def compare_BOLD_networks(self, group1_name, group2_name, comparison_method): group1_scan_list = self.get_scan_in_group(group1_name) group2_scan_list = self.get_scan_in_group(group2_name) self.group_nets[group1_name] = [loader.load_single_network(self.atlasobj, scan) for scan in group1_scan_list] self.group_nets[group2_name] = [loader.load_single_network(self.atlasobj, scan) for scan in group2_scan_list] stats_network, comp_p_network = netattr.networks_comparisons(self.group_nets[group2_name], self.group_nets[group1_name], comparison_method) return stats_network, comp_p_network
def test_real_data():
from mmdps.proc import loader, atlas
atlasobj = atlas.get('brodmann_lrce')
subject_1 = loader.load_single_network(atlasobj, 'caochangsheng_20161027')
subject_1 = subject_1.threshold(0.85)
subject_1_data = (np.abs(subject_1.data) > 0).astype(int)
subject_2 = loader.load_single_network(atlasobj, 'caochangsheng_20161114')
subject_2 = subject_2.threshold(0.85)
subject_2_data = (np.abs(subject_2.data) > 0).astype(int)
# subject_2_data = subject_1_data.copy()
# subject_2_data[0, 3] = 0
# subject_2_data[3, 0] = 0
print('graph similarity: ',
compare_two_graphs(subject_1_data, subject_2_data, 4, 4))
def calculate_real_healthy_range_threshold():
"""
This function loads in the healthy networks, calculate graph similarities, and plot them in one graph.
"""
atlasobj = atlas.get('brodmann_lrce')
healthy_group = group_manager.getHealthyGroup()
healthy_scans = [scan.filename for scan in healthy_group.scans]
healthy_nets = [
loader.load_single_network(atlasobj, filename)
for filename in healthy_scans
]
thresh_list = list(range(60, 100))
curve_mat = np.zeros((int(len(healthy_nets) * (len(healthy_nets) - 1) / 2),
len(thresh_list))) # each row is a curve
for count, threshold in enumerate(thresh_list):
print('Threshold: %d' % threshold)
threshold = threshold / 100.0
counter = 0 # counter in curve_mat row index
for i in range(len(healthy_nets)):
for j in range(i + 1, len(healthy_nets)):
# calculate similarities between each pair of networks
curve_mat[counter, count] = compare_two_graphs(
healthy_nets[i].binarize(threshold).data,
healthy_nets[j].binarize(threshold).data, 4, 4)
counter += 1
mdb = mongodb_database.MongoDBDatabase(None)
mdb.put_temp_data(
curve_mat, 'graph kernel inter-healthy',
'The inter-healthy network similarity calculated by graph kernel method with d = 4, h = 4. The binary threshold is varied range(60, 100). The data is of shape (cases, thresh_list length). Each row is a curve'
)
def calculate_SCI_vs_healthy():
"""
This function loads in healthy networks and SCI patients networks, calculate each
patients network similarity to each healthy controls, and plot each person's similarity
in one figure.
"""
atlasobj = atlas.get('brodmann_lrce')
healthy_group = group_manager.getHealthyGroup()
healthy_scans = [scan.filename for scan in healthy_group.scans]
healthy_nets = [
loader.load_single_network(atlasobj, filename)
for filename in healthy_scans
]
assistant = analysis_report.GroupAnalysisAssistant('jisuizhenjiaciji',
atlasobj)
study = assistant.study
group1 = study.getGroup('control 1')
scans = [scan.filename for scan in group1.scans]
group2 = study.getGroup('treatment 1')
scans += [scan.filename for scan in group2.scans]
SCI_nets = [loader.load_single_network(atlasobj, scan) for scan in scans]
curve_mat = np.zeros((len(scans) * len(healthy_scans),
len(thresh_list))) # each row is a curve
column_counter = 0
for subj_num, SCI_net in enumerate(SCI_nets):
for healthy_net in healthy_nets:
for count, threshold in enumerate(thresh_list):
print('Threshold: %d, pair: %d/%d' %
(threshold, column_counter, curve_mat.shape[0]))
threshold = threshold / 100.0
curve_mat[column_counter, count] = compare_two_graphs(
SCI_net.binarize(threshold).data,
healthy_net.binarize(threshold).data, 4, 4)
column_counter += 1
mdb = mongodb_database.MongoDBDatabase(None)
mdb.put_temp_data(
curve_mat, 'graph kernel SCI 1 x HC',
'The network similarity of each pair of SCI 1 and HC calculated by graph kernel method with d = 4, h = 4. The binary threshold is varied range(60, 100). The data is of shape (cases, thresh_list length). Each row is a curve.'
)
def calculate_real_range_parameters():
mdb = mongodb_database.MongoDBDatabase(None)
atlasobj = atlas.get('brodmann_lrce')
subject_1 = loader.load_single_network(atlasobj, 'caochangsheng_20161027')
subject_2 = loader.load_single_network(atlasobj, 'caochangsheng_20161114')
subject_3 = loader.load_single_network(atlasobj, 'chenyifan_20150612')
subject_4 = loader.load_single_network(atlasobj, 'chenyifan_20150629')
threshold_list = range(40, 60, 5)
h_list = list(range(1, int(atlasobj.count / 2)))
d_list = list(range(2, 10))
for thresh_count, threshold in enumerate(threshold_list, 1):
threshold = threshold / 100.0
sim_list_1 = np.zeros(len(h_list))
sim_list_2 = np.zeros(len(h_list))
sim_list_3 = np.zeros(len(h_list))
for idx, h in enumerate(h_list):
print('threshold: %d/%d, depth: %d/%d' %
(thresh_count, len(threshold_list), h, h_list[-1]))
sim_list_1[idx] = compare_two_graphs(
subject_1.binarize(threshold).data,
subject_2.binarize(threshold).data, h, 4)
sim_list_2[idx] = compare_two_graphs(
subject_3.binarize(threshold).data,
subject_4.binarize(threshold).data, h, 4)
sim_list_3[idx] = compare_two_graphs(
subject_1.binarize(threshold).data,
subject_3.binarize(threshold).data, h, 4)
threshold = int(threshold * 100.0)
mdb.put_temp_data(
sim_list_1, 'gk cao range h threshold %d' % threshold,
'caochangsheng_20161027 vs caochangsheng_20161114 network similarity measured by graph kernel with brodmann_lrce, threshold = %d, d = 4, h range(1, int(atlasobj.count/2))'
% threshold)
mdb.put_temp_data(
sim_list_2, 'gk chen range h threshold %d' % threshold,
'chenyifan_20150612 vs chenyifan_20150629 network similarity measured by graph kernel with brodmann_lrce, threshold = %d, d = 4, h range(1, int(atlasobj.count/2))'
% threshold)
mdb.put_temp_data(
sim_list_3, 'gk cao vs chen range h threshold %d' % threshold,
'caochangsheng_20161027 vs chenyifan_20150612 network similarity measured by graph kernel with brodmann_lrce, threshold = %d, d = 4, h range(1, int(atlasobj.count/2))'
% threshold)