import os from os import listdir from os.path import join, exists, isdir import nibabel as nib import numpy as np import SimpleITK import torch import logging import surface_distance as surfdist from Framework.dcm2nii import convert_rtss_to_nii_and_resample as dcm2nii from Framework.utils import group_dict_by_value from Framework.dcm_info_reader import get_pred_name_map logger = logging.getLogger(__name__) def readnii2numpy(pre_path, gt_path): pre = np.array(nib.load(pre_path).dataobj) gt = np.array(nib.load(gt_path).dataobj) return pre, gt # @profile def one_hot(label, class_num): label = label.cpu().numpy() # print(f'class_num, {class_num}') print(f'label shape, {label.shape}') print(label.max(), label.min()) print(np.unique(label)) #one_hot = np.zeros((label.shape[0], class_num), dtype=label.dtype) #one_hot[np.arange(label.size), label] = 1 # label = label > 0 label[
高效版本:计算图像分割结果评测的DICE打分,MSD,Haus Distance
最新推荐文章于 2024-07-11 14:54:41 发布
该博客介绍了如何使用Python进行深度学习图像分割的评估,具体涉及DICE相似系数、平均表面距离(MSD)和95% Hausdorff距离的计算。通过加载nifti格式的图像,将像素标签转换为one-hot编码,然后计算不同类别的 Dice 指数、表面距离等指标,以衡量预测分割结果与实际分割的准确性。
摘要由CSDN通过智能技术生成