-
Notifications
You must be signed in to change notification settings - Fork 1
/
drcValidFuncs.py
750 lines (600 loc) · 32 KB
/
drcValidFuncs.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
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
from env import *
import pandas as pd
import numpy as np
from matplotlib import pyplot as pl
import sklearn
import copy
import scipy
def addDRCValidData(validDf):
'''perform validation on DTI data from the DRC '''
#dtiSS = pd.read_csv('../data/DRC/DTI/DTI_summary_forRaz.xlsx')
dtiSS = pd.read_csv('data_processed/DTI_summary_forRaz.csv')
mappingIDtoRegion = {0 : ["Unclassified", "UNC"] ,
1: ["Middle cerebellar peduncle", "ICP"], # TBC
2: ["Pontine Crossing tract","PCT"], # TBC
3: ["Genu of corpus callosum", "GCC"],# cingulate or frontal
4: ["Body of corpus callosum", "BCC"],# cingulate or None
5: ["Splenium of corpus callosum", "SCC"], # cingulate or occipital
6: ["Fornix (column and body of fornix)","FX"], # hippo
7: ["Corticospinal tract R", "CST"],# frontal or None (wiki says 80-90% go to motor ctx from brain stem)
8: ["Corticospinal tract L", "CST"],# frontal
9: ["Medial lemniscus R", "ML"],## TBC
10: ["Medial lemniscus L","ML"], ## TBC
11: ["Inferior cerebellar peduncle R", "ICP"], ## TBC
12: ["Inferior cerebellar peduncle L", "ICP"], ## TBC
13: ["Superior cerebellar peduncle R", "SCP"], ## TBC
14: ["Superior cerebellar peduncle L", "SCP"], ## TBC
15: ["Cerebral peduncle R", "CP"], # TBC
16: ["Cerebral peduncle L", "CP"], # TBC
17: ["Anterior limb of internal capsule R", "ALIC"], # TBC or frontal
18: ["Anterior limb of internal capsule L", "ALIC"], # TBC or frontal
19: ["Posterior limb of internal capsule R", "PLIC"], # TBC or parietal (or None, wiki says connected to motor ctx.)
20: ["Posterior limb of internal capsule L", "PLIC"], # TBC or parietal (or None, wiki says connected to motor ctx.)
21: ["Retrolenticular part of internal capsule R", "RLIC"], # TBC or occipital
22: ["Retrolenticular part of internal capsule L", "RLIC"], # TBC or occipital
23: ["Anterior corona radiata R", "ACR"],# Frontal
24: ["Anterior corona radiata L", "ACR"],# Frontal
25: ["Superior corona radiata R", "SCR"], # Frontal or None
26: ["Superior corona radiata L", "SCR"],# Frontal or None
27: ["Posterior corona radiata R", "PCR"],# Parietal
28: ["Posterior corona radiata L", "PCR"],# Parietal
29: ["Posterior thalamic radiation R", "PTR"], # Parietal
30: ["Posterior thalamic radiation L", "PTR"], # Parietal
31: ["Sagittal stratum R", "SS"], # Temporal (connects inf temporal with sup parietal, passes through temporal)
32: ["Sagittal stratum L", "SS"],# Temporal
33: ["External capsule R", "EC"], # TBC
34: ["External capsule L", "EC"], # TBC
35: ["Cingulum (cingulate gyrus) R", "CGC"], # Cingulate
36: ["Cingulum (cingulate gyrus) L", "CGC"],# Cingulate
37: ["Cingulum (hippocampus) R", "CGH"], # hippocampus
38: ["Cingulum (hippocampus) L", "CGH"], # hippocampus
39: ["Fornix (cres) / Stria terminalis R", "FX"], # hippocampus
40: ["Fornix (cres) / Stria terminalis L", "FX"], # hippocampus
41: ["Superior longitudinal fasciculus R", "SLF"], # occip - update: wrong, should be parietal
42: ["Superior longitudinal fasciculus L", "SLF"], # occip - update: wrong, should be parietal
43: ["Superior fronto-occipital fasciculus R", "SFO"], # TBC or occip/frontal
44: ["Superior fronto-occipital fasciculus L", "SFO"], # TBC or occip/frontal
45: ["Uncinate fasciculus R", "UNC"], # TBC or temporal/frontal
46: ["Uncinate fasciculus L", "UNC"], # TBC or temporal/frontal
47: ["Tapetum R", "TP"],# TBC or temporal (these are fibers from corpus callosum that go to temporal)
48: ["Tapetum L", "TP"]}# TBC or temporal
dtiBiomkStructTemplate_updated = {
'CST':'TBC', # 'Frontal',
'ACR':'Frontal',
'SCR': 'TBC', #'Frontal',
'TP': 'Temporal', # 'TBC',
'PCR':'TBC', #Parietal',
'PTR': 'Parietal',
'SS': 'TBC', #'Temporal',
'UNC':'TBC',
'SLF':'TBC', #'Occipital',
'SFO':'TBC',
'CGC':'Cingulate',
'GCC': 'Frontal', #'TBC',
'BCC': 'TBC', #'Cingulate',
'SCC': 'Occipital', # 'Cingulate',
'CGH':'Hippocampus',
'FX':'Hippocampus',
'ALIC':'TBC',
'PLIC':'TBC',
'RLIC':'TBC',
'ICP':'TBC',
'SCP':'TBC',
'CP':'TBC',
'PCT':'TBC',
'EC':'TBC',
'ML':'TBC',
'n':'NA'
