def explore_statistics(n_clusters_dir, items, ylabels, colors, val_fmt=''):
stats_path = pjoin(n_clusters_dir, 'statistics.csv')
with open(stats_path) as f:
stats = f.read().splitlines()
stats_items = stats[0].split(',')
stats_content = [line.split(',') for line in stats[1:]]
stats_content = list(zip(*stats_content))
stats_dict = {}
for idx, item in enumerate(stats_items):
stats_dict[item] = stats_content[idx]
x = np.arange(len(stats) - 1)
width = auto_bar_width(x)
for idx, item in enumerate(items):
plt.figure()
y = [float(_) for _ in stats_dict[item]]
rects = plt.bar(x, y, width, color=colors[idx])
show_bar_value(rects, val_fmt)
plt.xlabel('subgroup label')
plt.ylabel(ylabels[idx])
plt.title(item)
plt.xticks(x, stats_dict['label'])
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.savefig(pjoin(n_clusters_dir, '{}.png'.format(item)))
def compare_plot_bar():
import numpy as np
import matplotlib.pyplot as plt
from os.path import join as pjoin
# https://www.statsmodels.org/dev/_modules/statsmodels/stats/multitest.html
from statsmodels.stats.multitest import multipletests
from commontool.io.io import CsvReader
from commontool.algorithm.plot import auto_bar_width, show_bar_value
stru_name = 'myelin'
project_dir = '/nfs/s2/userhome/chenxiayu/workingdir/study/FFA_clustering'
stru_dir = pjoin(project_dir,
's2_25_zscore/HAC_ward_euclidean/2clusters/structure')
compare_dir = pjoin(stru_dir, 'compare')
compare_file = pjoin(compare_dir, stru_name)
multi_test_corrected = True
alpha = 0.001
compare_dict = CsvReader(compare_file).to_dict(1)
ps = np.array(list(map(float, compare_dict['p'])))
if multi_test_corrected:
reject, ps, alpha_sidak, alpha_bonf = multipletests(ps, 0.05, 'fdr_bh')
sample_names = [
name for idx, name in enumerate(compare_dict['sample_name'])
if ps[idx] < alpha
]
ps = [p for p in ps if p < alpha]
print('\n'.join(list(map(str, zip(sample_names, ps)))))
fig, ax = plt.subplots()
x = np.arange(len(sample_names))
width = auto_bar_width(x)
rects_p = ax.bar(x, ps, width, color='g', alpha=0.5)
show_bar_value(rects_p, '.2f')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.set_title(stru_name)
ax.set_ylabel('p', color='g')
ax.tick_params('y', colors='g')
ax.axhline(0.05)
ax.axhline(0.01)
ax.axhline(0.001)
ax.set_xticks(x)
ax.set_xticklabels(sample_names)
plt.setp(ax.get_xticklabels(),
rotation=25,
ha='right',
rotation_mode='anchor')
plt.tight_layout()
plt.show()
def test_bar_plot():
x = np.arange(3)
y = [1, 2, 3]
x_ticks = ['11', '22', '33']
width = auto_bar_width(x)
plt.figure()
rects = plt.bar(x, y, width, color='r')
show_bar_value(rects)
plt.xticks(x, x_ticks)
x = np.arange(1)
y = [1.3241]
width = auto_bar_width(x)
plt.figure()
rects = plt.bar(x, y, width, color='g')
show_bar_value(rects, '.3f')
def plot_age_distribution(proj_name='HCPD'):
import pandas as pd
from matplotlib import pyplot as plt
from commontool.algorithm.plot import show_bar_value
age_type = 'age in years'
fname = f'{proj_name}_SubjInfo.csv'
info_file = pjoin(work_dir, fname)
data = pd.read_csv(info_file)
ages = data[age_type].to_list()
ages_uniq = np.unique(ages)
y = [ages.count(age) for age in ages_uniq]
rects = plt.bar(ages_uniq, y, edgecolor='k', facecolor='w')
show_bar_value(rects)
plt.xlabel(age_type)
plt.xticks(ages_uniq, ages_uniq, rotation=45)
plt.ylabel('#subjects')
plt.title(fname)
plt.tight_layout()
plt.show()
def plot_reliability():
import numpy as np
import pickle as pkl
from os.path import join as pjoin
from matplotlib import pyplot as plt
