def _load_data_infolder(self):
"""load training data and validation data and generate label for training data"""
print("loading...")
# train data
data1,_=read_multiNii_LC(self.patients_path, self.suffix)
data1=np.squeeze(np.array([np.array(data1).reshape(1,-1) for data1 in data1]))
data2,_=read_multiNii_LC(self.hc_path, self.suffix)
data2=np.squeeze(np.array([np.array(data2).reshape(1,-1) for data2 in data2]))
data=np.vstack([data1,data2])
# validation data
data_validation,self.name_val=read_multiNii_LC(self.val_path, self.suffix)
data_validation=np.squeeze(np.array([np.array(data_validation).reshape(1,-1) for data_validation in data_validation]))
# data in mask
mask,_=read_sigleNii_LC(self.mask)
mask=mask>=0.2
mask=np.array(mask).reshape(-1,)
self.data_train=data[:,mask]
self.data_validation=data_validation[:,mask]
# label_tr
self.label_tr=np.hstack([np.ones([len(data1),])-1,np.ones([len(data2),])])
print("loaded")
return self
def split_roi_for_one_subj(self, file_path):
"""load nii--split roi--save to """
# load nii
nii_name = file_path
nii_data, nii_object = read_sigleNii_LC(nii_name)
header = nii_object.header
affine = nii_object.affine
# split roi
uni_label = np.unique(nii_data)
uni_label = list(set(uni_label) - set([0])) # 去掉0背景
subjname = os.path.basename(file_path).split('.')[0]
# split and save to nii
for label in uni_label:
# creat folder
save_folder_name = os.path.join(self.out_path, subjname,
'ROI_' + str(label))
if not os.path.exists(save_folder_name):
os.makedirs(save_folder_name)
save_file_name = os.path.join(save_folder_name,
os.path.basename(nii_name))
if os.path.exists(save_file_name):
print('{} exist!\n'.format(save_file_name))
if self.overcopy:
print(f'overwrite!')
else:
continue
# split
roi_logic = np.array(nii_data == label, dtype=float)
roi = nii_data * roi_logic
# ndarry to nifti object
roi = nib.Nifti1Image(roi, affine=affine, header=header)
# save
save_nii(roi, save_file_name)
# -*- coding: utf-8 -*- """ Created on Wed Sep 12 19:10:02 2018 @author: lenovo """ from lc_read_nii import read_sigleNii_LC import sys sys.path.append(r'D:\myCodes\LC_MVPA\Python\MVPA_Python\utils') img_path_3d = r'D:\其他\陈逸凡\aLL_mask3D.nii' img_path_alff = r'D:\其他\陈逸凡\all_maskALFF.nii' data_3d = read_sigleNii_LC(img_path_3d) data_alff = read_sigleNii_LC(img_path_alff)
class SVCDeterministicTrAndTe():
def init(self):
# ==============================================================================
# input begin
patients_path = r'K:\XiaoweiJiang\KETI\3_新建文件夹（20190308）\DATA\REST\ALFF\ALFF_BD'
hc_path = r'K:\XiaoweiJiang\KETI\3_新建文件夹（20190308）\DATA\REST\ALFF\ALFF_HC1'
validation_path = r'K:\XiaoweiJiang\KETI\3_新建文件夹（20190308）\DATA\REST\ALFF\ALFF_HC1'
mask = r'G:\Softer_DataProcessing\spm12\spm12\tpm\Reslice3_TPM_greaterThan0.2.nii'
kfold = 5
# ==============================================================================
# mask
mask = read_sigleNii_LC(mask) >= 0.2
mask = np.array(mask).reshape(-1, )
# is_train
if_training = 1
# input end
def load_nii_and_gen_label(patients_path, hc_path, mask):
# train data
data1 = main(patients_path)
data1 = np.squeeze(
np.array([np.array(data1).reshape(1, -1) for data1 in data1]))
data2 = main(hc_path)
data2 = np.squeeze(
np.array([np.array(data2).reshape(1, -1) for data2 in data2]))
