def load_data_and_label(sel):
data_name_mean = os.path.join(sel.mat_path, sel.feature_mean)
data_name_std = os.path.join(sel.mat_path, sel.feature_std)
label_name = os.path.join(sel.mat_path, sel.label)
sel.data_mean = read_mat(data_name_mean, dataset_name=None)
sel.data_std = read_mat(data_name_std, dataset_name=None)
sel.label = pd.read_excel(label_name)['诊断']
return sel
def load_onemat_and_processing(sel, i, all_mat_, len_all, s):
# load mat
mat = read_mat(all_mat_, sel.dataset_name)
# 计算方差，均数等。可扩展。(如果时静态FC，则不执行)
if sel.feature == 'staticFC':
mat_std, mat_mean = mat, []
else:
mat_std, mat_mean = sel.calc_std(mat)
# 后处理特征，可扩展
if sel.feature == 'staticFC':
mat_std_1d, mat_mean_1d = sel.postprocessing_features(mat_std), []
else:
mat_std_1d = sel.postprocessing_features(mat_std)
mat_mean_1d = sel.postprocessing_features(mat_mean)
# 打印load进度
if i % 10 == 0 or i == 0:
print('{}/{}\n'.format(i, len_all))
if i % 50 == 0 and i != 0:
e = time.time()
remaining_running_time = (e - s) * (len_all - i) / i
print('\nremaining time={:.2f} seconds \n'.format(
remaining_running_time))
return mat_std_1d, mat_mean_1d
def load_data_and_label1(sel):
sel.data=read_mat(sel.mat_path,dataset_name=None)
sel.label=pd.read_excel(sel.label_path)['诊断']
sel.folder=pd.read_excel(sel.label_path)['folder']
return sel
def resultFusion(rootPath=r'D:\myCodes\LC_MVPA\Python\MVPA_Python\perm',
datasetName=['predict', 'dec', 'y_sorted', 'weight']):
# Fusion of all block results of permutation test
fileName = selectFile(rootPath)
dataset = []
for dsname in datasetName:
Ds = []
for flname in fileName:
_, ds = read_mat(flname, dsname)
Ds.append(ds)
dataset.append(Ds)
all_metrics = pd.DataFrame(dataset)
all_metrics = all_metrics.rename(index={
0: 'predict',
1: 'dec',
2: 'y_sorted',
3: 'weight'
})
y_true = all_metrics.loc['y_sorted'][0]
y_pred = all_metrics.loc['predict'][0]
y_true = pd.DataFrame(y_true)
y_pred = pd.DataFrame(y_pred)
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score
accuracy_score(y_true.T, y_pred.T)
confusion_matrix(y_true.T, y_pred.T)
pr[1]
return dataset
def machine_learning(sel, order=[3, 4]):
# label
y = pd.concat([
sel.label[sel.label.values == order[0]],
sel.label[sel.label.values == order[1]]
])
y = y.values
# x/sel.mat
if os.path.exists(os.path.join(sel.save_path, sel.feature + '.mat')):
sel.mat = pd.DataFrame(
read_mat(os.path.join(sel.save_path, sel.feature + '.mat'),
None))
x = pd.concat([
sel.mat.iloc[sel.label.values == order[0], :],
sel.mat.iloc[sel.label.values == order[1], :]
])
# #平衡测试
# y=np.hstack([y,y[-1:-70:-1]])
# x=pd.concat([x,x.iloc[-1:-70:-1]],axis=0)
# y=y[60:]
# x=x.iloc[60:,:]
# print(sum(y==0),sum(y==1))
# 置换y
# rand_ind=np.random.permutation(len(y))
# y=y[rand_ind]
# cross-validation
# 1) split data to training and testing datasets
x_train, x_test, y_train, y_test = \
train_test_split(x, y, random_state=sel.random_state)
# elasticNet
print('elasticNetCV')
sel = ENCV.elasticNetCV()
sel.train(x_train, y_train)
sel.test(x_test)
results = sel.test(x_test).__dict__
# =============================================================================
#
# # rfe
# import lc_svc_rfe_cv_V2 as lsvc
# model=lsvc.svc_rfe_cv(k=5,pca_n_component=0.85)
#
# results=model.main_svc_rfe_cv(x.values,y)
# =============================================================================
results = results.__dict__
return results
def load_allmat(sel):
# 多线程
s = time.time()
print('loading all mat...\n')
# 判断是否有FC mat文件
if os.path.exists(os.path.join(sel.save_path, sel.feature + '.mat')):
sel.mat = pd.DataFrame(
read_mat(os.path.join(sel.save_path, sel.feature + '.mat'),
None))
print(
'Already have {}\nloaded all mat!\nrunning time={:.2f}'.format(
sel.feature + '.mat',
time.time() - s))
else:
sel.all_mat = os.listdir(sel.file_path)
all_mat_path = [
os.path.join(sel.file_path, all_mat_)
for all_mat_ in sel.all_mat
]
cores = multiprocessing.cpu_count()
if sel.n_processess > cores:
sel.n_processess = cores - 1
len_all = len(all_mat_path)
sel.mat = pd.DataFrame([])
# 特征用std还是mean
if sel.feature == 'mean':
ith = 1
elif sel.feature == 'std':
ith = 0
elif sel.feature == 'staticFC':
ith = 0
else:
print('###还未添加其他衡量dFC的指标,默认使用std###\n')