}
# dtiBiomkStructTemplate = {
# 'Frontal' : ['CST', 'ACR', 'SCR'],
# 'Parietal' : ['PCR', 'PTR'],
# 'Temporal' : ['SS'],
# 'Occipital' : ['SLF'], # not only occipital, but also frontal & temporal
# 'Cingulate' : ['CGC', 'GCC', 'BCC', 'SCC'],
# 'Hippocampus': ['CGH', 'FX']
# }
##########################
# remove subj ID 1719, fa values too low due to presence of artifact.
dtiSS = dtiSS[~np.in1d(dtiSS.Scan1Study, [1719, 1496, 1760])]
dtiSS.reset_index(drop=True, inplace=True)
print('-------------------------\n\n')
dtiSS.replace({'Diagnosis': {'AD (PCA)':4, 'Control':5}}, inplace=True)
print(dtiSS['Diagnosis'])
# print(ads)
dtiSS_grouped = dtiSS.groupby(['region', 'metric'])
diagNrs = [4,5]
diagLabels = ['PCA', 'Control']
diagColors = ['r', 'g']
nrWMregions = 48
nrSubj = dtiSS_grouped.get_group((0,'fa')).shape[0]
sortedIdMS = np.zeros((nrWMregions,int(nrSubj/2)))
for r in range(nrWMregions):
for measure in ['fa', 'md', 'ad']:
# print('r = %s' % mappingIDtoRegion[r], 'meas=', measure)
# print(dtiSS_grouped.get_group((r,measure))[['mean', 'Diagnosis']])
colLabel = mappingIDtoRegion[r][0].replace("/", "")
outFile = 'resfiles/tad-drc/validDf_%s_%d%s.png' % (measure, r, colLabel)
currGroup = dtiSS_grouped.get_group((r,measure)).reset_index()
measureCol = currGroup['mean']
diagCol = currGroup['Diagnosis']
print('colLabel', colLabel)
# visDfCol(measureCol, colLabel, diagCol, diagNrs, diagLabels, diagColors, outFile)
ctlIndx = np.where(diagCol == diagNrs[1])[0]
measCtl = measureCol[diagCol == diagNrs[1]].as_matrix().reshape(-1)
minInd = np.argmin(measCtl)
indxSorted = np.argsort(measCtl)
# print(measCtl)
# print('indxSorted', indxSorted)
# print('ctlIndx[indxSorted]', ctlIndx[indxSorted])
sortedIdMS[r,:] = currGroup.loc[ctlIndx[indxSorted],'Scan1Study'].as_matrix()
# print('indxSorted', currGroup.loc[ctlIndx[indxSorted],'Scan1Study'])
# print('minInd', currGroup.loc[ctlIndx[minInd],:])
print('sortedIdMS',colLabel, sortedIdMS[r,:])
# import pdb
# pdb.set_trace()
# print('sortedIdMS', sortedIdMS)
# remove subj ID 1719, fa values too low due to presence of artifact.
# dtiSS = dtiSS[dtiSS.Scan1Study != 1719]
# dtiSS.reset_index(drop=True, inplace=True)
# print(dtiSS.Scan1Study)
# print(ads)
###########################
dtiSS['region'] = dtiSS['region'].map(lambda x: \
'DTI FA '+dtiBiomkStructTemplate_updated[mappingIDtoRegion[x][1]])
# print(dtiSS)
# print(asd)
dtiSS_means = dtiSS.groupby(['Scan1Study','region', 'metric'])['mean']\
.mean().reset_index()
print(dtiSS.groupby(['Scan1Study','region', 'metric']).mean())
# print(adsa)
idx = dtiSS_means.metric == 'fa'
print('idx', idx)
# dtiSS_means.drop(idx, inplace=True)
dtiSS_means = dtiSS_means[idx]
dtiSS_means.reset_index(drop=True, inplace=True)
# print(asd)
dtiSS_pivoted = dtiSS_means.\
pivot(index = 'Scan1Study', columns = 'region', values = 'mean')
unqScans_dti = np.unique(dtiSS_pivoted.index)
unqScans_tad = np.unique(validDf.scanID)
Scan_inter = list(set(unqScans_dti) & set(unqScans_tad))
validDf_u = validDf.set_index('scanID')
validDf_u.update(dtiSS_pivoted)
validDf_u = validDf_u.reset_index()