data = np.vstack([data1, data2])
# validation data
data_val = main(validation_path)
data_val = np.squeeze(
np.array(
[np.array(data_val).reshape(1, -1) for data_val in data_val]))
# data in mask
data_tr = data[:, mask]
data_val = data_val[:, mask]
# label_tr
label_tr = np.hstack(
[np.ones([
len(data1),
]) - 1, np.ones([
len(data2),
])])
return label_tr, data_tr, data_val
# training and test
def tr_te():
import lc_svc_rfe_cv_V2 as lsvc
svc = lsvc.SVCRefCv(pca_n_component=0.9,
show_results=1,
show_roc=0,
k=kfold)
if if_training:
results = svc.svc_rfe_cv(data_tr, label_tr)
return results
# run
data_tr, label_tr = datasets.make_classification(n_samples=200,
n_classes=2,
n_informative=50,
n_redundant=3,
n_features=100,
random_state=1)
label_tr, data_tr, data_val = load_nii_and_gen_label(
patients_path, hc_path, mask)
selector, weight = rfeCV(data_tr,
label_tr,
step=0.1,
cv=kfold,
n_jobs=1,
permutation=0)
results = tr_te()
results = results.__dict__
y_pred = selector.predict(x)
def read(self, nii_name):
nii_data, nii_object = read_sigleNii_LC(nii_name)
yield nii_data
)
# ==============================================================================
# input
# 外部数据
folder_p_2 = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_control\C_Weighted_selected'
folder_hc_2 = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_patient\P_Weighted_selected'
# 内部数据
folder_p_1 = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_control\C_Weighted_selected'
folder_hc_1 = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_patient\P_Weighted_selected'
# 灰质mask
mask = r'G:\Softer_DataProcessing\spm12\spm12\tpm\Reslice3_TPM_greaterThan0.2.nii'
mask = read_sigleNii_LC(mask) >= 0.2
mask = np.array(mask).reshape(-1, )
# 设置训练与否
if_training_inner_cv = 1
if_training_outer_cv = 0
if_show_data_distribution = 0 # 显示训练集和测试集数据分布
# ==============================================================================
def load_nii_and_gen_label(folder_p, folder_hc, mask):
# data
data_p = main(folder_p)
data_p = np.squeeze(
np.array([np.array(data_p).reshape(1, -1) for data_p in data_p]))
r'D:\My_Codes\LC_Machine_Learning\lc_rsfmri_tools\lc_rsfmri_tools_python\Utils'
)
sys.path.append(
r'D:\My_Codes\LC_Machine_Learning\lc_rsfmri_tools\lc_rsfmri_tools_python\Machine_learning\classfication'
)
from lc_read_nii import main
from lc_read_nii import read_sigleNii_LC
import numpy as np
# =====================================================================
patients_path = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_patient\P_Weighted_selected'
hc_path = r'D:\WorkStation_2018\Workstation_Old\WorkStation_2018-05_MVPA_insomnia_FCS\Degree\degree_gray_matter\Zdegree\Z_degree_control\C_Weighted_selected'
mask = r'G:\Softer_DataProcessing\spm12\spm12\tpm\Reslice3_TPM_greaterThan0.2.nii'
# =====================================================================
is_train = 1
mask, _ = read_sigleNii_LC(mask)
mask = mask >= 0.2
mask = np.array(mask).reshape(-1, )
def load_nii_and_gen_label(patients_path, hc_path, mask):
# data
data1, _ = main(patients_path, '.img')
data1 = np.squeeze(
np.array([np.array(data1).reshape(1, -1) for data1 in data1]))
data2, _ = main(hc_path, '.img')
data2 = np.squeeze(
np.array([np.array(data2).reshape(1, -1) for data2 in data2]))
data = np.vstack([data1, data2])