def loadTest(yObject,yMatch,features):
"""
Method to load the test data into the object. We might be interested
in loading new test data into, so we have explicitly defined this is
as a method.
Parameters:
- - - - -
y : SubjectFeatures object for a test brain
yMatch : MatchingFeaturesTest object containing vertLib attribute
detailing which labels each vertex in surface y maps to
in the training data
"""
features = list(features.split(','))
loadFeatures = copy.copy(features)
loadFeatures = list(set(features).difference({'label'}))
# load test subject data, save as attribtues
tObject = ld.loadH5(yObject,*['full'])
ID = tObject.attrs['ID']
parsedData = ld.parseH5(tObject,loadFeatures)
tObject.close()
data = parsedData[ID]
mtd = cu.mergeFeatures(data,features)
threshed = ld.loadMat(yMatch)
ltvm = cu.vertexMemberships(threshed,180)
return [threshed,mtd,ltvm]
def loadTest(yObject, yMatch, features):
"""
Method to load the test data into the object. We might be interested
in loading new test data into, so we have explicitly defined this is
as a method.
Parameters:
- - - - -
y : SubjectFeatures object for a test brain
yMatch : MatchingFeaturesTest object containing vertLib attribute
detailing which labels each vertex in surface y maps to
in the training data
features : feature to included in the test data numpy array
these must be the same as the training data
"""
nf = []
for f in features:
if f != 'label':
nf.append(f)
print 'Train features: {}'.format(features)
print 'Test features: {}'.format(nf)
# load test subject data, save as attribtues
tObject = ld.loadH5(yObject, *['full'])
ID = tObject.attrs['ID']
parsedData = ld.parseH5(tObject, nf)
tObject.close()
data = parsedData[ID]
mtd = cu.mergeFeatures(data, nf)
threshed = ld.loadMat(yMatch)
ltvm = cu.vertexMemberships(threshed, 180)
return [threshed, mtd, ltvm]
testOutput = '{}{}.{}'.format(outDirIter, test_subj,
outputExt)
#print 'Test Output: {}'.format(testOutput)
if not os.path.isfile(testOutput):
testObject = '{}{}.{}.{}'.format(
testDir, test_subj, hExt, testExt)
testMids = '{}{}.{}.{}'.format(
midsDir, test_subj, hExt, midsExt)
#print 'Test Mids: {}'.format(testMids)
testMatch = '{}{}.{}.{}'.format(
matchDir, test_subj, hExt, matchExt)
tm = ld.loadMat(testMatch)
#print 'Test Match: {}'.format(testMatch)
mids = ld.loadMat(testMids) - 1
if fExt == '.p':
# If model was a random forest,current model is a LIST
# of models. We feed this in to malp.parallelPredictiong
# along with the test data
if classifier == 'GMM':
[mm, mtd,
ltvm] = loadTest(currentModel, testObject,
testMatch)
midDir = args.midlineDir
midExt = args.midlineExt
subjectList = args.subjectList
with open(subjectList,'r') as inSubj:
subjects = inSubj.readlines()
subjects = [x.strip() for x in subjects]
hemiMap = {'Left': 'L',
'Right': 'R'}
for subj in subjects:
for h in hemiMap.keys():
matchMatrix = ''.join([matchDir,subj,'.',hemiMap[h],'.',matchExt])
midline = ''.join([midDir,subj,'.',hemiMap[h],'.',midExt])
outFile = ''.join([predDir,subj,'.',hemiMap[h],'.',predExt])
if os.path.exists(matchMatrix) and os.path.exists(midline):
if not os.path.exists(outFile):
matching = ld.loadMat(matchMatrix)
midline = ld.loadMat(midline) - 1
prediction = np.argmax(matching,axis=1)+1
prediction[midline] = 0
wg.writeGiftiImage(prediction,outFile,''.join(['Cortex',h]))
#print 'Test Output: {}'.format(testOutput)
# Check to make sure current subject hasn't been run yet
# If it has, skip
if not os.path.isfile(testOutput):
testObject = '{}{}.{}.{}'.format(testDir,test_subj,hExt,testExt)
#print 'Test Object: {}'.format(testObject)
testMids = '{}{}.{}.{}'.format(midsDir,test_subj,hExt,midsExt)
#print 'Test Mids: {}'.format(testMids)
testMatch = '{}{}.{}.{}'.format(matchDir,test_subj,hExt,matchExt)
#print 'Test Match: {}'.format(testMatch)
mids = ld.loadMat(testMids)-1
if fExt == '.p':
# If model was a random forest,current model is a LIST
# of models. We feed this in to malp.parallelPredictiong
# along with the test data
if classifier == 'RandomForest':
P = parallelPredictRF(currentModel,
testObject,
testMatch,
testMids)
P[mids] = 0
myl.darrays[0].data = np.array(P).astype(np.float32)
outDest = '{}{}.{}.{}'.format(fullDir, s, hstr, troExt)
data = h5py.File(trainingObject, mode='r+')
arrays = data[data.keys()[0]]
if 'fs_central' in arrays.keys():
del (arrays['fs_central'])
if 'vertVar' in arrays.keys():
del (arrays['vertVar'])
if 'subcort' in arrays.keys():
arrays['fs_subcort'] = arrays.pop('subcort')
curv = ld.loadGii(curvObject)
fsSubCort = ld.loadMat(fsSubCortObject)
print fsSubCort.shape
ptxCort = ld.loadMat(ptxCortObject)
ptxSubCort = ld.loadMat(ptxSubCortObject)
if 'fs_subcort' in arrays.keys():
del (arrays['fs_subcort'])
arrays['fs_subcort'] = fsSubCort
if 'curv' in arrays.keys():
del (arrays['curv'])
arrays['curv'] = curv
if 'pt_cort' in arrays.keys():
del (arrays['pt_cort'])
arrays['pt_cort'] = np.log(ptxCort)
# Loop over test subjects
for test_subj in testList:
print 'Test Subject: {}'.format(test_subj)
# Load functional file for saving prediction later
inFunc = '{}MyelinDensity/{}.{}.MyelinMap.32k_fs_LR.func.gii'.format(
baseDir, test_subj, hm)
assert os.path.isfile(inFunc)
func = nb.load(inFunc)
# Load midline indices
mid = '{}Midlines/{}.{}.Midline_Indices.mat'.format(baseDir, test_subj, hm)
assert os.path.isfile(mid)
mid = ld.loadMat(mid) - 1
# Construct output file name
outExtension = ''.join(['.', hm, '.', oute])
outPrediction = ''.join([outd, test_subj, outExtension])
if not os.path.isfile(outPrediction):
# Scale test data, load matching frequencies, compute label-to-vertex mappings
[data, match, ltvm] = pcd.testing(P, test_subj, trDir=objd, trExt=obje)
# Compute prediction
[baseline, threshold, predicted] = nnu.predict(data,
match,
model,
power=pw)
