sitk医疗图像分割评价指标Dice，hausdorff distance

医疗图像分割评价指标主要有三个，包括掩码相似度指标Dice，Jaccard，Overlap和边界距离指标hausdorff distance，平均hausdorff distance。掩码相似度指标很好理解，hausdorff distance指的是两个掩码边界像素的最近距离集合中的最大值。

sitk代码很简单，可以计算二维和三维的指标。

gt = sitk.GetImageFromArray(gt, isVector=False)
my_mask = sitk.GetImageFromArray(my_mask, isVector=False)
hausdorffcomputer = sitk.HausdorffDistanceImageFilter()
hausdorffcomputer.Execute(gt>0.5, my_mask>0.5)
AvgHD = hausdorffcomputer.GetAverageHausdorffDistance()
HD = hausdorffcomputer.GetHausdorffDistance()

dice_dist = sitk.LabelOverlapMeasuresImageFilter()
dice_dist.Execute(gt>0.5, my_mask>0.5)
dice = dice_dist.GetDiceCoefficient()