figureSavePath = '/media/onerva/KINGSTON/test-data/outcome/random-vs-brainnetome.pdf'
equalSized = False
nIterations = 10
nROIs = 246
allVoxelTs = io.loadmat(allVoxelTsPath)['roi_voxel_data'][0]['roi_voxel_ts'][0]
imgdata = nib.load(imgdataPath).get_data()
groupMaskPath = '/media/onerva/KINGSTON/test-data/group_roi_mask-30-4mm_with_subcortl_and_cerebellum.nii'
groupMask = nib.load(groupMaskPath).get_data()
nTime = imgdata.shape[3]
for i in range(nTime):
imgdata[:,:,:,i] = imgdata[:,:,:,i] * groupMask
_,_,originalVoxelCoordinates,originalROIMaps = cbc.readROICentroids(ROIInfoFile,readVoxels=True,fixCentroids=True)
originalROIs = []
for ROIMap in originalROIMaps:
ROI = []
for voxel in ROIMap:
voxelIndex = np.where((originalVoxelCoordinates == voxel).all(axis=1)==1)[0][0]
if len(cbc.findNeighbors(voxel,allVoxels=originalVoxelCoordinates)) > 0:
ROI.append(voxelIndex)
originalROIs.append(ROI)
VIs = np.zeros(nIterations)
for i in range(nIterations):
ROILabels, voxelCoordinates, radius = cbc.growSphericalROIs('random', imgdata, nROIs=nROIs, template=groupMask, equalSized=equalSized)
#import pdb; pdb.set_trace()
randomROIs = [list(np.where(ROILabels == j)[0]) for j in range(nROIs)]
import nibabel as nib
import clustering_by_consistency as cbc
testSubjectFolders = ['/media/onerva/KINGSTON/test-data/010/',
'/media/onerva/KINGSTON/test-data/011/']
ROIVoxelTsFileName = 'roi_voxel_ts_all_rois4mm_FWHM0.mat'
ROIVoxelTsInfoFileName = 'roi_voxel_ts_all_rois4mm_WHM0_info.mat'
allVoxelTsPath = testSubjectFolders[0] + ROIVoxelTsFileName
ROIInfoFile = '/media/onerva/KINGSTON/test-data/group_roi_mask-30-4mm_with_subcortl_and_cerebellum.mat'
imgdataPath = testSubjectFolders[0] + 'epi_STD_mask_detrend_fullreg.nii'
niiSavePath = testSubjectFolders[0] + '/optimized-rois-test-for-Tarmo-mean-weighted-consistency-thresholded-voxel-neighborhood.nii'
pickleSavePath = testSubjectFolders[0] + '/optimized-rois-test-for-Tarmo-mean-weighted-consistency-thresholded-voxel-neighborhood.pkl'
ROICentroids,_,voxelCoordinates,_ = cbc.readROICentroids(ROIInfoFile,readVoxels=True,fixCentroids=True)
allVoxelTs = io.loadmat(allVoxelTsPath)['roi_voxel_data'][0]['roi_voxel_ts'][0]
imgdata = nib.load(imgdataPath).get_data()
groupMaskPath = '/media/onerva/KINGSTON/test-data/group_roi_mask-30-4mm_with_subcortl_and_cerebellum.nii'
groupMask = nib.load(groupMaskPath).get_data()
nTime = imgdata.shape[3]
for i in range(nTime):
imgdata[:,:,:,i] = imgdata[:,:,:,i] * groupMask
cfg = {}
cfg['ROICentroids'] = ROICentroids
cfg['voxelCoordinates'] = voxelCoordinates
cfg['names'] = ''
cfg['allVoxelTs'] = allVoxelTs
cfg['imgdata'] = imgdata
testUpdateROI = False
testGrowROIs = False
testFindROICentroids = False
testRemoveVoxel = False
testFindVoxelLabels = False
testFindROIBoundary = False
testVoxelLabelsToROIInfo = False
testVoxelsInClustersToROIInfo = False
testFindInROINeighbors = False
testIsBoundary = False
testGetKendallW = True
# testing defSphericalROIs
if testDefSphericalROIs:
ROIPath = 'atlases/brainnetome/brainnetome_thr25_4mm_rois.mat'
centroids, _, voxelCoords, _ = cbc.readROICentroids(ROIPath,
