-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathevaluate.py
217 lines (178 loc) · 8.33 KB
/
evaluate.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
import torch
import os
from torch.utils.data import DataLoader
from dataset import DatasetImageMaskContourDist, mean_and_std
import glob
from models import UNet, UNet_DCAN, UNet_DMTN, PsiNet, UNet_ConvMCD
from tqdm import tqdm
import numpy as np
import cv2
from utils import create_validation_arg_parser
import scipy.io as scio
from utils import AverageMeter
import segmentation_models_pytorch as smp
from sklearn.metrics import cohen_kappa_score, accuracy_score, confusion_matrix, recall_score, f1_score, classification_report, jaccard_score
from train_seg_clf import CotrainingModelMulti
import pandas as pd
from scipy.special import softmax
import surface_distance
import scipy.spatial
from numpy import mean
def getDSC(testImage, resultImage):
"""Compute the Dice Similarity Coefficient."""
testArray = testImage.flatten()
resultArray = resultImage.flatten()
return 1.0 - scipy.spatial.distance.dice(testArray, resultArray)
def getJaccard(testImage, resultImage):
"""Compute the Dice Similarity Coefficient."""
testArray = testImage.flatten()
resultArray = resultImage.flatten()
return 1.0 - scipy.spatial.distance.jaccard(testArray, resultArray)
def getPrecisionAndRecall(testImage, resultImage):
testArray = testImage.flatten()
resultArray = resultImage.flatten()
TP = np.sum(testArray*resultArray)
FP = np.sum((1-testArray)*resultArray)
FN = np.sum(testArray*(1-resultArray))
precision = TP/(TP+FP)
recall = TP/(TP+FN)
return precision, recall
def build_model(model_type):
if model_type == "unet":
model = UNet(num_classes=2)
if model_type == "dcan":
model = UNet_DCAN(num_classes=2)
if model_type == "dmtn":
model = UNet_DMTN(num_classes=2)
if model_type == "psinet":
model = PsiNet(num_classes=2)
if model_type == "convmcd":
model = UNet_ConvMCD(num_classes=2)
return model
if __name__ == "__main__":
args = create_validation_arg_parser().parse_args()
test_path = os.path.join(args.test_path, "*.png")
model_file = args.model_file
save_path = args.save_path
model_type = args.model_type
distance_type = args.distance_type
cuda_no = args.cuda_no
CUDA_SELECT = "cuda:{}".format(cuda_no)
device = torch.device(CUDA_SELECT if torch.cuda.is_available() else "cpu")
train_file_names = glob.glob(os.path.join(args.train_path, "*.png"))
train_mean, train_std = mean_and_std(train_file_names)
test_file_names = glob.glob(test_path)
test_dataset = DatasetImageMaskContourDist(test_file_names, distance_type, train_mean, train_std, args.clahe)
testLoader = DataLoader(test_dataset, batch_size=4, num_workers=4, shuffle=True)
if not os.path.exists(save_path):
os.mkdir(save_path)
clf_accs = AverageMeter("Acc", ".8f")
clf_kappas = AverageMeter("Kappa", ".8f")
encoder = args.encoder
attention_type = args.attention
if args.pretrain in ['imagenet', 'ssl', 'swsl', 'instagram']:
pretrain = args.pretrain
else:
pretrain = None
usenorm = args.usenorm
print("clahe:", args.clahe)
model = CotrainingModelMulti(encoder, pretrain, usenorm, attention_type, args.classnum).to(device)
model.load_state_dict(torch.load(model_file))
model.eval()
name = []
prob = []
label = []
pred = []
dice_1o = []
dice_2o = []
jaccard_1o = []
jaccard_2o = []
HD_o = []
ASSD_o = []
for i, (img_file_name, inputs, targets1, targets2, targets3, targets4) in enumerate(tqdm(testLoader)):
inputs = inputs.to(device)
seg_labels = targets1.numpy()
targets1, targets2 = targets1.to(device), targets2.to(device)
targets3, targets4 = targets3.to(device), targets4.to(device)
targets = [targets1, targets2, targets3, targets4]
seg_outputs = model.seg_forward(inputs)
if not isinstance(seg_outputs, list):
seg_outputs = [seg_outputs]
clf_outputs = model.clf_forward(inputs, seg_outputs[3], seg_outputs[4], seg_outputs[5])