# dtiBiomkStructTemplate = {
# 'Frontal' : ['CST', 'ACR', 'SCR'],
# 'Parietal' : ['PCR', 'PTR'],
# 'Temporal' : ['SS'],
# 'Occipital' : ['SLF'], # not only occipital, but also frontal & temporal
# 'Cingulate' : ['CGC', 'GCC', 'BCC', 'SCC'],
# 'Hippocampus': ['CGH', 'FX']
# }
# print('validDf_u', validDf_u)
# multiply by the number of regions we averaged in the original ADNI model. double because of L+R
# TODO: make ADNI only take the mean, so we don't need to do this anymore.
validDf_u['DTI FA Frontal'] *= 6
validDf_u['DTI FA Parietal'] *= 4
validDf_u['DTI FA Temporal'] *= 2
validDf_u['DTI FA Occipital'] *= 2
validDf_u['DTI FA Cingulate'] *= 8
validDf_u['DTI FA Hippocampus'] *= 4
print('validDf_u', validDf_u)
# print(validDf_u)
# print(asda)
return validDf_u
def visValidDf(validDf, outFilePrefix):
fig = pl.figure(5)
# print(validDf.columns.tolist())
# print(adsa)
dtiCols = validDf.loc[:, 'DTI FA Cingulate' : 'DTI FA Temporal'].columns.tolist()
dtiDf = validDf[dtiCols]
# validDfByDiag = validDf.groupby([diag])
diagNrs = [4,5]
ctlVals = validDf.loc[validDf.diag == 4, dtiCols[0]].dropna()
patVals = validDf.loc[validDf.diag == 5, dtiCols[0]].dropna()
ridSortedCtlBS = np.zeros((len(dtiCols),ctlVals.shape[0]))
ridSortedPatBS = np.zeros((len(dtiCols), patVals.shape[0]))
for b in range(len(dtiCols)):
pl.clf()
ctlVals = validDf.loc[validDf.diag == 4, dtiCols[b]].dropna()
patVals = validDf.loc[validDf.diag == 5, dtiCols[b]].dropna()
pl.hist(ctlVals, color='g', label='ctl', histtype='step',fill=False)
pl.hist(patVals, color='r', label='pca', histtype='step',fill=False)
pl.title(dtiCols[b])
pl.legend()
fig.show()
outFile = 'resfiles/tad-drc/valid_%s_%d_%s.png' % (outFilePrefix, b, dtiCols[b])
fig.savefig(outFile)
# print(adas)
ctlScanID = validDf.loc[np.logical_and(validDf.diag == 4, ~np.isnan(validDf[dtiCols[b]])), 'scanID'].as_matrix().reshape(-1)
patScanID = validDf.loc[np.logical_and(validDf.diag == 5, ~np.isnan(validDf[dtiCols[b]])), 'scanID'].as_matrix().reshape(-1)
print('ctlScanID', ctlScanID)
print('patScanID', patScanID)
print('~np.isnan(validDf[dtiCols[b]])', np.sum(~np.isnan(validDf[dtiCols[b]])))
print('validDf.diag == 4 ', np.sum(validDf.diag == 4) )
# print(ads)
idxSortCtl = np.argsort(ctlVals)
idxSortPat = np.argsort(patVals)
ridSortedCtlBS[b,:] = ctlScanID[idxSortCtl]
ridSortedPatBS[b,:] = patScanID[idxSortPat]
print('ridSortedCtlBS', ridSortedCtlBS)
print('ridSortedPatBS', ridSortedPatBS)
# print(ads)
def visDfCol(dfCol, colLabel, diagCol, diagNrs, diagLabels, diagColors, outFile):
fig = pl.figure(1)
nrDiags = len(diagNrs)
pl.clf()
for d in range(nrDiags):
# print(dfCol)
# print(diagCol)
# print(diagNrs[d])
# print(diagCol == diagNrs[d])
# print(dfCol.loc[diagCol == diagNrs[d]])
# print(dfCol[diagCol == diagNrs[d]])
pl.hist(dfCol.loc[diagCol == diagNrs[d]].dropna(), color=diagColors[d], label=diagLabels[d], histtype='step',fill=False)
pl.legend()
pl.title(colLabel)
fig.show()
outFile = outFile.replace(" ", "_")
fig.savefig(outFile)
# print(ads)
def validateDRCBiomk(dpmObj, params):
# first predict subject DTI measures
diag = params['diag']
disNr = 1 # predict for DRC subjects
indxSubjToKeep = dpmObj.getIndxSubjToKeep(disNr)
import DPMModelGeneric
Xfilt, Yfilt = DPMModelGeneric.DPMModelGeneric.filterXYsubjInd(params['X'], params['Y'], indxSubjToKeep)
diagSubjCurrDis = diag[indxSubjToKeep]
#### construct sub-shifts for each biomarker
XshiftedDisModelBS, ysPredBS, xsOrigPred1S = dpmObj.getDataDisOverBiomk(disNr)
for b2 in range(dpmObj.nrBiomk):
assert len(params['X'][b2]) == len(params['Y'][b2])
assert len(XshiftedDisModelBS[b2]) == len(Yfilt[b2])
# now get the validation set. This is already only for the DRC subjects
Xvalid = params['Xvalid']
Yvalid = params['Yvalid']
RIDvalid = params['RIDvalid']
diagValid = params['diagValid']
labels = params['labels']
print('labels', labels)
nonMriBiomksList = [i for i in range(len(labels)) if labels[i].startswith('DTI')]
mriBiomksList = [i for i in range(len(labels)) if labels[i].startswith('Volume')]
assert len(ysPredBS) == len(Yvalid)
nrDtiCols = len(nonMriBiomksList)
mse = [0 for b in nonMriBiomksList]
# subjects who have DTI validation
subjWithValidIndx = np.where([ys.shape[0] > 0 for ys in Yvalid[nonMriBiomksList[0]]])[0]
nrSubjWithValid = subjWithValidIndx.shape[0]
XvalidFilt, YvalidFilt = DPMModelGeneric.DPMModelGeneric.filterXYsubjInd(Xvalid, Yvalid, subjWithValidIndx)
diagValidFilt = diagValid[subjWithValidIndx]
