def get_r_p_feature_name(self, sorted_number, show=False):
r_list = []
p_list = []
for feature_name in self.pd_feature.columns[1:]: # 注意不要casename
feature_value = self.pd_feature[feature_name]
new_t2_value, new_feature_value = self.remove_bad_data(
t2_value=self.t2_value, feature_value=feature_value)
r, p = pearsonr(new_t2_value, new_feature_value)
if p < 0.05:
r_list.append(r)
p_list.append(p)
r_min = sorted([abs(i) for i in r_list],
reverse=True)[sorted_number - 1]
new_r_list = []
new_p_list = []
new_feature_name = []
for feature_name in self.pd_feature.columns[1:]: # 注意不要casename
feature_value = self.pd_feature[feature_name]
new_t2_value, new_feature_value = self.remove_bad_data(
t2_value=self.t2_value, feature_value=feature_value)
r, p = pearsonr(new_t2_value, new_feature_value)
if p < 0.05 and abs(r) >= r_min: # 一定要注意不要忘记绝对值
r = round(r, 4)
p = round(p, 4)
new_r_list.append(r)
new_p_list.append(p)
new_feature_name.append(feature_name)
####################
# 显示相关系数图
####################
if show is True:
sns.regplot(
x=new_feature_value, # 特征值
y=new_t2_value, # t2值
scatter_kws={'s': 15},
)
plt.xlabel(feature_name)
plt.ylabel("T2 value")
plt.title("correlation value = " + str(round(r, 4)))
store_path = make_folder(
os.path.join(self.folder_path, self.feature_name))
plt.savefig(os.path.join(store_path,
feature_name + ".png"))
plt.clf()
# plt.show()
#####################
p_r_feature_name_result = pd.DataFrame(
data={
"feature_name": new_feature_name,
"r_value": new_r_list,
"p_value": new_p_list
})
store_path = make_folder(
os.path.join(self.folder_path, self.feature_name))
p_r_feature_name_result.to_csv(os.path.join(store_path,
"correlation_index.csv"),
index=None)
def get_dicom_by_echo():
"""获取dicom按照回波的顺序进行排序保存"""
root_path = r"Y:\DYB\2020832DATA\doctor_xie\normal_control"
for case in os.listdir(root_path):
case_path = os.path.join(root_path, case)
# 创建保存五个回波的文件夹
save_folder_name = ["new_folder_" + str(i) for i in range(1, 6)]
save_folder_path = [
os.path.join(case_path, i) for i in save_folder_name
]
# case TE 是一个case的TE值
case_TE = get_TE_in_one_case(case_path)
dicom_folder_path = os.path.join(case_path, "all_dicom")
for file in sort_file(dicom_folder_path): # 务必进行排序
dicom_path = os.path.join(dicom_folder_path, file) # 每一张dicom的路径
_, TE, _ = get_TE(dicom_path)
if TE == case_TE[0]:
shutil.copy(dicom_path, make_folder(save_folder_path[0]))
elif TE == case_TE[1]:
shutil.copy(dicom_path, make_folder(save_folder_path[1]))
elif TE == case_TE[2]:
shutil.copy(dicom_path, make_folder(save_folder_path[2]))
elif TE == case_TE[3]:
shutil.copy(dicom_path, make_folder(save_folder_path[3]))
else:
shutil.copy(dicom_path, make_folder(save_folder_path[4]))
print("case {} is finished!".format(case))
def __init__(self, src_folder_path):
self.src_folder_path = src_folder_path # 里面存放了train,test
self.train_path = os.path.join(self.src_folder_path, "train")
self.test_path = os.path.join(self.src_folder_path, "test")
self.store_trian_path = os.path.join(self.src_folder_path, "new_train")
make_folder(self.store_trian_path)
self.store_test_path = os.path.join(self.src_folder_path, "new_test")
make_folder(self.store_test_path)
def get_case(self):
for case in os.listdir(self.folder_path):
case_path = os.path.join(self.folder_path, case)
store_case_path = os.path.join(self.store_path, case)
make_folder(store_case_path)
for file in os.listdir(case_path):
if file != "SeriesRois":
file_path = os.path.join(case_path, file)
shutil.copy(file_path, store_case_path)
def move_test(self, store_name):
roi_number, data_path_list, roi_path_list = self.get_roi_number(
self.test_path)
