GuidedMorph/edge.py at main · mazurowski-lab/GuidedMorph · 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
import os
import torch
import numpy as np
import cv2
import SimpleITK as sitk
import matplotlib.pyplot as plt
from scipy.ndimage import gaussian_filter

def to_numpy(tensor):
    return tensor.detach().cpu().numpy() if tensor.requires_grad else tensor.cpu().numpy()

def canny_edge_detector(image, low_threshold=20, high_threshold=100, kernel_size=7, binary=False, binary_thresh=50):
    blurred_image = cv2.GaussianBlur(image, (kernel_size, kernel_size), 0)

    magnitude = cv2.Canny(blurred_image, low_threshold, high_threshold)

    magnitude = normalize_mri_percentile_to_255(magnitude)

    if binary:
        magnitude[magnitude > binary_thresh] = 255
        magnitude[magnitude != 255] = 0

    return magnitude

def normalize_mri_percentile_to_255(mri_data, lower_percentile=0, upper_percentile=100):
    lower_bound = np.percentile(mri_data, lower_percentile)
    upper_bound = np.percentile(mri_data, upper_percentile)
    mri_data_normalized = (mri_data - lower_bound) / (upper_bound - lower_bound)
    mri_data_scaled = mri_data_normalized * 255
    mri_data_scaled = np.round(mri_data_scaled).astype(np.uint8)

    return mri_data_scaled

def process_mri_volume(volume, **kwargs):
    processed_slices = []
    for slice_ in volume:
        processed_slice = canny_edge_detector(slice_, **kwargs)
        processed_slices.append(processed_slice)
    return np.array(processed_slices)

def extract_feature(numpy_array, seg_numpy_array, mode="tissue"):
    if isinstance(numpy_array, torch.Tensor):
        numpy_array = numpy_array.to('cpu').detach().numpy()
    if isinstance(seg_numpy_array, torch.Tensor):
        seg_numpy_array = seg_numpy_array.to('cpu').detach().numpy()
    if numpy_array.shape != seg_numpy_array.shape:
        raise ValueError("numpy_array and seg_numpy_array must have the same shape")

    mask = seg_numpy_array == 0

    if mode == "breast":
        enhancement_factor = 10
        numpy_array[mask] = (numpy_array[mask] * enhancement_factor).astype('float64')

    elif mode == "tissue":
        seg_numpy_array = cv2.dilate(seg_numpy_array, np.ones((10, 10), np.uint8), iterations=1)
        mask = seg_numpy_array == 0
        numpy_array[mask] = 0

    else:
        raise ValueError("mode must be 'tissue' or 'breast'")
    return numpy_array

def extract_feature_ssim(numpy_array, seg_numpy_array, mode="tissue"):
    if isinstance(numpy_array, torch.Tensor):
        numpy_array = numpy_array.to('cpu').detach().numpy()
    if isinstance(seg_numpy_array, torch.Tensor):
        seg_numpy_array = seg_numpy_array.to('cpu').detach().numpy()
    if numpy_array.shape != seg_numpy_array.shape:
        raise ValueError("numpy_array and seg_numpy_array must have the same shape")

    mask = seg_numpy_array == 0

    if mode == "breast":
        enhancement_factor = 10
        numpy_array[mask] = (numpy_array[mask] * enhancement_factor).astype('float64')

    elif mode == "tissue":
        mask = seg_numpy_array == 0
        numpy_array[mask] = 0

    else:
        raise ValueError("mode must be 'tissue' or 'breast'")
    return numpy_array

def get_edge_ssim(tensor_array, tensor_array_seg):
    if not isinstance(tensor_array, torch.Tensor):
        raise ValueError("Input must be a torch.Tensor")

    if not isinstance(tensor_array_seg, torch.Tensor):
        tensor_array_seg = torch.from_numpy(tensor_array_seg)

    batch, channel, depth, width, height = tensor_array.shape
    edges_array = torch.zeros_like(tensor_array)

    if tensor_array.dim() == 3:
        tensor_array = tensor_array.unsqueeze(0).unsqueeze(0)
    if tensor_array_seg.dim() == 3:
        tensor_array_seg = tensor_array_seg.unsqueeze(0).unsqueeze(0)

    for b in range(batch):
        for c in range(channel):
            slice_ = tensor_array[b, c, :, :, :]
            slice_seg_ = tensor_array_seg[b, c, :, :, :]
            slice_ = extract_feature_ssim(slice_, slice_seg_, mode = "tissue")
            slice_ = normalize_mri_percentile_to_255(slice_)
            edges_array[b, c, :, :, :] =  torch.from_numpy(slice_).to(torch.uint8)

    return edges_array

def get_edge(tensor_array, tensor_array_seg):
    if not isinstance(tensor_array, torch.Tensor):
        raise ValueError("Input must be a torch.Tensor")

    if not isinstance(tensor_array_seg, torch.Tensor):
        tensor_array_seg = torch.from_numpy(tensor_array_seg)

    batch, channel, depth, width, height = tensor_array.shape
    edges_array = torch.zeros_like(tensor_array)

    if tensor_array.dim() == 3:
        tensor_array = tensor_array.unsqueeze(0).unsqueeze(0)
    if tensor_array_seg.dim() == 3:
        tensor_array_seg = tensor_array_seg.unsqueeze(0).unsqueeze(0)

    for b in range(batch):
        for c in range(channel):
            slice_ = tensor_array[b, c, :, :, :]
            slice_seg_ = tensor_array_seg[b, c, :, :, :]
            slice_ = extract_feature(slice_, slice_seg_, mode = "tissue")
            slice_ = normalize_mri_percentile_to_255(slice_)
            edges_array[b, c, :, :, :] =  torch.from_numpy(slice_).to(torch.uint8)

    return edges_array

def apply_gaussian_filter(volume, sigma=1):
    volume_np = volume.detach().cpu().numpy()

    for c in range(volume_np.shape[1]):
        for z in range(volume_np.shape[2]):
            volume_np[0, c, z, :, :] = gaussian_filter(volume_np[0, c, z, :, :], sigma=sigma)

    smoothed_volume = torch.from_numpy(volume_np).to(volume.device)
    return smoothed_volume