def scorer(pred,label,vxlspacing):
"""
:param pred:
:param label:
:param voxelspacing:
:return:
"""
volscores = {}
volscores['dice'] = metric.dc(pred,label)
volscores['jaccard'] = metric.binary.jc(pred,label)
volscores['voe'] = 1. - volscores['jaccard']
volscores['rvd'] = metric.ravd(label,pred)
if np.count_nonzero(pred) ==0 or np.count_nonzero(label)==0:
volscores['assd'] = 0
volscores['msd'] = 0
else:
evalsurf = Surface(pred,label,physical_voxel_spacing = vxlspacing,mask_offset = [0.,0.,0.], reference_offset = [0.,0.,0.])
volscores['assd'] = evalsurf.get_average_symmetric_surface_distance()
volscores['msd'] = metric.hd(label,pred,voxelspacing=vxlspacing)
logging.info("\tDice " + str(volscores['dice']))
logging.info("\tJaccard " + str(volscores['jaccard']))
logging.info("\tVOE " + str(volscores['voe']))
logging.info("\tRVD " + str(volscores['rvd']))
logging.info("\tASSD " + str(volscores['assd']))
logging.info("\tMSD " + str(volscores['msd']))
return volscores
def get_assd(preds_path, labels_path, dicoms_path, result_path, ):
test_time = time.strftime("%Y.%m.%d.%H.%M.%S", time.localtime())
result = os.path.join(result_path, "results.txt")
assd = {}
all_assd = 0
# 获取病人姓名
patients = os.listdir(preds_path)
for patient in patients:
# 获取标签集合和预测集合
pred_path = os.path.join(preds_path, patient)
label_path = os.path.join(labels_path, patient)
dicom_path = os.path.join(dicoms_path, patient)
preds = sorted(glob.glob(os.path.join(pred_path, "*.bmp")))
labels = sorted(glob.glob(os.path.join(label_path, "*.bmp")))
# 创建array变量，存储每个病人图片数据
[width, height] = imageio.imread(preds[0]).shape
pred_rawData = np.zeros((width, height, len(preds)))
label_rawData = np.zeros((width, height, len(labels)))
i = 0
for pred, label in zip(preds, labels):
# 读取图片，获取numpy array
pred_array = imageio.imread(pred)
label_array = imageio.imread(label)
# 进行二值化处理
pred_rawData[:, :, i] = pred_array / 255
label_rawData[:, :, i] = label_array / 255
# 循环将每张图片数据进行转换并保存
i += 1
# 对array进行数据格式转换
pred_rawData = pred_rawData.astype(int)
label_rawData = label_rawData.astype(int)
# print("*"*30)
print(dicom_path)
pixel_info = get_pixel_info(dicom_path)
print(patient, "_像素间距:", pixel_info)
# print("*"*30)
evalsurf = Surface(pred_rawData, label_rawData, physical_voxel_spacing=pixel_info, mask_offset=[0., 0., 0.],
reference_offset=[0., 0., 0.])
assd[patient] = evalsurf.get_average_symmetric_surface_distance()
per_result = "'%s'_the average symmetric surface distance：%.2f" % (patient, assd[patient])
print(per_result)
with open(result, mode='a') as file:
file.write(test_time)
file.write(" ")
file.write(per_result)
file.write("\n")
for key in assd.keys():
all_assd += assd.get(key)
average_assd = all_assd / len(assd)
average_result = "AllPatients_the average symmetric surface distance：%.2f" % (average_assd)
print(average_result)
with open(result, mode='a') as file:
file.write(test_time)
file.write(" ")
file.write(average_result)
file.write("\n")
return assd
def compute_segmentation_scores(prediction_mask, reference_mask,
voxel_spacing):
"""
Calculates metrics scores from numpy arrays and returns an dict.
Assumes that each object in the input mask has an integer label that
defines object correspondence between prediction_mask and
reference_mask.
:param prediction_mask: numpy.array, int
:param reference_mask: numpy.array, int
:param voxel_spacing: list with x,y and z spacing
:return: dict with dice, jaccard, voe, rvd, assd, rmsd, and msd
"""
scores = {
'dice': [],
'jaccard': [],
'voe': [],
'rvd': [],
'assd': [],
'rmsd': [],
'msd': []
}
p = (prediction_mask > 0)
r = (reference_mask > 0)
if np.any(p) and np.any(r):
dice = metric.dc(p, r)
jaccard = dice / (2. - dice)
scores['dice'].append(dice)
scores['jaccard'].append(jaccard)
scores['voe'].append(1. - jaccard)
scores['rvd'].append(metric.ravd(r, p))
evalsurf = Surface(p,
r,
physical_voxel_spacing=voxel_spacing,
mask_offset=[0., 0., 0.],
reference_offset=[0., 0., 0.])
assd = evalsurf.get_average_symmetric_surface_distance()
rmsd = evalsurf.get_root_mean_square_symmetric_surface_distance()
msd = evalsurf.get_maximum_symmetric_surface_distance()
scores['assd'].append(assd)
scores['rmsd'].append(rmsd)
scores['msd'].append(msd)
else:
# There are no objects in the prediction, in the reference, or both
scores['dice'].append(0)
scores['jaccard'].append(0)
scores['voe'].append(1.)
# Surface distance (and volume difference) metrics between the two
# masks are meaningless when any one of the masks is empty. Assign
# maximum penalty. The average score for these metrics, over all
# objects, will thus also not be finite as it also loses meaning.
scores['rvd'].append(LARGE)
scores['assd'].append(LARGE)
scores['rmsd'].append(LARGE)
scores['msd'].append(LARGE)
return scores
def get_scores(pred,label,vxlspacing):
volscores = {}
volscores['dice'] = metric.dc(pred,label)
volscores['jaccard'] = metric.binary.jc(pred,label)
volscores['voe'] = 1. - volscores['jaccard']
volscores['rvd'] = metric.ravd(label,pred)
if np.count_nonzero(pred) ==0 or np.count_nonzero(label)==0:
volscores['assd'] = 0
volscores['msd'] = 0
else:
evalsurf = Surface(pred,label,physical_voxel_spacing = vxlspacing,mask_offset = [0.,0.,0.], reference_offset = [0.,0.,0.])
volscores['assd'] = evalsurf.get_average_symmetric_surface_distance()
volscores['msd'] = metric.hd(label,pred,voxelspacing=vxlspacing)
return volscores