ith = 0
# load mat...
with ThreadPoolExecutor(sel.n_processess) as executor:
for i, all_mat_ in enumerate(all_mat_path):
task = executor.submit(sel.load_onemat_and_processing, i,
all_mat_, len_all, s)
sel.mat = pd.concat(
[sel.mat, pd.DataFrame(task.result()[ith]).T], axis=0)
# 保存后处理后的mat文件
if sel.if_save_post_mat:
write_mat(fileName=os.path.join(sel.save_path,
sel.feature + '.mat'),
dataset_name=sel.feature,
dataset=np.mat(sel.mat.values))
print('saved all {} mat!\n'.format(sel.feature))
Created on Tue Nov 13 20:55:51 2018
@author: lenovo
"""
import sys
sys.path.append(r'D:\myCodes\MVPA_LIChao\MVPA_Python\MVPA\utils')
from lc_read_write_Mat import read_mat
import numpy as np
import pandas as pd
hcPath = r'I:\dynamicALFF\Results\DALFF\50_0.9\Statistical_Results\Signal\ROISignals_ROISignal_FWHM4_HC.mat'
szPath = r'I:\dynamicALFF\Results\DALFF\50_0.9\Statistical_Results\Signal\ROISignals_ROISignal_FWHM4_SZ.mat'
bdPath = r'I:\dynamicALFF\Results\DALFF\50_0.9\Statistical_Results\Signal\ROISignals_ROISignal_FWHM4_BD.mat'
mddPath = r'I:\dynamicALFF\Results\DALFF\50_0.9\Statistical_Results\Signal\ROISignals_ROISignal_FWHM4_MDD.mat'
dataset_struct, datasetHC = read_mat(hcPath, 'ROISignals')
dataset_struct, datasetSZ = read_mat(szPath, 'ROISignals')
dataset_struct, datasetBD = read_mat(bdPath, 'ROISignals')
dataset_struct, datasetMDD = read_mat(mddPath, 'ROISignals')
meanHC = pd.DataFrame(np.mean(datasetHC, axis=0))
meanSZ = pd.DataFrame(np.mean(datasetSZ, axis=0))
meanBD = pd.DataFrame(np.mean(datasetBD, axis=0))
meanMDD = pd.DataFrame(np.mean(datasetMDD, axis=0))
allData = pd.concat([meanHC, meanSZ, meanBD, meanMDD], axis=1)
allData.index = [
'左侧额中回/额上回 ', '右侧额上回（靠内)', '右侧前扣带回 ', '右侧尾状核', '左侧尾状核', '右侧putamen',
'左侧putamen', '右侧前岛叶', '左侧前岛叶', '右侧杏仁核 ', '左侧杏仁核 ', '右侧海马', '左侧海马',
'右侧海马旁回', '左侧海马旁回', '右侧舌回', '左侧舌回', '右侧cuneus', '左侧cuneus',
'右侧angular gyrus', '右侧中央后回'
plotting.plot_matrix(matrix,
vmin=-1.,
vmax=1.,
colorbar=True,
title='Power correlation matrix')
# Tweak edge_threshold to keep only the strongest connections.
plotting.plot_connectome(matrix,
coords,
title='Power correlation graph',
edge_threshold='99.8%',
node_size=20,
colorbar=True)
file = r'D:\WorkStation_2018\WorkStation_dynamicFC_V3\Data\results\results_dfc\group_centroids_2.mat'
matrix = read_mat(file)
yeo_file = r'D:\My_Codes\Github_Related\Github_Code\Template_CBIG\stable_projects\brain_parcellation\Yeo2011_fcMRI_clustering\1000subjects_reference\Yeo_JNeurophysiol11_SplitLabels\MNI152\Yeo2011_17Networks_N1000.split_components.FSL_MNI152_2mm.nii.gz'
from nilearn import input_data
smoothed_img = input_data.smooth_img(yeo_file)
yeo = datasets.fetch_atlas_yeo_2011()
print('Yeo atlas nifti image (3D) with 17 parcels and liberal mask is located '
'at: %s' % yeo['thick_17'])
coords = plotting.find_parcellation_cut_coords(labels_img=yeo['thick_17'])
plotting.plot_connectome(matrix,
coords,
title='Power correlation graph',
edge_threshold='99.8%',
node_size=20,
colorbar=True)
""" Created on Tue Aug 28 15:07:49 2018 @author: lenovo """ from lc_svc_oneVsRest import oneVsRest import numpy as np import pandas as pd from lc_read_write_Mat import read_mat import sys sys.path.append(r'D:\myCodes\LC_MVPA\Python\MVPA_Python\utils') sys.path.append(r'D:\myCodes\LC_MVPA\Python\MVPA_Python\classfication') # X fileName = r'J:\分类测试_20180828\Ne-L_VS_Ne-R_n=709' dataset_name = 'coef' dataset_struct, dataset = read_mat(fileName, dataset_name) X = dataset X = pd.DataFrame(X) # y s = pd.read_excel(r'J:\分类测试_20180828\机器学习-ID.xlsx') dgns = s['诊断'].values # comb xandy = pd.concat([pd.DataFrame(dgns), X], axis=1) # NaN xandy = xandy.dropna() # X = xandy.iloc[:, 1:].values y = xandy.iloc[:, 0].values X = np.reshape(X, [len(X), X.shape[1]]) y = [int(d) for d in y] # predict and test