RIDvalidFilt = RIDvalid[subjWithValidIndx]
ridCurrDis = params['RID'][indxSubjToKeep]
XvalidShifFilt = [[[] for s in range(nrSubjWithValid)] for b in range(dpmObj.nrBiomk)]
###### construct the shifts of the subjects in validation set #############
for b2 in range(nrDtiCols):
mseList = []
for s in range(RIDvalidFilt.shape[0]):
# for each validation subject
idxCurrDis = np.where(RIDvalidFilt[s] == ridCurrDis)[0][0]
xsOrigFromModel = xsOrigPred1S[idxCurrDis]
assert np.where(xsOrigFromModel == XvalidFilt[nonMriBiomksList[b2]][s][0])[0].shape[0] == 1
idxXsWithValid = np.where(xsOrigFromModel == XvalidFilt[nonMriBiomksList[b2]][s][0])[0][0]
ysPredCurrSubj = ysPredBS[nonMriBiomksList[b2]][idxCurrDis][idxXsWithValid]
assert YvalidFilt[nonMriBiomksList[b2]][s].shape[0] > 0
mseList += [(ysPredCurrSubj - YvalidFilt[nonMriBiomksList[b2]][s][0]) ** 2]
# also compose the shifted Xs for the validation subjects
xsShiftedFromModel = XshiftedDisModelBS[0][idxCurrDis]
XvalidShifFilt[nonMriBiomksList[b2]][s] = np.array([xsShiftedFromModel[idxXsWithValid]])
assert XvalidShifFilt[nonMriBiomksList[b2]][s].shape[0] == YvalidFilt[nonMriBiomksList[b2]][s].shape[0]
mse[b2] = np.mean(mseList)
# part 2. plot the inferred dynamics for DRC data:
# every biomarker against original DPS
# also plot extra validation data on top
xsTrajX = dpmObj.getXsMinMaxRange(disNr)
predTrajXB = dpmObj.predictBiomkSubjGivenXs(xsTrajX, disNr)
trajSamplesBXS = dpmObj.sampleBiomkTrajGivenXs(xsTrajX, disNr, nrSamples=100)
### build a simpler linear predictor from MR to DTI for every ROI independently.
# Train it on ADNI MR+DTI data and use it to predict DRC-DTI from DRC-MR.
dataDfAll = params['dataDfAll']
colsList = dataDfAll.columns.tolist()
mriBiomksColsInd = [i for i in range(len(colsList)) if colsList[i].startswith('Volume')]
nonMriBiomksColsInd = [i for i in range(len(colsList)) if colsList[i].startswith('DTI')]
nrMriBiomks = len(mriBiomksColsInd)
dataAllTrainDf = dataDfAll.loc[dataDfAll.dataset == 1,:]
dataAllTrain = dataAllTrainDf.values
# print('dataAllTrain', dataAllTrain)
nrNonMriBiomk = len(mriBiomksColsInd)
YvalidDktNonMri = [0 for f in range(nrNonMriBiomk)]
YvalidLinModelNonMri = [0 for f in range(nrNonMriBiomk)]
YvalidSplineModelNonMri = [0 for f in range(nrNonMriBiomk)]
YvalidMultivarModelNonMri = [0 for f in range(nrNonMriBiomk)]
mseDpm = np.zeros(nrNonMriBiomk)
mseLin = np.zeros(nrNonMriBiomk)
mseSpline = np.zeros(nrNonMriBiomk)
mseMultivar = np.zeros(nrNonMriBiomk)
squaredErrorsDpm = [[] for f in range(nrNonMriBiomk)]
squaredErrorsLin = [[] for f in range(nrNonMriBiomk)]
squaredErrorsSpline = [[] for f in range(nrNonMriBiomk)]
squaredErrorsMultivar = [[] for f in range(nrNonMriBiomk)]
# select just the DTI biomarkers
nonMriColsArrayIndx = np.array(nonMriBiomksList)
mriColsArrayIndx = np.array(mriBiomksList)
predTrajNonMriXB = predTrajXB[:,nonMriColsArrayIndx]
predTrajMriXB = predTrajXB[:, mriColsArrayIndx]
trajSamplesNonMriBXS = trajSamplesBXS[nonMriColsArrayIndx,:,:]
XvalidShifNonMriFilt = [XvalidShifFilt[b] for b in nonMriBiomksList]
YvalidFiltNonMri = [YvalidFilt[b] for b in nonMriBiomksList]
YvalidFiltMriClosestToNonMri = [[[] for s in range(nrSubjWithValid) ] for b in nonMriBiomksColsInd] # only the MRI where Non-MRI exists
nonMriValValidAll = [[] for b in range(nrNonMriBiomk)]
nonMriPredValidDktAll = [[] for b in range(nrNonMriBiomk)]
nonMriPredValidLinAll = [[] for b in range(nrNonMriBiomk)]
nonMriPredValidSplineAll = [[] for b in range(nrNonMriBiomk)]
nonMriPredValidMultivarAll = [[] for b in range(nrNonMriBiomk)]
corrDpm = np.zeros(nrNonMriBiomk)
pValDpm = np.zeros(nrNonMriBiomk)
corrLin = np.zeros(nrNonMriBiomk)
pValLin = np.zeros(nrNonMriBiomk)
corrSpline = np.zeros(nrNonMriBiomk)
pValSpline = np.zeros(nrNonMriBiomk)
corrMultivar = np.zeros(nrNonMriBiomk)
pValMultivar = np.zeros(nrNonMriBiomk)
for b in range(nrNonMriBiomk):
mriDataCurrCol = dataAllTrain[:, mriBiomksColsInd[b]]
nonMriDataCurrCol = dataAllTrain[:, nonMriBiomksColsInd[b]]
nnMask = ~np.isnan(mriDataCurrCol) & ~np.isnan(nonMriDataCurrCol)
linModel = sklearn.linear_model.LinearRegression(fit_intercept=True)
linModel.fit(mriDataCurrCol[nnMask].reshape(-1,1),
nonMriDataCurrCol[nnMask].reshape(-1,1))
from scipy.interpolate import Rbf, UnivariateSpline
sortedXInd = np.argsort(mriDataCurrCol[nnMask])
splineModel = UnivariateSpline(mriDataCurrCol[nnMask][sortedXInd],
nonMriDataCurrCol[nnMask][sortedXInd], k=3)
