Models:
Samuel Callahan
SCallahan/
mri-breast-tumor-segm
0
Downloads: 1
[9ab38d]: / Segmenation.py

Download this file

221 lines (149 with data), 8.5 kB 

  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
import numpy as np

import torch

from scipy import ndimage as nd  

import torch.nn.parallel

from torch.autograd import Variable



def Generator_multichannels(image, sizeofchunk, sizeofchunk_expand,numofchannels): 

    sizeofimage = np.shape(image)[1:4]

            

    nb_chunks = (np.ceil(np.array(sizeofimage)/float(sizeofchunk))).astype(int)

    

    pad_image = np.zeros(([numofchannels,nb_chunks[0]*sizeofchunk,nb_chunks[1]*sizeofchunk,nb_chunks[2]*sizeofchunk]), dtype='float32')

    pad_image[:,:sizeofimage[0], :sizeofimage[1], :sizeofimage[2]] = image

            

    width = int(np.ceil((sizeofchunk_expand-sizeofchunk)/2.0))

            

    size_pad_im = np.shape(pad_image)[1:4]

    size_expand_im = np.array(size_pad_im) + 2 * width

    expand_image = np.zeros(([numofchannels,size_expand_im[0],size_expand_im[1],size_expand_im[2]]), dtype='float32')

    expand_image[:,width:-width, width:-width, width:-width] = pad_image

  

            

    batchsize = np.prod(nb_chunks)

    idx_chunk = 0

    chunk_batch = np.zeros((batchsize,numofchannels,sizeofchunk_expand,sizeofchunk_expand,sizeofchunk_expand),dtype='float32')

    idx_xyz = np.zeros((batchsize,3),dtype='uint16')

    for x_idx in range(nb_chunks[0]):

        for y_idx in range(nb_chunks[1]):

            for z_idx in range(nb_chunks[2]):

                

                idx_xyz[idx_chunk,:] = [x_idx,y_idx,z_idx]

                        

                         



                chunk_batch[idx_chunk,:,...] = expand_image[:,x_idx*sizeofchunk:x_idx*sizeofchunk+sizeofchunk_expand,\

                           y_idx*sizeofchunk:y_idx*sizeofchunk+sizeofchunk_expand,\

                           z_idx*sizeofchunk:z_idx*sizeofchunk+sizeofchunk_expand]             

                        

                idx_chunk += 1

    

    return chunk_batch, nb_chunks, idx_xyz, sizeofimage    



def Chunks_Image(segment_chunks, nb_chunks, sizeofchunk, sizeofchunk_expand, idx_xyz, sizeofimage):

    

    batchsize = np.size(segment_chunks,0)



    segment_image = np.zeros((nb_chunks[0]*sizeofchunk,nb_chunks[1]*sizeofchunk,nb_chunks[2]*sizeofchunk))

    

    for idx_chunk in range(batchsize):

        

        idx_low = idx_xyz[idx_chunk,:] * sizeofchunk

        idx_upp = (idx_xyz[idx_chunk,:]+1) * sizeofchunk

        

        segment_image[idx_low[0]:idx_upp[0],idx_low[1]:idx_upp[1],idx_low[2]:idx_upp[2]] = \

        segment_chunks[idx_chunk,0,...]

        



    segment_image = segment_image[:sizeofimage[0], :sizeofimage[1], :sizeofimage[2]]

    return segment_image







def BreastSeg(image,scale_subject,model,opt):

    modelpath = "Models/"       

    modelname = modelpath+"model_breast.pth"  

    checkpoint = torch.load(modelname)

    model.load_state_dict(checkpoint)





    numofseg =1



    commonspacing = [1.5,1.5,1.5]



    imageshape = image.shape



    scale_subject = scale_subject[::-1]/commonspacing    

    image = nd.interpolation.zoom(image,scale_subject,order=1)    

    imagesize = image.shape

    

    sizeofchunk = 20

    sizeofchunk_expand = 108    

    if opt.cuda:

        sizeofchunk = 132

        sizeofchunk_expand = 220            

    image_one = np.zeros((1,imagesize[0],imagesize[1],imagesize[2]),dtype='float32')

    image_one[0,...] =image 

    chunk_batch, nb_chunks, idx_xyz, sizeofimage = Generator_multichannels(image_one,sizeofchunk,sizeofchunk_expand,1)



    seg_batch = np.zeros((np.size(chunk_batch,0),numofseg,sizeofchunk,sizeofchunk,sizeofchunk),dtype='float32')

    for i_chunk in range(np.size(chunk_batch,0)):

        input = Variable(torch.from_numpy(chunk_batch[i_chunk:i_chunk+1,...]),volatile=True)

        model.eval()

        if opt.cuda:

            input = input.cuda()

        prediction = model(input)   

        

        seg_batch[i_chunk,0,...] = (prediction.data).cpu().numpy()

        