readVoxels=True)
# testcase 1: ROIs with radius < 1 should contain only ROI centroids
spheres = cbc.defineSphericalROIs(centroids, voxelCoords, 0)
sphereMaps = []
for sphereMap in spheres['ROIMaps']:
sphereMaps.append(sphereMap)
sphereMaps = np.array(sphereMaps)
diff = np.sum(np.abs(sphereMaps - centroids))
if diff == 0:
print("defSphericalROIs: testcase 1/3 OK")
# testcase 2: small example space
space = np.zeros((4, 4, 4))
for i, (clusteringMethod, tempalateNamesPerMethod, consistencySaveName) in enumerate(zip(clusteringMethods,templateNames,consistencySaveNames)):
for j, templateName in enumerate(templateNamesPerMethod):
for timeWindow in timeWindows:
consistencyPath = consistencySaveName + '_' + timeWindow + '.pkl'
f = open(consistencyPath, 'r')
data = pickle.load(f)
f.close()
if 'consistencies' in data.keys():
pooledConsistencies[i][j].extend(data['consistencies'])
else:
pooledConsistencies[i][j].extend(data['spatialConsistencies'])
if not excludeSizes:
if 'ROI_sizes' in data.keys():
ROISizes[i][j].extend(data['ROI_sizes'])
elif clusteringMethod == 'template': # TODO: this is a hack to be removed later
originalCentroids,_,originalVoxelCoordinates,originalROIMaps = cbc.readROICentroids('/media/onerva/KINGSTON/test-data/group_roi_mask-30-4mm_with_subcortl_and_cerebellum.mat',readVoxels=True,fixCentroids=True)
ROISizes[i][j].extend([len(ROIMap) for ROIMap in originalROIMaps])
else:
for i, (clusteringMethod, templateNamesPerMethod, netIdentificatorsPerMethod) in enumerate(zip(clusteringMethods,templateNames,netIdentificators)):
for j, (templateName, netIdentificatorsPerTemplate) in enumerate(zip(templateNamesPerMethod,netIdentificatorsPerMethod)):
savePath = subjectFolder + '/spatialConsistency_' + clusteringMethod + '_' + templateName + '.pkl'
f = open(savePath,'r')
spatialConsistencyData = pickle.load(f)
f.close()
ROISizes[i][j].extend(spatialConsistencyData['roi_sizes'])
pooledConsistencies[i][j].extend(spatialConsistencyData['spatial_consistencies'])
# calculating and visualizing consistency distributions
consFig = plt.figure(1)
consAx = consFig.add_subplot(111)
data,
statistic='count',
bins=nBins)
pdf = count / float(np.sum(count) * (binEdges[1] - binEdges[0]))
binCenters = 0.5 * (binEdges[:-1] + binEdges[1:])
return pdf, binCenters
fig = plt.figure(1)
ax1 = fig.add_subplot(111)
sizeFig = plt.figure(2)
sizeAx = sizeFig.add_subplot(111)
originalCentroids, _, originalVoxelCoordinates, originalROIMaps = cbc.readROICentroids(
originalROIInfoFile, readVoxels=True, fixCentroids=True)
centroidIndices = np.zeros(len(originalCentroids), dtype=int)
for i, originalCentroid in enumerate(originalCentroids):
centroidIndices[i] = np.where(
(originalVoxelCoordinates == originalCentroid).all(axis=1) == 1)[0][0]
for i, subject in enumerate(subjects):
optimizedROIInfoPath = subject + optimizedROIInfoFile + '.pkl'
tempPath = subject + optimizedROIInfoFile + '_temp.pkl'
allVoxelTsPath = subject + allVoxelTsFileName
allVoxelTs = io.loadmat(
allVoxelTsPath)['roi_voxel_data'][0]['roi_voxel_ts'][0]
f = open(optimizedROIInfoPath, "rb")
ROIInfo = pickle.load(f)
f.close()