# Now do multivariate prediction. Based on all MRI values at each lobe, predict one DTI biomk.
nnMaskMulti = (np.sum(np.isnan(dataAllTrain[:, mriBiomksColsInd]),axis=1) == 0) & ~np.isnan(nonMriDataCurrCol)
# print('print(np.sum(nnMaskMulti))', np.sum(nnMaskMulti))
assert np.isfinite(dataAllTrain[nnMaskMulti, :][:, mriBiomksColsInd]).all()
assert np.isfinite(nonMriDataCurrCol[nnMaskMulti]).all()
nnDataMri = dataAllTrain[nnMaskMulti, :][:, mriBiomksColsInd]
# print(nnDataMri[:,0].shape[0])
# print(nonMriDataCurrCol[nnMaskMulti].shape[0])
assert nnDataMri[:,0].shape[0] == nnDataMri[:, 1].shape[0]
assert nnDataMri[:,0].shape[0] == nonMriDataCurrCol[nnMaskMulti].shape[0]
# multivarModel = Rbf(nnDataMri[:,0], nnDataMri[:,1], nnDataMri[:,2], nnDataMri[:,3], nnDataMri[:,4],
# nnDataMri[:,5], nonMriDataCurrCol[nnMaskMulti], function='cubic')
# print(ads)
from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn.gaussian_process.kernels import RBF, WhiteKernel
multivarModelRbf = Rbf(nnDataMri[:,0], nnDataMri[:,1], nnDataMri[:,2], nnDataMri[:,3], nnDataMri[:,4],
nnDataMri[:,5], nonMriDataCurrCol[nnMaskMulti], function='cubic')
kernel = 1.0 * RBF(length_scale=1, length_scale_bounds=(0.1, 10)) \
+ WhiteKernel(noise_level=0.3, noise_level_bounds=(0.1, 0.1))
# print('nnDataMri[:,0:5].shape', nnDataMri[:,0:6].shape)
# print(nonMriDataCurrCol[nnMaskMulti].shape)
multivarModel = GaussianProcessRegressor(kernel=kernel,
alpha=0.0).fit(nnDataMri[:,0:6], nonMriDataCurrCol[nnMaskMulti])
YvalidDktNonMri[b] = [[] for s in range(nrSubjWithValid)] # Non-MRI predictions of DKT model for subj in validation set
YvalidLinModelNonMri[b] = [[] for s in range(nrSubjWithValid) ] # Non-MRI predictions of linear model for subj in validation set
YvalidSplineModelNonMri[b] = [[] for s in range(nrSubjWithValid)] # Non-MRI predictions of linear model for subj in validation set
YvalidMultivarModelNonMri[b] = [[] for s in range(nrSubjWithValid)] # Non-MRI predictions of linear model for subj in validation set
for s in range(nrSubjWithValid):
mrValsValidCurrSubj = np.array(YvalidFilt[mriBiomksList[b]][s]).reshape(-1,1)
nonMriValValidCurrSubj = YvalidFilt[nonMriBiomksList[b]][s][0]
xMriCurr = np.array(XvalidFilt[mriBiomksList[b]][s])
xDTICurr = XvalidFilt[nonMriBiomksList[b]][s][0]
closestMriIdx = np.argmin(np.abs(xMriCurr - xDTICurr))
YvalidFiltMriClosestToNonMri[b][s] = mrValsValidCurrSubj[closestMriIdx]
nonMriPredValidLin = linModel.predict(mrValsValidCurrSubj[closestMriIdx].reshape(-1,1))
nonMriPredValidLin = nonMriPredValidLin[0][0]
nonMriPredValidSpline = splineModel(mrValsValidCurrSubj[closestMriIdx].astype(float))
mrValsToPredictCurrSubj = [YvalidFilt[mriBiomksList[u]][s][closestMriIdx]
for u in range(nrMriBiomks)]
# print('mrValsToPredictCurrSubj', mrValsToPredictCurrSubj)
assert len(mrValsToPredictCurrSubj) == nrMriBiomks
# nonMriPredValidMultivar = multivarModel(mrValsToPredictCurrSubj[0], mrValsToPredictCurrSubj[1],
# mrValsToPredictCurrSubj[2], mrValsToPredictCurrSubj[3], mrValsToPredictCurrSubj[4],
# mrValsToPredictCurrSubj[5])
nonMriPredValidMultivar = multivarModel.predict(np.array(mrValsToPredictCurrSubj).reshape(1, -1))
# print(np.array([nonMriPredValidLin]))
# print(mrValsValidCurrSubj[closestMriIdx])
# print(dasdsa)
assert nonMriPredValidLin != []
YvalidLinModelNonMri[b][s] = [nonMriPredValidLin]
YvalidSplineModelNonMri[b][s] = [nonMriPredValidSpline]
YvalidMultivarModelNonMri[b][s] = [nonMriPredValidMultivar]
indOfXTrajClosestToCurrSubj = np.argmin(np.abs(XvalidShifNonMriFilt[b][s][0] - xsTrajX))
