 feature maps - bias
h_conv3 = tf.nn.relu(conv3d(h_pool2, W_conv3,"third layer") + b_conv3)
# input is 24X24X8X64 ----> 24X24X8X64
h_pool3 = max_pool_2x2(h_conv3,"third layer")
# input is 24X24X8X64 ----> 12X12X4X64
'''
##### fourth convolutional layer #####
W_conv4 = weight_variable([3, 3, 3, 64, 64], "W_conv4")
# 3X3 receptive field ,64 input channel, 128 feature maps
b_conv4 = bias_variable([64],"b_conv4")
#128 feature maps - bias
h_conv4 = tf.nn.relu(conv3d(h_pool2, W_conv4,"fourth layer") + b_conv4)
# input is 12X12X4X64 ----> 12X12X4X128
h_pool4 = max_pool_2x2(h_conv4,"fourth layer")
# input is 12X12X4X128 ----> 6X6X2128
# first fully connected layer
W_fc1 = weight_variable([6*6*4*64, 64],"W_fc1")
b_fc1 = bias_variable([64],"b_fc1")
h_pool4_flat = tf.reshape(h_pool4, [-1, 6*6*4*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool4_flat, W_fc1) + b_fc1)
'''
# second fully connected layer
W_fc2 = weight_variable([64, 64],"W_fc2")
b_fc2 = bias_variable([64],"b_fc2")
#h_pool2_flat = tf.reshape(h_pool2, [-1, 6 * 6 * 128])
h_fc2 = tf.nn.relu(tf.matmul(h_fc1, W_fc2) + b_fc2)
'''
# dropout
#keep_prob = tf.placeholder(tf.float32)
#h_fc_drop = tf.nn.dropout(h_fc2, keep_prob)
# softmax
W_fc3 = weight_variable([64, 3],"W_fc3")
b_fc3 = bias_variable([3],"b_fc3")
y_conv = tf.nn.softmax(tf.matmul(h_fc1, W_fc3) + b_fc3)
print ("-----CNN architecture built-----")
#cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(tf.clip_by_value(y_conv, 1e-10,1.0)), reduction_indices=[1]))
#cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1]))
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y_conv))
#tf.summary.scalar('cross_entropy', cross_entropy)
train_step = tf.train.AdamOptimizer(1e-6).minimize(cross_entropy)
# with tf.name_scope('train'):
# #train_step = tf.train.AdamOptimizer(FLAGS.learning_rate).minimize(cross_entropy)
#train_step = tf.train.GradientDescentOptimizer(1e-3).minimize(cross_entropy)
#with tf.name_scope('accuracy'):
# with tf.name_scope('correct_prediction'):
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
# with tf.name_scope('accuracy'):
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# tf.summary.scalar('accuracy', accuracy)
##correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_,1))
# when the y_conv is equal to given y_ then correct_prediction==1
# when the y_conv prediction is wrong, correct_prediction==1
##accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# we want to minimize the ~accuracy~ parameter, when it is zero - all predictions are correct
saver = tf.train.Saver()
sess = tf.Session()
#merged = tf.summary.merge_all()
#train_writer = tf.summary.FileWriter(DIR,sess.graph)
init = tf.global_variables_initializer()
sess.run(init)
print ("started session")
#restoring older weights if such exists
ckpt = tf.train.get_checkpoint_state(DIR)
print(ckpt)
if ckpt:
print(ckpt.model_checkpoint_path)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
train_lables = []
train_stack = []
for i in range(train_all.shape[0]):
label = [0,0,0]
label[trY_all[i]-1] = 1
label = np.asarray(label)
train_lables.append(label)
#a = train_all[i].reshape(96,96,1)
a = np.expand_dims(train_all[i],96)
train_stack.append(a)
####################### valid set loading and modeling to CNN size #############
valid_tmp = np.load('../img_array_valid_6k_1.npy')
for i in range(2,6):
valid_cur = np.load('../img_array_valid_6k_%d.npy' %i)
valid_tmp = np.vstack((valid_tmp, valid_cur))
# print("sanity check: valid_tmp size suppose to be 26,970, it is: ",valid_tmp.shape)
#
valid_allX_trim_list = []
for i in range (valid_tmp.shape[0]):
#if ((i%62)>19 and (i%62)<40):
valid_allX_trim_list.append(valid_tmp[i])
