-
Notifications
You must be signed in to change notification settings - Fork 0
/
pipeline - Validation.py
185 lines (156 loc) · 7.7 KB
/
pipeline - Validation.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
# -*- coding: utf-8 -*-
"""
Created on Mon Apr 1 11:10:26 2019
@author: s159890, group 4 of 8DM20, Capita Selecta prostate
"""
# God is in his heaven, all is right with the world
#pipeline - Validation.py
# Script used for generation validation segmentation of Group 4 for 8DM20, CS Prostate
# Import all necessary methods written by group
from newmethods import decisionfusing as df
from newmethods import Mutualinformation as mi
from newmethods.dice import dice
from Import_Files import Import_Files, Import_Files_string
from Transform_GT import Transform_GT
from Registration import register
# Import all packages relevant for pipeline
from scipy.stats import wilcoxon
import numpy as np
import SimpleITK as sitk
import matplotlib.pyplot as plt
import glob
import os
# CONSTANTS
TRESHOLD = 0.7
RUNREGISTRATION = True
# Define paths: these must be altered to use the script on your local machine!
FIXED_IMAGES_PATH = r'C:\Users\s159890\Documents\Q3 Jaar 1 (BME)\Capita selecta in image analysis (8DM20)\Registration\Assignment 2\DatasetValidatie'
MOVING_IMAGES_PATH = r'C:\Users\s159890\Documents\Q3 Jaar 1 (BME)\Capita selecta in image analysis (8DM20)\Registration\Assignment 2\Dataset'
ELASTIX_PATH = r'C:\Users\s159890\Documents\Q3 Jaar 1 (BME)\Capita selecta in image analysis (8DM20)\Registration\Assignment 2\Practical\Software\elastix_windows64\elastix.exe'
TRANSFORMIX_PATH = r'C:\Users\s159890\Documents\Q3 Jaar 1 (BME)\Capita selecta in image analysis (8DM20)\Registration\Assignment 2\Practical\Software\elastix_windows64\transformix.exe'
RESULT_PATH = r'D:\Leander\8DM20 Capita Selecta Image Analysis\Run 5 - Validation'
if RUNREGISTRATION:
# Register all the images with grountruth on the unseen, fixed images, return MI array (mxn)
MI = register(FIXED_IMAGES_PATH, MOVING_IMAGES_PATH, ELASTIX_PATH, RESULT_PATH)
# Perform calculated registrations on masks and write them to disk
Transform_GT(MOVING_IMAGES_PATH, RESULT_PATH, TRANSFORMIX_PATH)
else:
MI = np.load(RESULT_PATH + r'\Mutualinformation.npy')
# Get images for comparisons during atlas filtering
images = Import_Files(FIXED_IMAGES_PATH, files_to_import = 'images')
# Collect registered images & masks
images_registered = []
masks_registered = []
im_reg_paths = glob.glob(RESULT_PATH + '/p*')
for i, path in zip(range(len(im_reg_paths)), im_reg_paths):
image_registered, mask_registered = Import_Files(path)
# Add dummy images to image/mask_registered to compensate for lack of diagonal
# to ease with indexing
image_registered = np.insert(image_registered, i, np.zeros(image_registered.shape[1:]), axis = 0)
mask_registered = np.insert(mask_registered, i, np.zeros(mask_registered.shape[1:]), axis = 0)
images_registered.append(image_registered)
masks_registered.append(mask_registered)
images_registered = np.stack(images_registered) # Stack images into numpy array for easy storage
masks_registered = np.stack(masks_registered) # Idem
# Derive segmentations of prostate of images via the following four methods:
# 1. Simple majority voting decision fusing
# 2. Weighted decision fusing
# 3. Atlas filtering & method 1
# 4. Atlas filtering & method 2
# Calculate which masks survive the treshold (Atlas filtering)
survivors = mi.atlasfiltering(MI, TRESHOLD)
# Execute methods
method_1_segmentations = []
method_2_segmentations = []
method_3_segmentations = []
method_4_segmentations = []
for i, fixed_im, im_regs, im_segs in zip(range(len(image_registered)), images, images_registered, masks_registered):
# Method 3
majority_vote_filter = df.decisionfusion_majority(im_segs[survivors[i]])
method_3_segmentations.append(majority_vote_filter)