nonMriPredValidDkt = predTrajNonMriXB[indOfXTrajClosestToCurrSubj, b]
YvalidDktNonMri[b][s] = [np.array(nonMriPredValidDkt)]
# if diagValidFilt[s] == PCA:
squaredErrorsDpm[b] += [(nonMriValValidCurrSubj - nonMriPredValidDkt) ** 2]
squaredErrorsLin[b] += [(nonMriValValidCurrSubj - nonMriPredValidLin) ** 2]
squaredErrorsSpline[b] += [(nonMriValValidCurrSubj - nonMriPredValidSpline) ** 2]
squaredErrorsMultivar[b] += [(nonMriValValidCurrSubj - nonMriPredValidMultivar) ** 2]
nonMriValValidAll[b] += [nonMriValValidCurrSubj]
nonMriPredValidDktAll[b] += [nonMriPredValidDkt]
nonMriPredValidLinAll[b] += [nonMriPredValidLin]
nonMriPredValidSplineAll[b] += [nonMriPredValidSpline]
nonMriPredValidMultivarAll[b] += [nonMriPredValidMultivar]
YvalidFiltMriClosestToNonMri[b][s] = np.array(YvalidFiltMriClosestToNonMri[b][s])
YvalidDktNonMri[b][s] = np.array(YvalidDktNonMri[b][s])
YvalidLinModelNonMri[b][s] = np.array(YvalidLinModelNonMri[b][s])
YvalidSplineModelNonMri[b][s] = np.array(YvalidSplineModelNonMri[b][s])
YvalidMultivarModelNonMri[b][s] = np.array(YvalidMultivarModelNonMri[b][s])
print(nonMriPredValidLin)
print(YvalidLinModelNonMri[b][s])
assert YvalidLinModelNonMri[b][s].shape[0] > 0
nonMriValValidAll[b] = np.array(nonMriValValidAll[b]).reshape(-1, 1).astype(float)
nonMriPredValidDktAll[b] = np.array(nonMriPredValidDktAll[b]).reshape(-1, 1).astype(float)
nonMriPredValidLinAll[b] = np.array(nonMriPredValidLinAll[b]).reshape(-1, 1).astype(float)
nonMriPredValidSplineAll[b] = np.array(nonMriPredValidSplineAll[b]).reshape(-1, 1).astype(float)
nonMriPredValidMultivarAll[b] = np.array(nonMriPredValidMultivarAll[b]).reshape(-1, 1).astype(float)
# print('nonMriValValidAll', nonMriValValidAll[b].shape, nonMriValValidAll[b])
# print('nonMriPredValidLinAll', nonMriPredValidLinAll[b].shape, nonMriPredValidLinAll[b])
# print('nonMriPredValidDktAll', nonMriPredValidDktAll[b].shape, nonMriPredValidDktAll[b])
corrDpm[b], pValDpm[b] = scipy.stats.spearmanr(nonMriValValidAll[b],
nonMriPredValidDktAll[b])
corrLin[b], pValLin[b] = scipy.stats.spearmanr(nonMriValValidAll[b],
nonMriPredValidLinAll[b])
corrSpline[b], pValSpline[b] = scipy.stats.spearmanr(nonMriValValidAll[b],
nonMriPredValidSplineAll[b])
corrMultivar[b], pValMultivar[b] = scipy.stats.spearmanr(nonMriValValidAll[b],
nonMriPredValidMultivarAll[b])
for b in range(nrNonMriBiomk):
for s in range(nrSubjWithValid):
print(YvalidFiltMriClosestToNonMri[b][s].shape)
print(YvalidLinModelNonMri[b][s].shape)
if YvalidFiltMriClosestToNonMri[b][s].shape[0] != YvalidLinModelNonMri[b][s].shape[0]:
print('b s', b, s)
print(YvalidFiltMriClosestToNonMri[b][s])
print(YvalidLinModelNonMri[b][s])
raise ValueError('array shapes do not match')
if YvalidFiltMriClosestToNonMri[b][s].shape[0] != YvalidDktNonMri[b][s].shape[0]:
print('b s', b, s)
print(YvalidFiltMriClosestToNonMri[b][s])
print(YvalidDktNonMri[b][s])
raise ValueError('array shapes do not match')
if YvalidFiltMriClosestToNonMri[b][s].shape[0] != YvalidFiltNonMri[b][s].shape[0]:
print('b s', b, s)
print(YvalidFiltMriClosestToNonMri[b][s])
print(YvalidFiltNonMri[b][s])
raise ValueError('array shapes do not match')
for b in range(nrNonMriBiomk):
squaredErrorsDpm[b] = np.array(squaredErrorsDpm[b])
squaredErrorsLin[b] = np.array(squaredErrorsLin[b])
squaredErrorsSpline[b] = np.array(squaredErrorsSpline[b])
squaredErrorsMultivar[b] = np.array(squaredErrorsMultivar[b])
# nonMriValValidAll[b] = nonMriValValidAll[b]
# nonMriPredValidDktAll[b] = nonMriPredValidDktAll[b]
# nonMriPredValidLinAll[b] = nonMriPredValidLinAll[b]
nrBootStraps = 50
mseDpmUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
mseLinUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
mseSplineUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
mseMultivarUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
corrDpmUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
corrLinUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
corrSplineUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
corrMultivarUB = np.zeros((nrNonMriBiomk, nrBootStraps), float)
for b in range(nrNonMriBiomk):
nrSubjWithValidAndChosen = len(squaredErrorsLin[b])
assert nonMriValValidAll[b].shape[0] == len(squaredErrorsLin[b])
for b2 in range(nrBootStraps):
idxBootCurr = np.array(np.random.choice(nrSubjWithValidAndChosen,
nrSubjWithValidAndChosen), int)
# print(len(squaredErrorsLin[b]))
# print(idxBootCurr)
mseDpmUB[b, b2] = np.mean(squaredErrorsLin[b][idxBootCurr])
mseLinUB[b, b2] = np.mean(squaredErrorsDpm[b][idxBootCurr])
mseSplineUB[b, b2] = np.mean(squaredErrorsSpline[b][idxBootCurr])
mseMultivarUB[b, b2] = np.mean(squaredErrorsMultivar[b][idxBootCurr])
idxBootCorrCurr = np.array(np.random.choice(nrSubjWithValidAndChosen,
nrSubjWithValidAndChosen), int)
print(idxBootCorrCurr)
print(nonMriValValidAll[b])
print(nonMriValValidAll[b][idxBootCorrCurr])
print(nonMriPredValidDktAll[b][idxBootCorrCurr])
print(corrDpmUB[b, b2])
corrDpmUB[b, b2], _ = scipy.stats.spearmanr(nonMriValValidAll[b][idxBootCorrCurr],
nonMriPredValidDktAll[b][idxBootCorrCurr])
corrLinUB[b, b2], _ = scipy.stats.spearmanr(nonMriValValidAll[b][idxBootCorrCurr],
nonMriPredValidLinAll[b][idxBootCorrCurr])
corrSplineUB[b, b2], _ = scipy.stats.spearmanr(nonMriValValidAll[b][idxBootCorrCurr],
nonMriPredValidSplineAll[b][idxBootCorrCurr])
corrMultivarUB[b, b2], _ = scipy.stats.spearmanr(nonMriValValidAll[b][idxBootCorrCurr],
nonMriPredValidMultivarAll[b][idxBootCorrCurr])
metrics = {}
metrics['dpm'] = {}
metrics['dpm']['corrUB'] = corrDpmUB
# metrics['dpm']['pValsU'] = pValDpm
metrics['dpm']['mseUB'] = mseDpmUB
metrics['lin'] = {}
metrics['lin']['corrUB'] = corrLinUB
# metrics['lin']['pValsU'] = pValLin
metrics['lin']['mseUB'] = mseLinUB
metrics['spline'] = {}
metrics['spline']['corrUB'] = corrSplineUB
metrics['spline']['mseUB'] = mseSplineUB
metrics['multivar'] = {}
metrics['multivar']['corrUB'] = corrMultivarUB
metrics['multivar']['mseUB'] = mseMultivarUB
labelsNonMri = [params['labels'][b] for b in nonMriBiomksList]
metrics['labelsNonMri'] = labelsNonMri
# plot against MRI vals instead of DPS time-shifts
# also plot training data DTI[b] in MRI[b] space
YDti = [params['Y'][b] for b in nonMriBiomksList]
YMriClosestToDti = [[0 for s in range(len(YDti[b]))] for b in mriBiomksList] # only the MRI where DTI exists
# idxWithDti = [s for s in range(len(YDti)) ]
# print('YDti', YDti)
# print(adsa)
for b in range(nrNonMriBiomk):
for s in range(len(YDti[b])):
YMriClosestToDti[b][s] = np.array([])
if YDti[b][s].shape[0] > 0:
xsMriCurrSubj = params['X'][mriBiomksList[b]][s]
xsDtiCurrSubj = params['X'][nonMriBiomksList[b]][s]
mriValsCorrespToDtiCurrSubj = []
for t in range(xsDtiCurrSubj.shape[0]):
mriIndClosestToCurrDtiScan = np.argmin(np.abs(xsDtiCurrSubj[t] - xsMriCurrSubj))
mriValsCorrespToDtiCurrSubj += [params['Y'][mriBiomksList[b]][s][mriIndClosestToCurrDtiScan]]
YMriClosestToDti[b][s] = np.array(mriValsCorrespToDtiCurrSubj)
# print(YMriClosestToDti[b][s].shape[0])
# print(YDti[b][s].shape[0])
assert YMriClosestToDti[b][s].shape[0] == YDti[b][s].shape[0]