from commontool.algorithm.plot import show_bar_value
hemi = 'both' # lh, rh, or both
thr = None
trg_dir = '/nfs/t3/workingshop/chenxiayu/study/FFA_pattern/analysis/s4_reliability'
if hemi == 'both':
reliability_lh = pkl.load(
open(pjoin(trg_dir, 'reliability_lh.pkl'), 'rb'))
reliability_rh = pkl.load(
open(pjoin(trg_dir, 'reliability_rh.pkl'), 'rb'))
corrs = reliability_lh['corr'] + reliability_rh['corr']
else:
reliability = pkl.load(
open(pjoin(trg_dir, f'reliability_{hemi}.pkl'), 'rb'))
corrs = reliability['corr']
if thr is not None:
corrs = [i for i in corrs if i > 0.5]
print('#correlation:', len(corrs))
bins = np.linspace(min(corrs), max(corrs), 30)
# plt.figure(figsize=(9, 4))
_, _, patches = plt.hist(corrs, bins, color='white', edgecolor='black')
plt.xlabel('correlation')
plt.title(f'{hemi} activation pattern reliability')
show_bar_value(patches, '.0f')
plt.tight_layout()
plt.show()
def roi_mean_plot(roi_mean_file, ROIitems, colors, xticklabels, ylabel=None, title=None, plot_style='violin'):
roi_mean_dict = CsvReader(roi_mean_file).to_dict(axis=1)
roi_means_list = [list(map(float, roi_mean_dict[ROIitem])) for ROIitem in ROIitems]
ROIitem_num = len(ROIitems)
for i in range(ROIitem_num):
for j in range(i+1, ROIitem_num):
print('{} vs. {}'.format(ROIitems[i], ROIitems[j]),
ttest_ind(roi_means_list[i], roi_means_list[j]))
plt.figure()
if plot_style == 'violin':
violin_parts = plt.violinplot(roi_means_list, showmeans=True)
for idx, pc in enumerate(violin_parts['bodies']):
# https://stackoverflow.com/questions/26291479/changing-the-color-of-matplotlibs-violin-plots
pc.set_color(colors[idx])
plt.xticks(range(1, ROIitem_num + 1), xticklabels)
elif plot_style == 'bar':
x = np.arange(ROIitem_num)
y = [np.mean(roi_means) for roi_means in roi_means_list]
sems = [sem(roi_means) for roi_means in roi_means_list]
width = auto_bar_width(x)
rects = plt.bar(x, y, width, edgecolor=colors[0], yerr=sems, facecolor='white')
show_bar_value(rects, '.2f')
plt.xticks(x, xticklabels)
else:
raise RuntimeError("Invalid plot style: {}".format(plot_style))
if ylabel is not None:
plt.ylabel(ylabel)
if title is not None:
plt.title(title)
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.tight_layout()
def plot_roi_info():
import numpy as np
import pickle as pkl
from matplotlib import pyplot as plt
from commontool.algorithm.plot import show_bar_value, auto_bar_width
roi_info_file = '/nfs/s2/userhome/chenxiayu/workingdir/study/FFA_pattern/analysis/s2_rh/zscore/' \
'HAC_ward_euclidean/100clusters/activation/ROIs/v3/rois_info.pkl'
roi_infos = pkl.load(open(roi_info_file, 'rb'))
# -plot n_group and n_subject-
x_labels = list(roi_infos.keys())
n_roi = len(x_labels)
x = np.arange(n_roi)
width = auto_bar_width(x)
# plot n_group
y_n_group = [info['n_group'] for info in roi_infos.values()]
rects_group = plt.bar(x,
y_n_group,
width,
facecolor='white',
edgecolor='black')
show_bar_value(rects_group)
plt.xticks(x, x_labels)
plt.ylabel('#group')
# plot n_subject
plt.figure()
y_n_subj = [info['n_subject'] for info in roi_infos.values()]
rects_subj = plt.bar(x,
y_n_subj,
width,
facecolor='white',
edgecolor='black')
show_bar_value(rects_subj)
plt.xticks(x, x_labels)
plt.ylabel('#subject')
# -plot sizes-
for roi, info in roi_infos.items():
plt.figure()
sizes = info['sizes']
bins = np.linspace(min(sizes), max(sizes), 50)
_, _, patches = plt.hist(sizes, bins, color='white', edgecolor='black')
plt.xlabel('#vertex')
plt.title(f'distribution of {roi} sizes')
