-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathgenerate_main_plots.py
304 lines (235 loc) · 10.8 KB
/
generate_main_plots.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from scipy import io
from confusion_matrix import plot_confusion_matrix
from sklearn import metrics
import matplotlib.patches as mpatches
from nilearn import plotting, image
from scipy import interp
from sklearn.preprocessing import label_binarize
import os
if not os.path.exists('./plots'):
os.mkdir('./plots')
class_names = [' VIS', ' SM', 'DA', 'VA','L', 'FP','DMN', 'SUB', 'CER']
n_classes= len(class_names)
colors = ['dark blue','tan','lavender','olive','peach','puke green','blue grey', 'black','brown']
list_models=[[268,268,268,268],
[200,200,200,200],
[100,100,100,100],
[50,50,50,50],
[268,200,100,50],
[268,128,64,32],
[128,64,32,16],
[512,256,128,64],
[250,250,250,250],
[250,250,250],
[250,250],
[250],
[500,500,500,500],
[500,500,500],
[500,500],
[500],
[125,125,125,125],
[125,125,125],
[125,125],
[125]
]
# FIGURE 2
# Manual work: select the best model to plot.
# In our case, the best case is neural network [512,256,128,64]
# (7th model out of the neural network list)
folder_path = './results/task_tr_task_test_nn_models/'
name_model = 'categorical_' + str(7)
list_cm=[]
for i in range(50):
results=pd.read_csv(folder_path + name_model + '_'+str(i)+'_pred.csv')
cm = metrics.confusion_matrix(results['true'].values,results['pred'].values)
cm = cm.astype('float')/cm.sum(axis=1)[:, np.newaxis]
list_cm.append(cm)
cm_mean=np.mean(list_cm, axis=0)
cm_std=np.std(list_cm, axis=0)
fig, ax = plt.subplots()
plot_confusion_matrix(cm_mean, cm_std, classes=class_names, normalize=False,
title='Task-Training \n Task-Test (%)', cmap = plt.cm.YlGn)
plt.savefig('plots/Fig2.tiff',dpi=300,bbox_inches='tight')
#FIGURE 3
yeoROIs=io.loadmat('data/Shen268_yeo_RS7.mat',squeeze_me=True)['yeoROIs']-1
rsn_ROIs= [class_names[i] for i in yeoROIs]
array_dirs = sorted([f for f in os.listdir('./data/data_task_icafix/') if f.startswith('sub-')])
corrs_list = [np.corrcoef(np.loadtxt('./data/data_task_icafix/' + f + '/func_mean.txt')) for f in array_dirs]
corrs_list = [mat - np.identity(mat.shape[0]) for mat in corrs_list]
corr_task = np.asarray(corrs_list)
corr_task = corr_task.mean(axis=0)
task_corr = pd.DataFrame(corr_task[np.argsort(yeoROIs),:].transpose()).corr().values
fig= plt.figure(figsize=(10,5))
ax1=plt.subplot(1,2,1)
hm = sns.heatmap(task_corr,square=True,cbar_kws={"shrink": .6},
xticklabels=False, yticklabels=True, cmap=plt.cm.RdBu_r, ax=ax1)
x=0
y=0
w = [sum(yeoROIs==i) for i in range(9)]
y_ticks=[]
for i in range(9):
hm.add_patch(mpatches.Rectangle((x,y), w[i], w[i], fill=False,edgecolor='k',lw=3))
y_ticks.append(y + 0.5*w[i])
x = x + w[i]
y = y + w[i]
dat_to_plot=[]
for i, name in enumerate(class_names):
rsn_inds = np.where(np.sort(yeoROIs)==i)
mat = task_corr[np.meshgrid(rsn_inds[0],rsn_inds[0], indexing ='ij')]
#take upper off diagonal terms
dat_to_plot.append(mat[np.triu_indices(mat.shape[0], k=1)])
ax1.set_yticks(y_ticks)
ax1.set_yticklabels(class_names, {'fontweight':'bold'} )
ax2=plt.subplot(1, 2, 2)
x_for_boxplot = [np.repeat(i+1, len(dat_to_plot[i])) for i in range(9)]
x_for_boxplot = np.hstack(x_for_boxplot)
y_for_boxplot = np.hstack(dat_to_plot)
