DiffuseReg/test_bootstrap.py at master · streamflowmaster/DiffuseReg · GitHub

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
import argparse
import os
import torch
from timeit import default_timer
from model.diffusion_3D.unet import RecursiveCascadeNetwork, SpatialTransform
from model.diffusion_3D.loss import loss_RCN

import core.logger as Logger
import core.metrics as Metrics
import data as Data
from math import *

from torch.optim import Adam
from torch.optim.lr_scheduler import MultiStepLR
from torch.utils.tensorboard import SummaryWriter
import time
import numpy as np
import matplotlib.pyplot as plt
import SimpleITK as sitk


def main(args):
    opt = Logger.parse(args)
    # Convert to NoneDict, which return None for missing key.
    opt = Logger.dict_to_nonedict(opt)
    writer = SummaryWriter(opt['path']["tb_logger"])

    # dataset
    phase = 'test'
    finesize = opt['model']['diffusion']['image_size']
    dataset_opt = opt['datasets']['test']
    test_set = Data.create_dataset_ACDC(dataset_opt, finesize, "test50")
    test_loader = Data.create_dataloader(test_set, dataset_opt, phase)
    print('Dataset Initialized')

    reconstruction = SpatialTransform(finesize).cuda()
    model = RecursiveCascadeNetwork(n_cascades=opt['model']['bootstrap']['n_cas'],
                                    im_size=finesize,
                                    network=opt['model']['bootstrap']['module'],
                                    stn=reconstruction).cuda()
    params_dict = torch.load(args.weights)
    # model.stems = [torch.nn.DataParallel(submodel) for submodel in model.stems]
    for i, submodel in enumerate(model.stems):
        submodel.load_state_dict(params_dict["cascade {}".format(i)])

    registDice = np.zeros((len(test_set), 3))
    originDice = np.zeros((len(test_set), 3))
    registTime = []
    print('Begin Model Evaluation.')
    model.eval()
    print(len(test_loader))
    flow_xL = []
    flow_yL = []
    flow_zL = []

    for istep, test_data in enumerate(test_loader):
        t0 = default_timer()
        with torch.no_grad():
            fixed, moving = test_data["F"].cuda(), test_data["M"].cuda()
            fixed = fixed.squeeze().unsqueeze(0).unsqueeze(0)
            moving = moving.squeeze().unsqueeze(0).unsqueeze(0)

            flows, warps, results = model(fixed, moving)
            flow = flows[-1]

        t1 = default_timer()
        origin_seg = test_data['MS'].squeeze().unsqueeze(0).unsqueeze(0).cuda()
        regist_seg = reconstruction(origin_seg.type(torch.float32), flow, mode="nearest")
        label_seg = test_data['FS'].squeeze().unsqueeze(0).unsqueeze(0).cuda()
        regist_img = reconstruction(moving.type(torch.float32), flow, mode="bilinear")

        tmp_MS = sitk.GetImageFromArray(origin_seg.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_MS, "./toy_sample/VTN_origin_seg.nii.gz")
        tmp_M = sitk.GetImageFromArray(moving.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_M, "./toy_sample/VTN_origin.nii.gz")
        tmp_WS = sitk.GetImageFromArray(regist_seg.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_WS, "./toy_sample/VTN_regist_seg.nii.gz")
        tmp_W = sitk.GetImageFromArray(regist_img.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_W, "./toy_sample/VTN_regist.nii.gz")
        tmp_FS = sitk.GetImageFromArray(label_seg.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_FS, "./toy_sample/VTN_label_seg.nii.gz")
        tmp_F = sitk.GetImageFromArray(fixed.squeeze().cpu().numpy())
        sitk.WriteImage(tmp_F, "./toy_sample/VTN_label.nii.gz")
        flow_vis = sitk.GetImageFromArray(flow.detach().squeeze().permute(1, 2, 3, 0).cpu().numpy())
        sitk.WriteImage(flow_vis, f"./toy_sample/VTN_flow.nii.gz")


        vals_regist = Metrics.dice_BraTS(regist_seg.cpu().numpy(), label_seg.cpu().numpy())[::3]
        vals_origin = Metrics.dice_BraTS(origin_seg.cpu().numpy(), label_seg.cpu().numpy())[::3]
        registDice[istep] = vals_regist
        originDice[istep] = vals_origin
        print('---- Original Dice: %03f | Deformed Dice: %03f' % (np.mean(vals_origin), np.mean(vals_regist)))

