-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathFiANet.py
225 lines (175 loc) · 7.21 KB
/
FiANet.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
#*****************************************************
#
# This is the PyTorch code for our paper:
# Multi-channel attention-fusion neural network for brain age estimation: Accuracy, generality, and interpretation with 16,705 healthy MRIs across lifespan
# Medical Image Analysis, Volume 72, 2021, 102091
#
# @email: [email protected]
#
#*****************************************************
import torch
import torch.nn as nn
def conv3d(in_planes, out_planes, stride=1, kernel_size=3, groups=1, dilation=1):
"""3x3 convolution with padding"""
padding = (kernel_size-1)//2
return nn.Conv3d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
padding=padding, groups=groups, bias=False, dilation=dilation)
class BasicBlock(nn.Module):
expansion = 1
__constants__ = ['downsample']
def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
base_width=64, dilation=1, norm_layer=None,conv_layer=conv3d):
super(BasicBlock, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2d
#if groups != 1 or base_width != 64:
#raise ValueError('BasicBlock only supports groups=1 and base_width=64')
if dilation > 1:
raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
# Both self.conv1 and self.downsample layers downsample the input when stride != 1
self.conv1 = conv_layer(inplanes, planes, stride)
self.bn1 = norm_layer(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv_layer(planes, planes)
self.bn2 = norm_layer(planes)
self.downsample = downsample
self.stride = stride
def forward(self, x):
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class first_conv(nn.Module):
def __init__(self,inplace,outplace,norm_layer,conv_layer):
super().__init__()
self.conv = conv_layer(inplace,outplace,kernel_size=1)
self.bn = norm_layer(outplace)
self.relu = nn.ReLU(inplace=True)
def forward(self,x):
x = self.conv(x)
x = self.bn(x)
x = self.relu(x)
return x
class Resnet(nn.Module):
def __init__(self,num_classes=1,in_channels=1):
super().__init__()
self.bn3 = nn.InstanceNorm3d
self.cv3 = conv3d
self.dilation = 1
self.groups = 1
self.inplanes = 64
self.base_width = 64
layers = [2, 2, 2, 2]
n_channel = [64,128,256,512]
self.conv3d_1 = first_conv(in_channels,outplace=self.inplanes,norm_layer=self.bn3,conv_layer=self.cv3)
self.conv3d_layer1 = self._make_layer(BasicBlock, n_channel[0], layers[0],norm_layer=self.bn3,conv_layer=self.cv3)
self.conv3d_layer2 = self._make_layer(BasicBlock, n_channel[1], layers[1],stride=2,norm_layer=self.bn3,conv_layer=self.cv3)
self.conv3d_layer3 = self._make_layer(BasicBlock, n_channel[2], layers[2],stride=2,norm_layer=self.bn3,conv_layer=self.cv3)
self.conv3d_layer4 = self._make_layer(BasicBlock, n_channel[3], layers[3],stride=2,norm_layer=self.bn3,conv_layer=self.cv3)
self.avgpool_3d = nn.AdaptiveAvgPool3d(1)
self.regression_3d = nn.Linear(n_channel[3],num_classes)
def _make_layer(self, block, planes, blocks, norm_layer, conv_layer,stride=1, dilate=False):
downsample = None
previous_dilation = self.dilation
if dilate:
self.dilation *= stride
stride = 1
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
conv_layer(self.inplanes, planes * block.expansion, stride,kernel_size=1),
norm_layer(planes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
self.base_width, previous_dilation, norm_layer,conv_layer=conv_layer))
self.inplanes = planes * block.expansion
for _ in range(1, blocks):
layers.append(block(self.inplanes, planes, groups=self.groups,
base_width=self.base_width, dilation=self.dilation,
norm_layer=norm_layer,conv_layer=conv_layer))
return nn.Sequential(*layers)
def forward(self,x):
x3d_0 = self.conv3d_1(x)
x3d_1 = self.conv3d_layer1(x3d_0)
x3d_2 = self.conv3d_layer2(x3d_1)
x3d_3 = self.conv3d_layer3(x3d_2)
x3d_4 = self.conv3d_layer4(x3d_3)
feat3d = self.avgpool_3d(x3d_4)
feat3d = torch.flatten(feat3d, 1)
o3d = self.regression_3d(feat3d)
return x3d_1,x3d_2,x3d_3,x3d_4,o3d
class fusion(nn.Module):
def __init__(self,inplace,outplace,first):
super().__init__()
self.first = first
self.conv = nn.Conv3d(4 * inplace, outplace,3,1,1,bias=False)
self.bn = nn.InstanceNorm3d(outplace)
self.relu = nn.ReLU(inplace=True)
if self.first:
self.conv2 = nn.Conv3d(outplace, outplace,3,1,1,bias=False)
else:
self.conv2 = nn.Conv3d(outplace+inplace, outplace,3,1,1,bias=False)
self.bn2 = nn.InstanceNorm3d(outplace)
self.kernel_se_conv = nn.Conv3d(2*inplace,inplace,3,1,1)
self.convs1 = nn.Conv3d(inplace,inplace,3,1,1,bias=False)
self.convs2 = nn.Conv3d(inplace,inplace,3,1,1,bias=False)
self.convm1 = nn.Conv3d(inplace,inplace,3,1,1,bias=False)
self.convm2 = nn.Conv3d(inplace,inplace,3,1,1,bias=False)
#self.softmax = nn.Softmax(dim=1)
def forward(self,x1,x2,z=None):
x = torch.cat([x1,x2],1)
x = self.kernel_se_conv(x)
x = torch.sigmoid(x)
y1 = x * x1 + (1-x) * x2
y2 = torch.sigmoid(self.convs1(x1)) * x2
y3 = torch.sigmoid(self.convs2(x2)) * x1
y4 = torch.max(self.convm1(x1),self.convm2(x2))
y = torch.cat([y1,y2,y3,y4],1)
y = self.relu(self.bn(self.conv(y)))
if self.first:
y = self.relu(self.bn2(self.conv2(y)))
else:
y = torch.cat([y,z],1)
y = self.relu(self.bn2(self.conv2(y)))
return y
class fusNet(nn.Module):
def __init__(self,num_classes=1):
super().__init__()
self.net1 = Resnet(num_classes)
self.net2 = Resnet(num_classes)
self.fus1 = fusion(64,128,True)
self.fus2 = fusion(128,256,False)
self.fus3 = fusion(256,512,False)
self.fus4 = fusion(512,512,False)
self.maxp = nn.MaxPool3d(2,2)
self.avgpool_3d = nn.AdaptiveAvgPool3d(1)
self.regression_3d = nn.Linear(512,num_classes)
def forward(self,ix1,ix2):
x1,x2,x3,x4,f1 = self.net1(ix1)
y1,y2,y3,y4,f2 = self.net2(ix2)
z = self.fus1(x1,y1)
z = self.maxp(z)
z = self.fus2(x2,y2,z)
z = self.maxp(z)
z = self.fus3(x3,y3,z)
z = self.maxp(z)
z = self.fus4(x4,y4,z)
z = self.maxp(z)
z = self.avgpool_3d(z)
z = torch.flatten(z,1)
f3 = self.regression_3d(z)
return f1,f2,f3
if __name__ == '__main__':
x = torch.rand(1,1,64,64,64)
y = torch.rand(1,1,64,64,64)
mod = fusNet()
f = mod(x,y)
for i in f:
print(i.shape)