show_bar_value(patches, '.0f')
plt.tight_layout()
plt.show()
def explore_label_dice(n_clusters_dir):
import nibabel as nib
from commontool.algorithm.tool import calc_overlap
c1_r = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster1_ROI_z2.3.label'))
c2_r = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster2_ROI_z2.3.label'))
c3_r = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster3_ROI_z2.3.label'))
c4_r1 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster4_ROI1_z2.3.label'))
c4_r2 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster4_ROI2_z2.3.label'))
c5_r1 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster5_ROI1_z2.3.label'))
c5_r2 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster5_ROI2_z2.3.label'))
c6_r1 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster6_ROI1_z2.3.label'))
c6_r2 = nib.freesurfer.read_label(
pjoin(n_clusters_dir, 'cluster6_ROI2_z2.3.label'))
c1_6_acti_top10 = nib.load(
pjoin(n_clusters_dir, 'top_acti_ROIs_percent10.0.nii.gz')).get_data()
c123_r_z = [c1_r, c2_r, c3_r]
c123_r_top = c1_6_acti_top10[:3]
c123_dice = []
c123_xticks = [
'c1_c2_z2.3', 'c1_c3_z2.3', 'c2_c3_z2.3', 'c1_c2_top10', 'c1_c3_top10',
'c2_c3_top10'
]
c456_r_z = [
np.concatenate((c4_r1, c4_r2)),
np.concatenate((c5_r1, c5_r2)),
np.concatenate((c6_r1, c6_r2))
]
c456_r_top = c1_6_acti_top10[3:]
c456_dice = []
c456_xticks = [
'c4_c5_z2.3', 'c4_c6_z2.3', 'c5_c6_z2.3', 'c4_c5_top10', 'c4_c6_top10',
'c5_c6_top10'
]
for idx, i in enumerate(c123_r_z[:-1]):
for j in c123_r_z[idx + 1:]:
c123_dice.append(calc_overlap(i, j))
for idx, i in enumerate(c123_r_top[:-1]):
for j in c123_r_top[idx + 1:]:
c123_dice.append(calc_overlap(i, j, 1, 1))
for idx, i in enumerate(c456_r_z[:-1]):
for j in c456_r_z[idx + 1:]:
c456_dice.append(calc_overlap(i, j))
for idx, i in enumerate(c456_r_top[:-1]):
for j in c456_r_top[idx + 1:]:
c456_dice.append(calc_overlap(i, j, 1, 1))
x = np.arange(6)
width = auto_bar_width(x)
plt.figure()
rects = plt.bar(x, c123_dice, width, color='b')
show_bar_value(rects, '.2%')
plt.ylabel('dice')
plt.xticks(x, c123_xticks)
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.figure()
rects = plt.bar(x, c456_dice, width, color='b')
show_bar_value(rects, '.2%')
plt.ylabel('dice')
plt.xticks(x, c456_xticks)
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
def explore_roi_stats(n_clusters_dir):
roi_path = pjoin(n_clusters_dir, 'mean_map_ROIs.nii.gz')
stats_path = pjoin(n_clusters_dir, 'statistics.csv')
roi_maps = read_nifti(roi_path)
stats_reader = CsvReader(stats_path)
row_dict = stats_reader.to_dict(keys=['#subjects'])
numb_items = ['1', '2']
numb_dict = OrderedDict()
for item in numb_items:
numb_dict[item] = 0
type_items = ['r_pFFA', 'r_mFFA', 'both', 'unknown']
type_dict = OrderedDict()
for item in type_items:
type_dict[item] = 0
for idx, roi_map in enumerate(roi_maps):
map_set = set(roi_map)
subjects_num = int(row_dict['#subjects'][idx])
if 0 not in map_set:
raise RuntimeError('Be careful! There is no zero in one roi_map')
if len(map_set) == 2:
numb_dict['1'] += subjects_num
if 1 in map_set:
type_dict['r_pFFA'] += subjects_num
elif 2 in map_set:
type_dict['r_mFFA'] += subjects_num
elif 3 in map_set:
type_dict['unknown'] += subjects_num
else:
raise RuntimeError(
'Be careful! the only one ROI label is not in (1, 2, 3)')
elif len(map_set) == 3:
numb_dict['2'] += subjects_num
if 1 in map_set and 2 in map_set:
type_dict['both'] += subjects_num
else:
raise RuntimeError(
'Be careful! the two ROI labels are not 1 and 2')
else:
raise RuntimeError(
'Be careful! the number of ROI labels is not 1 or 2')
plt.figure()
x = np.arange(len(numb_items))
width = auto_bar_width(x)
rects = plt.bar(x, numb_dict.values(), width, color='b')
show_bar_value(rects)
plt.ylabel('#subjects')
plt.xticks(x, numb_dict.keys())
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.savefig(pjoin(n_clusters_dir, 'numb_count.png'))
plt.figure()
x = np.arange(len(type_items))
width = auto_bar_width(x)
rects = plt.bar(x, type_dict.values(), width, color='b')
show_bar_value(rects)
plt.ylabel('#subjects')
plt.xticks(x, type_dict.keys())
ax = plt.gca()
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.savefig(pjoin(n_clusters_dir, 'type_count.png'))
male_num, female_num = explore_Gender(behavior_dict, label)
male_nums.append(male_num)
female_nums.append(female_num)
for item in float_items:
float_data_dict[item].append(
explore_float_data(behavior_dict, item, label))
# plot
x = np.arange(label_num)
width = auto_bar_width(x, 2)
plt.figure()
ax = plt.gca()
rects1 = ax.bar(x, male_nums, width, color='b')
rects2 = ax.bar(x + width, female_nums, width, color='r')
show_bar_value(rects1)
show_bar_value(rects2)
ax.legend((rects1[0], rects2[1]), ('male', 'female'))
ax.set_xticks(x + width / 2.0)
ax.set_xticklabels(labels)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.set_title('Gender')
ax.set_xlabel('subgroup label')
ax.set_ylabel('count')
# plt.savefig(pjoin(cluster_num_dir, 'Gender.png'))
width = auto_bar_width(x)
for item in float_items:
plt.figure()
ax = plt.gca()
def plot_mean_sem():
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from os.path import join as pjoin
from commontool.algorithm.plot import auto_bar_width, show_bar_value
stru_name = 'myelin'
project_dir = '/nfs/s2/userhome/chenxiayu/workingdir/study/FFA_clustering'
stru_dir = pjoin(project_dir,
's2_25_zscore/HAC_ward_euclidean/2clusters/structure')
mean_sem_file = pjoin(stru_dir, 'mean_sem/{}'.format(stru_name))
mean_sems = pd.read_csv(mean_sem_file)
intra_inter_pairs = np.array(
[[
'G1_acti_corr_G1_{}_lFFA'.format(stru_name),
'G1_acti_corr_G2_{}_lFFA'.format(stru_name)
],
[
'G2_acti_corr_G2_{}_lFFA'.format(stru_name),
'G2_acti_corr_G1_{}_lFFA'.format(stru_name)
],
[
'G1_acti_corr_G1_{}_rFFA'.format(stru_name),
'G1_acti_corr_G2_{}_rFFA'.format(stru_name)
],
[
'G2_acti_corr_G2_{}_rFFA'.format(stru_name),
'G2_acti_corr_G1_{}_rFFA'.format(stru_name)
]])
names = mean_sems['names'].to_list()
means = list(map(float, mean_sems['means']))
sems = list(map(float, mean_sems['sems']))
fig, ax = plt.subplots()
x = np.arange(intra_inter_pairs.shape[0])
item_num = intra_inter_pairs.shape[1]
width = auto_bar_width(x, item_num)
for idx in range(item_num):
sub_means = [means[names.index(i)] for i in intra_inter_pairs[:, idx]]
sub_sems = [sems[names.index(i)] for i in intra_inter_pairs[:, idx]]
rects = ax.bar(x + width * idx,
sub_means,
width,
color='k',
alpha=1. / ((idx + 1) / 2 + 0.5),
yerr=sub_sems)
show_bar_value(rects, '.3f')
xticklabels1 = [name for name in intra_inter_pairs[:, 0]]
xticklabels2 = [name for name in intra_inter_pairs[:, 1]]
xticklabels = xticklabels1 + xticklabels2
ax.set_xticks(np.r_[x, x + width * (item_num - 1)])
ax.set_xticklabels(xticklabels)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.set_ylabel('pearsonr')
plt.setp(ax.get_xticklabels(),
rotation=25,
ha='right',
rotation_mode='anchor')
plt.tight_layout()
plt.show()