hippocampus-segmentation/assessment.py at main · Lisa-suger/hippocampus-segmentation · 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

import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
from scipy.spatial.distance import directed_hausdorff
from medpy.metric.binary import hd, hd95, dc, jc,positive_predictive_value

def dice(pred, gt, class_label):
    # im1 = im1 == tid
    # im2 = im2 == tid
    # im1 = np.asarray(im1).astype(np.bool_)
    # im2 = np.asarray(im2).astype(np.bool_)
    # if im1.shape != im2.shape:
    #     raise ValueError("Shape mismatch: im1 and im2 must have the same shape.")
    pred_binary = (pred == class_label).astype(np.uint8)
    gt_binary = (gt == class_label).astype(np.uint8)
    assert pred_binary.shape == gt_binary.shape, "预测和真实标注的尺寸必须相同"
    # Compute Dice coefficient
    dsc = dc(pred_binary, gt_binary)
    return dsc

def jaccard(pred, gt, class_label):
    # im1 = im1 == tid
    # im2 = im2 == tid
    # im1 = np.asarray(im1, dtype=np.bool_)
    # im2 = np.asarray(im2, dtype=np.bool_)
    #
    # if im1.shape != im2.shape:
    #     raise ValueError("Shape mismatch: im1 and im2 must have the same shape.")

    pred_binary = (pred == class_label).astype(np.uint8)
    gt_binary = (gt == class_label).astype(np.uint8)
    assert pred_binary.shape == gt_binary.shape, "预测和真实标注的尺寸必须相同"

    jac = jc(pred_binary, gt_binary)

    return jac

def calculate_ppv(pred, gt, class_label):
    # 将预测和真实标注二值化，只保留当前类别的点
    pred_binary = (pred == class_label).astype(np.uint8)
    gt_binary = (gt == class_label).astype(np.uint8)
    assert pred_binary.shape == gt_binary.shape, "预测和真实标注的尺寸必须相同"
    ppv = positive_predictive_value(pred_binary, gt_binary)

    return ppv


def calculate_hd95(pred, gt, class_label):
    # 计算豪斯多夫距离
    pred_binary = (pred == class_label).astype(np.uint8)
    gt_binary = (gt == class_label).astype(np.uint8)
    assert pred_binary.shape == gt_binary.shape, "预测和真实标注的尺寸必须相同"

    #hausdorff_distance = hd(pred, gt)
    hausdorff_distance95 = hd95(pred_binary,gt_binary)
    return hausdorff_distance95