valid_allX_trim = np.asarray(valid_allX_trim_list)
valid_allX_trim = valid_allX_trim.reshape(-1, 96, 96, 62)
demo = pd.read_csv('../adni_demographic_master_kaggle.csv')
validY_subjs = demo[(demo['train_valid_test']==1)]
validY_before_dup = np.asarray(validY_subjs.diagnosis)
validY_trim_after_dup = []
for n in range(len(validY_before_dup)):
#for i in range(20): # duplicating diagnosis for each slice
validY_trim_after_dup.append(validY_before_dup[n])
validY_trim_after_dup = np.asarray(validY_trim_after_dup)
# validY_trim_after_dup holds all labels for valid set - duplicated 40 times for each subject
validY_stack = []
validX_stack = []
for i in range(len(validY_trim_after_dup)):
label = [0,0,0]
label[validY_trim_after_dup[i]-1] = 1
label = np.asarray(label)
validY_stack.append(label)
# a = valid_allX_trim[i].reshape(96,96,1)
a = np.expand_dims(valid_allX_trim[i],96)
validX_stack.append(a)
##################################################################################
## print sizes of vectors ##
##################################################################################
print("size of train_all should == to train_stack")
#print("train_all ", len(train_all))
print("train_stack ", len(train_stack))
print("train_lables.shape ", len(train_lables))
print("validX_stack ", len(validX_stack))
print("validY_stack ", len(validY_stack))
batch_size = 16 #32
batch_valid_accuracy = 0
batch_train_accuracy = 0
accuracy_sum = 0
valid_accuracy = 0
#print("len(train_all)/batch_size) is " ,math.floor(len(train_all)/batch_size))
#batch_num = int(math.floor(len(train_all)/batch_size))
batch_num = int(math.floor(train_all.shape[0]/batch_size))
batch_num_valid = int(math.floor(valid_allX_trim.shape[0]/batch_size))
print("batch_num = %d" %(batch_num))
for train_iter in range(25000):
print("train_iter number: ", train_iter)
for i in range(batch_num):
Xtr_train, Ytr_train = next_batch(batch_size, train_stack, train_lables)
sess.run(train_step, feed_dict={x:Xtr_train, y_:Ytr_train}) # , keep_prob:0.5
#sess.run(train_step, feed_dict={x:train_stack[batch_size*i:batch_size*i+(batch_size-1)], y_:train_lables[batch_size*i:batch_size*i+(batch_size-1)], keep_prob:0.5})
if i%50 == 0:
#Xtr_validate, Ytr_validate = next_batch(batch_size, validX_stack, validY_stack)
train_accuracy = sess.run(accuracy,feed_dict={x:Xtr_train, y_:Ytr_train}) # , keep_prob:0.5
#valid_accuracy = sess.run(accuracy,feed_dict={x:Xtr_validate, y_:Ytr_validate})
valid_accuracy = 0
accuracy_sum = 0
for j in range(batch_num_valid):
Xtr_validate, Ytr_validate = next_batch_valid(batch_size, validX_stack, validY_stack, j)
valid_accuracy_cur = sess.run(accuracy, feed_dict={x: Xtr_validate, y_: Ytr_validate})
accuracy_sum = accuracy_sum + valid_accuracy_cur
valid_accuracy = accuracy_sum/(batch_num_valid+1)
#batch_valid_accuracy = batch_valid_accuracy + valid_accuracy*100
print("step %d,train accuracy, %f" %(i,train_accuracy*100))
print("step %d, validation accuracy, %f" %(i,valid_accuracy*100))
print("cross_entropy = ",sess.run(cross_entropy,feed_dict={x:Xtr_train, y_:Ytr_train})) # , keep_prob:0.5
#print("correct_prediction = ",sess.run(correct_prediction,feed_dict={x:Xtr_train, y_:Ytr_train}))
###print("step %d,valid accuracy, %f" %(i,valid_accuracy*100))
#print("iteration valid accuracy: %f" %(batch_valid_accuracy/(51)))
batch_valid_accuracy = 0
#summary, _ = sess.run([merged, train_step], feed_dict={x:Xtr_train, y_:Ytr_train, keep_prob:0.5})
#train_writer.add_summary(summary, i)
a = saver.save(sess, DIR+"mode_3d_with_MCI.ckpt")
##print("saved information is:", a)
##print("saved session weights to file")
train_writer.close()
#test_writer.close()