bp= sns.boxplot(x =x_for_boxplot, y=y_for_boxplot, linewidth=2.5,
whis = 1.5,showfliers=False, width=0.6,ax=ax2)
for i,box in enumerate(bp.artists):
box.set_edgecolor('k')
# iterate over whiskers and median lines
for j in range(5*i,5*(i+1)):
bp.lines[j].set_color('k')
bp.lines[5*i+4].set_color(sns.xkcd_rgb['off white'])
for patch, color in zip(bp.artists, sns.xkcd_palette(colors)):
patch.set_facecolor(color)
sns.swarmplot(x =x_for_boxplot, y=y_for_boxplot, size=1.2, color='0.3', ax=ax2)
bp.set_xticklabels(class_names, {'fontweight':'bold'})
ax2.set_ylim(-0.8,1)
plt.xticks(rotation=90)
plt.ylabel("Pearson similarity", size=20, fontweight='bold')
fig.subplots_adjust(wspace=0.3)
ax2.set_aspect(4)
plt.savefig('plots/Fig3.tiff',dpi=300,bbox_inches='tight')
#FIGURE 4
list_cm=[]
for i in range(50):
results=pd.read_csv('./results/task_tr_rest_tr/task_tr_resting_ts_'+str(i)+'_pred.csv')
cm = metrics.confusion_matrix(results['true'].values,results['pred'].values)
cm = cm.astype('float')/cm.sum(axis=1)[:, np.newaxis]
list_cm.append(cm)
cm_mean=np.mean(list_cm, axis=0)
cm_std=np.std(list_cm, axis=0)
fig, ax = plt.subplots()
plot_confusion_matrix(cm_mean, np.zeros((9,9)), classes=class_names, normalize=False,
title='Task-Training \n Resting-Test (%)', cmap = plt.cm.YlGn)
plt.savefig('plots/Fig4.tiff',dpi=300,bbox_inches='tight')
#FIGURE 5
list_scores=[]
list_res =[]
for i in range(50):
results=pd.read_csv('./results/task_tr_rest_tr/task_tr_resting_ts_'+str(i)+'_pred.csv')
list_res.append(results.loc[:,['pred','true']].values)
scores=np.load('./results/task_tr_rest_tr/task_tr_resting_ts_'+str(i) +'_pred_probs.npy')
list_scores.append(scores)
tpr_model= np.array([metrics.recall_score(list_res[i][:,1],list_res[i][:,0],average =None) \
for i in range(50)])
fpr_model= np.array([[sum(list_res[i][:,0][list_res[i][:,1]!= clas_id]==clas_id)/float(sum(list_res[i][:,1]!= clas_id)) \
for clas_id in range(n_classes)] for i in range(50)])
fig, ax = plt.subplots()
for i, color in zip(range(n_classes), sns.xkcd_palette(colors)):
tprs=[]
aucs=[]
mean_fpr = np.linspace(0, 1, 1000)
for fold in range(50):
y_test=label_binarize(list_res[fold][:,1],[0,1,2,3,4,5,6,7,8,9])
y_score=list_scores[fold]
fpr, tpr, thresholds = metrics.roc_curve(y_test[:,i], y_score[:,i])
auc_roc = metrics.roc_auc_score(y_test[:, i], y_score[:, i])
aucs.append(auc_roc)
tprs.append(interp(mean_fpr, fpr, tpr))
mean_tpr=np.mean(tprs, axis=0)
mean_auc= np.mean(aucs, axis=0)
std_auc= np.std(aucs, axis=0)
mean_tpr[-1] = 1.0
l, = plt.plot(mean_fpr, mean_tpr, color=color, lw=2, zorder=1,
label= class_names[i] + ' (%0.2f $\pm$ %0.2f)' % (mean_auc, std_auc))
plt.plot([0, 1], [0, 1], linestyle='--', lw=2, color='r',alpha=.3)
plt.scatter(np.mean(fpr_model,axis=0), np.mean(tpr_model,axis=0),marker='x', c ="#363737", s=100,zorder=2,
linewidths=100, edgecolors=None)
plt.legend(loc="lower right", ncol =1,
prop={'weight':'bold'},labelspacing=0.1,columnspacing=0.2, frameon=False)
plt.title('ROC curve', fontsize = 20, fontweight='bold')
plt.xlabel('FALSE POSITIVE RATE', fontsize = 15,fontweight='bold')
plt.ylabel('TRUE POSITIVE RATE', fontsize = 15,fontweight='bold')
plt.xticks(fontweight='bold')
plt.yticks(fontweight='bold')
plt.xlim([-0.05,1.05])
plt.ylim([-0.05,1.05])