    

    for i_seg in range(numofseg):

        prob_image = Chunks_Image(seg_batch[:,i_seg:i_seg+1,...], nb_chunks, sizeofchunk, sizeofchunk_expand, idx_xyz, sizeofimage)

        up_image = nd.interpolation.zoom(prob_image,1/scale_subject,order=1)

        up_image_norm = np.zeros(imageshape,dtype='float32')

        temp_image = up_image[0:imageshape[0],0:imageshape[1],0:imageshape[2]]

        shape_tempimage =np.shape(temp_image)

        up_image_norm[0:shape_tempimage[0],0:shape_tempimage[1],0:shape_tempimage[2]] = temp_image

        

        threshold = 0.5

        idx = up_image_norm > threshold

        up_image_norm[idx] = 1

        up_image_norm[~idx] = 0               

        seg_img = up_image_norm.astype('uint8')

    return seg_img









def BreastTumor(image_sub,image_post,image_mask,scale_subject,model1st,model2nd,opt):

    modelpath = "Models/"       

    modelname = modelpath+"/model_tumor_1st.pth" 

    checkpoint = torch.load(modelname)

    model1st.load_state_dict(checkpoint)





    numofseg =1



    commonspacing = [0.7,0.7,0.7]

    imageshape = image_sub.shape





    scale_subject = scale_subject[::-1]/commonspacing

    

    

    image_sub = nd.interpolation.zoom(image_sub,scale_subject,order=1)

    image_post = nd.interpolation.zoom(image_post,scale_subject,order=1)

    image_mask = nd.interpolation.zoom(image_mask,scale_subject,order=1)

    

    imagesize = np.shape(image_sub)

    

    image_one = np.zeros((3,imagesize[0],imagesize[1],imagesize[2]),dtype='float32')

    

    image_one[0,...] = 1.0*image_sub



    image_one[1,...] = 1.0*image_post

    image_one[2,...] = 1.0*image_mask







    

    sizeofchunk = 12

    sizeofchunk_expand = 52    

    if opt.cuda:

#        sizeofchunk = 148

#        sizeofchunk_expand = 188      

        sizeofchunk = 108

        sizeofchunk_expand = 148  

               



    chunk_batch, nb_chunks, idx_xyz, sizeofimage = Generator_multichannels(image_one,sizeofchunk,sizeofchunk_expand,3)



    seg_batch = np.zeros((np.size(chunk_batch,0),numofseg,sizeofchunk,sizeofchunk,sizeofchunk),dtype='float32')

    for i_chunk in range(np.size(chunk_batch,0)):

        input = Variable(torch.from_numpy(chunk_batch[i_chunk:i_chunk+1,...]),volatile=True)

        model1st.eval()

        if opt.cuda:

            input = input.cuda()

        prediction = model1st(input)   

        

        seg_batch[i_chunk,0,...] = (prediction.data).cpu().numpy()

        

    

    for i_seg in range(numofseg):

        prob_image = Chunks_Image(seg_batch[:,i_seg:i_seg+1,...], nb_chunks, sizeofchunk, sizeofchunk_expand, idx_xyz, sizeofimage)

        prob_image[prob_image<0.01] =0



    image_one[2,...] = prob_image

# Just for saving output of 1st stage             

    up_image = nd.interpolation.zoom(prob_image,1/scale_subject,order=1)

    up_image_norm = np.zeros(imageshape,dtype='float32')

    temp_image = up_image[0:imageshape[0],0:imageshape[1],0:imageshape[2]]

    shape_tempimage =np.shape(temp_image)

    up_image_norm[0:shape_tempimage[0],0:shape_tempimage[1],0:shape_tempimage[2]] = temp_image             

    prob_output =  up_image_norm

#            

             

             

    del model1st           

    modelname = modelpath+"/model_tumor_2nd.pth"  

    checkpoint = torch.load(modelname)

    model2nd.load_state_dict(checkpoint)             

             

             

    chunk_batch, nb_chunks, idx_xyz, sizeofimage = Generator_multichannels(image_one,sizeofchunk,sizeofchunk_expand,3)



    seg_batch = np.zeros((np.size(chunk_batch,0),numofseg,sizeofchunk,sizeofchunk,sizeofchunk),dtype='float32')

    for i_chunk in range(np.size(chunk_batch,0)):

        input = Variable(torch.from_numpy(chunk_batch[i_chunk:i_chunk+1,...]),volatile=True)

        model2nd.eval()

        if opt.cuda:

            input = input.cuda()

        prediction = model2nd(input)   

        

        seg_batch[i_chunk,0,...] = (prediction.data).cpu().numpy()

        

    

    for i_seg in range(numofseg):

        prob_image = Chunks_Image(seg_batch[:,i_seg:i_seg+1,...], nb_chunks, sizeofchunk, sizeofchunk_expand, idx_xyz, sizeofimage)

        up_image = nd.interpolation.zoom(prob_image,1/scale_subject,order=1)

        up_image_norm = np.zeros(imageshape,dtype='float32')

        temp_image = up_image[0:imageshape[0],0:imageshape[1],0:imageshape[2]]

        shape_tempimage =np.shape(temp_image)

        up_image_norm[0:shape_tempimage[0],0:shape_tempimage[1],0:shape_tempimage[2]] = temp_image

        threshold = 0.5

        idx = up_image_norm > threshold

        up_image_norm[idx] = 1

        up_image_norm[~idx] = 0               

        seg_img = up_image_norm.astype('uint8')           

    return prob_output,seg_img