for data, roi, count in zip(data_path_list, roi_path_list,
range(1, roi_number)):
store_data_path = os.path.join(self.store_test_path,
store_name + "_test_" + str(count))
make_folder(store_data_path)
shutil.copy(data, store_data_path)
shutil.copy(roi, store_data_path)
def get_dicom_and_roi():
"""将数据的dicom和roi放在一起"""
root_dicom_path = r"Y:\DYB\2020832DATA\doctor_xie\normal_control"
root_roi_path = r"Y:\DYB\2020832DATA\doctor_xie\Normal_data_nii_and_T2"
store_path = r"Y:\DYB\2020832DATA\doctor_xie\data"
for dicom, roi in zip(os.listdir(root_dicom_path), os.listdir(root_roi_path)):
dicom_path = os.path.join(root_dicom_path, dicom, "folder_1")
roi_path = os.path.join(root_roi_path, roi, "roi.nii")
store_case_path = make_folder(os.path.join(store_path, dicom))
store_case_dicom = make_folder(os.path.join(store_case_path, "dicom"))
shutil.copy(roi_path, store_case_path)
for file in os.listdir(dicom_path):
dicom_file_path = os.path.join(os.path.join(dicom_path, file))
shutil.copy(dicom_file_path, store_case_dicom)
print(dicom)
def __init__(self):
# 这里最后初始化的时候，就传入case_path就可以了
self.case_path = r"Y:\DYB\2020832DATA\doctor_xie\normal_control\case0"
self.nii_folder_path = os.path.join(self.case_path, "nii_folder")
self.dcm_folder_path = os.path.join(self.case_path, "all_dicom")
self.store_dicom_path = make_folder(
os.path.join(self.case_path, "store_dicom"))
def get_all_dicom():
root_path = r"Y:\DYB\2020832DATA\doctor_xie\split_data_ACLR_T2"
for case in os.listdir(root_path):
case_path = os.path.join(root_path, case)
store_all_folder = make_folder(os.path.join(case_path, "all_dicom"))
for folder in os.listdir(case_path):
folder_path = os.path.join(case_path, folder)
for file in os.listdir(folder_path):
file_path = os.path.join(folder_path, file)
shutil.copy(file_path, store_all_folder)
print(case)
def check_all_case_roi():
root_path = r"Y:\DYB\PETCT_EGFR"
store_path = r"Y:\DYB\2020832DATA\doctor_gao\EGFR_PET_CT_DATA"
for folder in os.listdir(root_path):
folder_path = os.path.join(root_path, folder)
store_folder_path = make_folder(os.path.join(store_path, folder))
for case in os.listdir(folder_path):
store_case_path = make_folder(os.path.join(store_folder_path, case))
case_path = os.path.join(folder_path, case)
ct_path = os.path.join(case_path, "ct.nii")
pet_path = os.path.join(case_path, "new_pet_resize.nii")
roi_path = os.path.join(case_path, "ct_roi.nii")
shutil.copy(ct_path, store_case_path)
shutil.copy(pet_path, store_case_path)
shutil.copy(roi_path, store_case_path)
print("case {} is finished!".format(case))
def get_2d_roi():
root_path = r"Y:\LS\liusheng _shidaoai_train_and_test"
store_path = r"Y:\LS\xxxxxxxxxxxx"
for folder in os.listdir(root_path):
folder_path = os.path.join(root_path, folder)
store_folder_path = make_folder(os.path.join(store_path, folder))
for case in os.listdir(folder_path):
case_path = os.path.join(folder_path, case)
data_path = os.path.join(case_path, "data.nii.gz")
roi_path = os.path.join(case_path, "ROI.nii.gz")
data_array = get_array_from_path(data_path)
roi_array = get_array_from_path(roi_path)
# 存放一个方向的结果
for i in range(roi_array.shape[-1]):
if np.sum(roi_array[..., i]) != 0:
# 开始存放：
# data 1 是从一个方向看的结果
data_1 = data_array[..., i-1:i+2]
roi_1 = roi_array[..., i-1:i+2]
store_case_path = make_folder(os.path.join(store_folder_path, str(i)+"_1_"+case))
np.save(os.path.join(store_case_path, "data.npy"), data_1)
np.save(os.path.join(store_case_path, "roi.npy"), roi_1)
print("---1---")
# 存放另一个方向的结果
for j in range(roi_array.shape[0]):
if np.sum(roi_array[j, ...]) != 0:
# 开始存放
# data 2 是从第二个方向看的结果
data_2 = data_array[j-1:j+2, ...]
roi_2 = roi_array[j-1:j+2, ...]