plt.savefig('plots/Fig5.tiff',dpi=300,bbox_inches='tight')
#FIGURE 6
rec_model = np.array([metrics.recall_score(list_res[i][:,1],list_res[i][:,0],average =None) for i in range(50)])
prec_model = np.array([metrics.precision_score(list_res[i][:,1],list_res[i][:,0],average =None) for i in range(50)])
fig, ax = plt.subplots()
for i, color in zip(range(n_classes), sns.xkcd_palette(colors)):
precs=[]
avg=[]
mean_rec = np.linspace(0, 1, 1000)
for fold in range(50):
y_test=label_binarize(list_res[fold][:,1],[0,1,2,3,4,5,6,7,8,9])
y_score=list_scores[fold]
precision, recall, thresholds = metrics.precision_recall_curve(y_test[:,i], y_score[:,i])
average_precision = metrics.average_precision_score(y_test[:, i], y_score[:, i])
avg.append(average_precision)
precs.append(interp(mean_rec, recall[::-1], precision[::-1]))
mean_prec=np.mean(precs, axis=0)
mean_prec[-1] = 0.0
mean_avg= np.mean(avg, axis=0)
std_avg= np.std(avg, axis=0)
l, = plt.plot(mean_rec, mean_prec, color=color, lw=2, zorder=1,
label= class_names[i] + ' (%0.2f $\pm$ %0.2f)' % (mean_avg, std_avg))
plt.plot([0, 1], [1, 0], linestyle='--', lw=2, color='r',alpha=.3)
plt.scatter(np.mean(rec_model,axis=0), np.mean(prec_model, axis=0),marker='x', c ="#363737", s=100,zorder=2,
linewidths=100, edgecolors=None)
plt.legend(loc='lower left', ncol =2, prop={'weight':'bold'},labelspacing=0.1,columnspacing=0.2, frameon=False)
plt.title('PR curve', fontsize = 20, fontweight='bold')
plt.xlabel('Recall', fontsize = 15,fontweight='bold')
plt.ylabel('Precision', fontsize = 15,fontweight='bold')
plt.xticks(fontweight='bold')
plt.yticks(fontweight='bold')
plt.xlim([-0.05,1.05])
plt.ylim([-0.05,1.05])
plt.savefig('plots/Fig6.tiff',dpi=300,bbox_inches='tight')
#FIGURE 7
# Manual Work: All these pieces are then put together using inkscape
rsn_labels=io.loadmat('data/Shen268_yeo_RS7.mat')['yeoROIs'].flatten()
rois_ind_all=[np.load('./results/task_tr_rest_tr/task_tr_resting_ts_'+ str(fold)+'.npz')['rois_ids'] for fold in range(50)]
res_all=[pd.read_csv('./results/task_tr_rest_tr/task_tr_resting_ts_'+ str(fold)+'_pred.csv') for fold in range(50)]
nodes_acc=[np.asarray([metrics.accuracy_score(res_all[fold].iloc[rois_ind_all[fold]==idx,2],
res_all[fold].iloc[rois_ind_all[fold]==idx,1])
for idx in np.arange(268)]) for fold in range(50)]
nodes_acc = np.vstack(nodes_acc)
nodes_acc = np.mean(nodes_acc, axis=0)
vmax = max(nodes_acc)
vmin = min(nodes_acc)
#we use 1mm resolution for a finner result
img = image.load_img("./data/atlas/shen_1mm_268_parcellation.nii.gz")
for i in range(1,10):
a = img.get_data().copy()
sub = np.where(rsn_labels==i)[0] + 1
for index in np.ndindex(a.shape[0],a.shape[1],a.shape[2]):
xx=index[0]
yy=index[1]
zz=index[2]
if a[xx,yy,zz] in sub:
a[xx,yy,zz] = nodes_acc[int(a[xx,yy,zz])-1]
else:
a[xx,yy,zz] = 0
tmap = image.new_img_like(img,a, copy_header =True)
if i==9:
display = plotting.plot_glass_brain(tmap, threshold=vmin,display_mode='lzr',
title = class_names[i-1],
vmax=vmax,black_bg=False,
colorbar=True,symmetric_cbar=False)
else:
display = plotting.plot_glass_brain(tmap, threshold=vmin,display_mode='lzr',
title = class_names[i-1],
vmax=vmax,black_bg=False,
symmetric_cbar=False)
display.savefig("./plots/" + class_names[i-1]+"_glass_brain.png", dpi=300)
display.close()