outputs1 = seg_outputs[0].detach().cpu().numpy().squeeze()
outputs2 = seg_outputs[1].detach().cpu().numpy().squeeze()
outputs3 = seg_outputs[2].detach().cpu().numpy().squeeze()
seg_preds = np.round(outputs1)
clf_labels = torch.argmax(targets[3], dim=2).squeeze(1).detach().cpu().item()
clf_preds = torch.argmax(clf_outputs, dim=1).detach().cpu().numpy().item()
dsc_loss = smp.utils.losses.DiceLoss()
jac_loss = smp.utils.losses.JaccardLoss()
seg_prs = seg_preds
dice_1 = f1_score(seg_labels.squeeze(), seg_prs, average='micro')
dice_2 = getDSC(seg_labels, seg_prs)
jaccard_1 = jaccard_score(seg_labels.squeeze(), seg_prs, average='micro')
jaccard_2 = getJaccard(seg_labels, seg_prs)
label_seg = np.array(seg_labels.squeeze(), dtype=bool)
predict = np.array(seg_preds, dtype=bool)
surface_distances = surface_distance.compute_surface_distances(label_seg, predict, spacing_mm=(1, 1))
HD = surface_distance.compute_robust_hausdorff(surface_distances, 95)
distances_gt_to_pred = surface_distances["distances_gt_to_pred"]
distances_pred_to_gt = surface_distances["distances_pred_to_gt"]
surfel_areas_gt = surface_distances["surfel_areas_gt"]
surfel_areas_pred = surface_distances["surfel_areas_pred"]
ASSD = (np.sum(distances_pred_to_gt * surfel_areas_pred) + np.sum(distances_gt_to_pred * surfel_areas_gt))/(np.sum(surfel_areas_gt)+np.sum(surfel_areas_pred))
output_path_m = os.path.join(
save_path, "m_" + os.path.basename(img_file_name[0])
)
output_path_d = os.path.join(
save_path, "d_" + os.path.basename(img_file_name[0])
)
output_path_dmat = os.path.join(
save_path, "d_" + os.path.basename(img_file_name[0]).replace('.png', '.mat')
)
output_path_p = os.path.join(
save_path, os.path.basename(img_file_name[0])
)
output_path_b = os.path.join(
save_path, "b_" + os.path.basename(img_file_name[0])
)
output_path_bmat = os.path.join(
save_path, "b_" + os.path.basename(img_file_name[0]).replace('.png', '.mat')
)
cv2.imwrite(output_path_p, (outputs1*255.))
cv2.imwrite(output_path_m, (seg_preds*255.))
cv2.imwrite(output_path_b, (outputs2*255.))
cv2.imwrite(output_path_d, (outputs3*255.))
scio.savemat(output_path_bmat, {'boundary': outputs2})
scio.savemat(output_path_dmat, {'dist': outputs3})
name.append(os.path.basename(img_file_name[0]))
prob.append(softmax(clf_outputs.detach().cpu().numpy().squeeze()))
label.append(clf_labels)
pred.append(clf_preds)
dice_1o.append(dice_1)
dice_2o.append(dice_2)
jaccard_1o.append(jaccard_1)
jaccard_2o.append(jaccard_2)
HD_o.append(HD)
ASSD_o.append(ASSD)
kappa = cohen_kappa_score(label, pred)
acc = accuracy_score(label, pred)
recall = recall_score(label, pred, average='micro')
f1 = f1_score(label, pred, average='weighted')
c_matrix = confusion_matrix(label, pred)
if args.classnum == 3:
target_names = ['N', 'D', 'M']
clas_report = classification_report(label, pred, target_names=target_names, digits=5)
elif args.classnum == 2:
target_names = ['N', 'D']
clas_report = classification_report(label, pred, target_names=target_names, digits=5)
name_flag = args.val_path[11:12].replace('/', '_')
print(name_flag)
dataframe = pd.DataFrame({'case': name, 'prob': prob, 'label': label, 'pred': pred, 'dice1': dice_1o, 'dice2': dice_2o, 'jaccard1': jaccard_1o, 'jaccard2': jaccard_2o, 'HD': HD_o, 'ASSD': ASSD_o})
dataframe.to_csv(save_path + "/" + name_flag + "_class&seg.csv", index=False, sep=',')
resultframe = pd.DataFrame({'acc': acc, 'kappa': kappa, 'recall': recall, 'f1score': f1, 'seg_dice1': mean(dice_1o), 'seg_dice2': mean(dice_2o), 'jaccard1': mean(jaccard_1o), 'jaccard2': mean(jaccard_2o), 'HD': mean(HD_o), 'ASSD': mean(ASSD_o)}, index=[1])
resultframe.to_csv(save_path + "/" + name_flag + "_acc_kappa.csv", index=0)
with open(save_path + "/" + name_flag + "_cmatrix.txt", "w") as f:
f.write(str(c_matrix))
with open(save_path + "/" + name_flag + "_clas_report.txt", "w") as f:
f.write(str(clas_report))