store_case_path = make_folder(os.path.join(store_folder_path, str(j) + "_2_" + case))
np.save(os.path.join(store_case_path, "data.npy"), data_2)
np.save(os.path.join(store_case_path, "roi.npy"), roi_2)
print("---2---")
# 存放第三个方向看的结果
for k in range(roi_array.shape[1]):
if np.sum(roi_array[:, k, :]) != 0:
# 开始存放第三个方向的结果 data 3
data_3 = data_array[:, k-1:k+2, :]
roi_3 = roi_array[:, k-1:k+2, :]
store_case_path = make_folder(os.path.join(store_folder_path, str(k) + "_3_" + case))
np.save(os.path.join(store_case_path, "data.npy"), data_3)
np.save(os.path.join(store_case_path, "roi.npy"), roi_3)
print("---3---")
print("case {} is finished!".format(case))
def get_nii_by_echo():
"""将每个回波的图像转为nii"""
# 正常对照组
# root_path = r"Y:\DYB\2020832DATA\doctor_xie\normal_control"
# 有病的组
root_path = r"Y:\DYB\2020832DATA\doctor_xie\split_data_ACLR_T2"
for case in os.listdir(root_path):
case_path = os.path.join(root_path, case)
# 创建保存五个回波的文件夹
save_folder_name = ["folder_" + str(i) for i in range(1, 6)]
save_folder_path = [
os.path.join(case_path, i) for i in save_folder_name
]
for folder in save_folder_path:
save_folder = make_folder(os.path.join(case_path, folder + "_nii"))
folder_path = os.path.join(case_path, folder)
ProcessSeries(folder_path, save_folder)
def __init__(self, root_path, save_folder_name):
self.root_path = root_path
self.store_path = r"Y:\DYB\2020832DATA\doctor_gao\pet_ct_egfr"
self.save_folder_path = make_folder(
os.path.join(self.store_path, save_folder_name))
def get_k_means_label(data_path, store_path, store_name):
data = pd.read_csv(data_path)
data_load = LoadData()
data_load.set_pd_data(data)
case_name, label, feature_name, array = data_load.get_element()
array = StandardScaler().fit_transform(array)
array = np.transpose(array)
k_means_model = KMeans(n_clusters=6).fit(array)
labels = k_means_model.labels_
# result = update_frame_by_data(case_name=feature_name, feature_name=case_name,
# feature_value=array, label=labels)
# result.to_csv(os.path.join(r"Y:\DYB\data_and_result\doctor tao\roi_3", "聚类.csv"),
# index=None)
label_0_feature = ["CaseName", "label"] + []
label_1_feature = ["CaseName", "label"] + []
label_2_feature = ["CaseName", "label"] + []
label_3_feature = ["CaseName", "label"] + []
label_4_feature = ["CaseName", "label"] + []
label_5_feature = ["CaseName", "label"] + []
for feature, l in zip(feature_name, labels):
if l == 0:
label_0_feature.append(feature)
elif l == 1:
label_1_feature.append(feature)
elif l == 2:
label_2_feature.append(feature)
elif l == 3:
label_3_feature.append(feature)
elif l == 4:
label_4_feature.append(feature)
elif l == 5:
label_5_feature.append(feature)
origin_feature = pd.read_csv(data_path)
group_1_feature = origin_feature[label_0_feature]
group_2_feature = origin_feature[label_1_feature]
group_3_feature = origin_feature[label_2_feature]
group_4_feature = origin_feature[label_3_feature]
group_5_feature = origin_feature[label_4_feature]
group_6_feature = origin_feature[label_5_feature]
group_1_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_1")), store_name + ".csv"),
index=None)
group_2_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_2")), store_name + ".csv"),
index=None)
group_3_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_3")), store_name + ".csv"),
index=None)
group_4_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_4")), store_name + ".csv"),
index=None)
group_5_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_5")), store_name + ".csv"),
index=None)
group_6_feature.to_csv(os.path.join(
make_folder(os.path.join(store_path, "group_6")), store_name + ".csv"),
index=None)
def __init__(self, data_folder_path, store_folder_path):
self.data_folder_path = data_folder_path
self.store_folder_path = make_folder(store_folder_path)
import os
import shutil
from Tool.FolderProcress import make_folder, remove_folder
store_path = r"E:\Data\doctor_xie\新建文件夹 (2)"
data_path = r"E:\Data\doctor_xie\新建文件夹"
for case in os.listdir(data_path):
store_case_path = os.path.join(store_path, case)
make_folder(store_case_path)
case_path = os.path.join(data_path, case)
for file in os.listdir(case_path):
if file != "data_1.nii" and file != "roi.nii.gz":
file_path = os.path.join(case_path, file)
shutil.move(file_path, store_case_path)