# print(ads)
labelsDti = [params['labels'][b] for b in nonMriBiomksList]
metrics['labelsDti'] = labelsDti
# change diagnosis numbers to get different plotting behaviour (specific labels, colors and markers)
diagValidFiltLinModel = copy.deepcopy(diagValidFilt)
diagValidFiltLinModel[diagValidFiltLinModel == CTL2] = CTL_OTHER_MODEL
diagValidFiltLinModel[diagValidFiltLinModel == PCA] = PCA_OTHER_MODEL
diagValidFiltDktModel = copy.deepcopy(diagValidFilt)
diagValidFiltDktModel[diagValidFiltDktModel == CTL2] = CTL_DKT
diagValidFiltDktModel[diagValidFiltDktModel == PCA] = PCA_DKT
#plot just the trajectories by modality groups
for d in range(dpmObj.nrDis):
fig = dpmObj.plotter.plotTrajInDisSpaceOverlap(dpmObj, d, params, replaceFig=True)
figName = '%s/trajDisSpaceOverlap_%s_%s' % (params['outFolder'],
params['disLabels'][d], params['expName'])
fig.savefig('%s.png' % figName)
fig.savefig('%s.pdf' % figName)
print('Fig saved:%s.pdf ' % figName)
plotFigs = False
if plotFigs:
# for u in range(dpmObj.nrFuncUnits):
# trajStructUnitModel = dpmObj.unitModels[u].plotter.getTrajStructWithTrueParams(dpmObj.unitModels[u])
# fig = dpmObj.unitModels[u].plotter.plotTraj(dpmObj.unitModels[u], trajStructUnitModel,
# legendExtraPlot=True, rowsAuto=True)
# fig.savefig('%s/unit%d_allTraj.png' % (params['outFolder'], u))
# for d in range(dpmObj.nrDis):
# # yNormMode = dpmObj.params['plotTrajParams']['yNormMode']
# yNormMode = 'unscaled'
# trajStructDisModel = dpmObj.disModels[d].plotter.getTrajStructWithTrueParams(dpmObj.disModels[d], yNormMode)
# fig = dpmObj.disModels[d].plotter.plotAllTrajZeroOne(dpmObj.disModels[d], trajStructDisModel)
# fig.savefig('%s/dis%d_%s_allTrajZeroOne.png' % (params['outFolder'], d, dpmObj.params['disLabels'][d]))
# plot DTI over MRI space: traj, validation data, predictions of linear model, training data.
fig = dpmObj.plotter.plotTrajInBiomkSpace(dpmObj=dpmObj,
xsTrajXB=predTrajMriXB, predTrajXB=predTrajNonMriXB, trajSamplesBXS=trajSamplesNonMriBXS,
XsubjData1BSX=YvalidFiltMriClosestToNonMri, YsubjData1BSX=YvalidFiltNonMri, diagData1S=diagValidFilt,
XsubjData2BSX=YvalidFiltMriClosestToNonMri, YsubjData2BSX=YvalidLinModelNonMri, diagData2S=diagValidFiltLinModel,
XsubjData3BSX=YMriClosestToDti, YsubjData3BSX=YDti, diagData3S=params['diag'],
labels=labelsDti,
ssdDKT=mseDpm, ssdNoDKT=mseLin, replaceFig=True)
fig.savefig('%s/validTrajDtiOverMriPCA.png' % params['outFolder'])
# plot DTI over MRI space: DKT predictions, predictions of linear model, validation data.
fig = dpmObj.plotter.plotTrajInBiomkSpace(dpmObj=dpmObj,
xsTrajXB=None, predTrajXB=None, trajSamplesBXS=None,
XsubjData1BSX=YvalidFiltMriClosestToNonMri, YsubjData1BSX=YvalidFiltNonMri, diagData1S=diagValidFilt,
XsubjData2BSX=YvalidFiltMriClosestToNonMri, YsubjData2BSX=YvalidLinModelNonMri, diagData2S=diagValidFiltLinModel,
XsubjData3BSX=YvalidFiltMriClosestToNonMri, YsubjData3BSX=YvalidDktNonMri, diagData3S=diagValidFiltDktModel,
labels=labelsDti,
ssdDKT=None, ssdNoDKT=None, replaceFig=True)
fig.savefig('%s/validPredDtiOverMriPCA.png' % params['outFolder'])
# fig = dpmObj.plotterObj.plotTrajInDisSpace(xsTrajX, predTrajDtiXB, trajSamplesDtiBXS,
# XvalidShifDtiFilt, YvalidFiltDti, diagValidFilt,
# XvalidShifDtiFilt, YvalidLinModelNonMri, diagValidFiltLinModel,
# XsubjData3BSX=None, YsubjData3BSX=None, diagData3S=None,
# labelsDti, mseDpm, mseLin,
# replaceFig=False)
# fig.savefig('%s/validDtiPCA.png' % params['outFolder'])
return metrics