        # vals_regist = Metrics.mask_metrics(regist_seg, label_seg)
        # vals_origin = Metrics.mask_metrics(origin_seg, label_seg)
        # registDice[istep] = vals_regist.item()
        # originDice[istep] = vals_origin.item()
        # print('---- Original Dice: %03f | Deformed Dice: %03f' % (vals_origin, vals_regist))

        # vals_regist = Metrics.dice_ACDC(regist_seg.cpu().numpy(), label_seg.cpu().numpy())[::3]
        # vals_origin = Metrics.dice_ACDC(origin_seg.cpu().numpy(), label_seg.cpu().numpy())[::3]
        # registDice[istep] = vals_regist
        # originDice[istep] = vals_origin
        # print('---- Original Dice: %03f | Deformed Dice: %03f' % (np.mean(vals_origin), np.mean(vals_regist)))

        flow_np = flow.squeeze().cpu().numpy()
        flow_x = flow_np[0]
        flow_y = flow_np[1]
        flow_z = flow_np[2]
        # print(flow_x.mean(), flow_x.std(), flow_x.min(), flow_x.max())
        # print(flow_y.mean(), flow_y.std(), flow_y.min(), flow_y.max())
        # print(flow_z.mean(), flow_z.std(), flow_z.min(), flow_z.max())

        flow_xL.append(flow_x.flatten())
        flow_yL.append(flow_y.flatten())
        flow_zL.append(flow_z.flatten())

        time.sleep(1)
    omdice, osdice = np.mean(originDice), np.std(originDice)
    mdice, sdice = np.mean(registDice), np.std(registDice)
    flow_xL = np.concatenate(flow_xL, axis=None)
    flow_yL = np.concatenate(flow_yL, axis=None)
    flow_zL = np.concatenate(flow_zL, axis=None)

    print()
    print(flow_xL.mean(), flow_xL.std(), flow_xL.min(), flow_xL.max())  # 0.0014  0.4636 -14.0123 10.3754
    print(flow_yL.mean(), flow_yL.std(), flow_yL.min(), flow_yL.max())  # -0.0758 1.1375 -15.7695 16.4528
    print(flow_zL.mean(), flow_zL.std(), flow_zL.min(), flow_zL.max())  # -0.1493 1.2221 -18.4860 15.3057
    # plt.hist(flow_xL, bins=50, label="x")
    # plt.legend()
    # plt.show()
    # plt.hist(flow_yL, bins=50, label="y")
    # plt.legend()
    # plt.show()
    # plt.hist(flow_zL, bins=50, label="z")
    # plt.legend()
    # plt.show()

    print('---------------------------------------------')
    print('Total Dice and Time Metrics------------------')
    print('---------------------------------------------')
    print('origin Dice | mean = %.3f, std= %.3f' % (omdice, osdice))
    print(f'origin detailed Dice | mean = {np.mean(originDice, axis=0)}({np.std(originDice, axis=0)})')
    print('Deform Dice | mean = %.3f, std= %.3f' % (mdice, sdice))
    print(f'Deform detailed Dice | mean = {np.mean(registDice, axis=0)}({np.std(registDice, axis=0)})')


if __name__ == '__main__':
    np.random.seed(0)
    torch.manual_seed(0)
    torch.cuda.manual_seed_all(0)

    parser = argparse.ArgumentParser()

    parser.add_argument('-g', "--gpu_ids", type=str,
                        default="1")
    parser.add_argument("--strategy", type=str,
                        default="plain")
    parser.add_argument('-c', '--config', type=str,
                        default='config/test_VTN.json')
    parser.add_argument('-w', '--weights', type=str,
                        default="./experiments/.../checkpoint/E2000.pth")
    args = parser.parse_args()

    main(args)