# Method 4
weighted_vote_filter = df.decisionfusing_weighted(fixed_im, im_regs[survivors[i]], im_segs[survivors[i]])
method_4_segmentations.append(weighted_vote_filter)
# Delete dummies for the methods that do not include filtering and accidentally include dummies
im_regs = np.delete(im_segs, i, axis = 0)
im_segs = np.delete(im_segs, i, axis = 0)
# Method 1
majority_vote = df.decisionfusion_majority(im_segs)
method_1_segmentations.append(majority_vote)
# Method 2
weighted_vote = df.decisionfusing_weighted(fixed_im, im_regs, im_segs)
method_2_segmentations.append(weighted_vote)
# Merge segmentations into numpy arrays that are convenient to use
method_1_segmentations = np.stack(method_1_segmentations)
method_2_segmentations = np.stack(method_2_segmentations)
method_3_segmentations = np.stack(method_3_segmentations)
method_4_segmentations = np.stack(method_4_segmentations)
# Write images to .mhd in result folder
if not os.path.exists(RESULT_PATH + r"\Results method 1"):
os.mkdir(RESULT_PATH + r"\Results method 1")
os.mkdir(RESULT_PATH + r"\Results method 2")
os.mkdir(RESULT_PATH + r"\Results method 3")
os.mkdir(RESULT_PATH + r"\Results method 4")
writer = sitk.ImageFileWriter()
for i in range(len(method_1_segmentations)):
# Multiply image segmentations by 255 to convert from bool to int (0..255)
# Method 1
image = sitk.GetImageFromArray(method_1_segmentations[i] * 255)
writer.SetFileName(RESULT_PATH + r"\Results method 1" + r"\segmentation_{}.mhd".format(i+1))
writer.Execute(image)
# Method 2
image = sitk.GetImageFromArray(method_2_segmentations[i] * 255)
writer.SetFileName(RESULT_PATH + r"\Results method 2" + r"\segmentation_{}.mhd".format(i+1))
writer.Execute(image)
# Method 3
image = sitk.GetImageFromArray(method_3_segmentations[i] * 255)
writer.SetFileName(RESULT_PATH + r"\Results method 3" + r"\segmentation_{}.mhd".format(i+1))
writer.Execute(image)
# Method 4
image = sitk.GetImageFromArray(method_4_segmentations[i] * 255)
writer.SetFileName(RESULT_PATH + r"\Results method 4" + r"\segmentation_{}.mhd".format(i+1))
writer.Execute(image)
## Performing 'evaluation' of segmentations via calculation of dice score
#dice_scores_method_1 = []
#dice_scores_method_2 = []
#dice_scores_method_3 = []
#dice_scores_method_4 = []
#for i in range(len(images)):
# gt = masks[i].flatten()
#
# dice_method_1_i = dice(gt, method_1_segmentations[i].flatten())
# dice_scores_method_1.append(dice_method_1_i)
# dice_method_2_i = dice(gt, method_2_segmentations[i].flatten())
# dice_scores_method_2.append(dice_method_2_i)
# dice_method_3_i = dice(gt, method_3_segmentations[i].flatten())
# dice_scores_method_3.append(dice_method_3_i)
# dice_method_4_i = dice(gt, method_4_segmentations[i].flatten())
# dice_scores_method_4.append(dice_method_4_i)
#
## PROBABLY BEST TO DELETE EVERYTHING UNDER THIS
#
## Making boxplots
#methods= [dice_scores_method_1, dice_scores_method_2, dice_scores_method_3, dice_scores_method_4]
#
#fig, ax = plt.subplots()
#bp= ax.boxplot(methods)
#ax.set_xticklabels(["Majority voting", "Weighted voting", "Atlas filtering, \n majority voting", "Atlas filtering,\n weighted voting"],
# rotation = 0, fontsize = 11)
#ax.set_xlabel('Methods', fontsize = 11)
#ax.set_ylabel('Dice score', fontsize = 11)
#ax.yaxis.grid(True, linestyle='-', which='major', color='lightgrey', alpha=0.5)
#ax.set_axisbelow(True)
#ax.set_ylim(-0.05, 1.0)
#plt.show()
#
## Test significance of difference between results
#T1, p1 = wilcoxon(dice_scores_method_1, dice_scores_method_2)
#T2, p2 = wilcoxon(dice_scores_method_1, dice_scores_method_3)
#T3, p3 = wilcoxon(dice_scores_method_1, dice_scores_method_4)
## Outcome: Methods 2, 3, 4 all differe significantly from 1
#
#T4, p4 = wilcoxon(dice_scores_method_2, dice_scores_method_3)
#T5, p5 = wilcoxon(dice_scores_method_2, dice_scores_method_4)
## Outcome: Methods 3 and 4 do not vary significantly from 2
#
#T6, p6 = wilcoxon(dice_scores_method_3, dice_scores_method_4)