def slicer(patientDir):
'''
Split NRRD CT Scan file into 64/64/64 disjoint cubes for inference
'''
multiple = 4
desired = np.multiply(IN_SIZE, multiple)
newshape = np.array(IN_SIZE)
cuber = Cubify(oldshape=desired, newshape=newshape)
sliceDir = patientDir + "sliced/"
if not os.path.exists(sliceDir):
os.mkdir(sliceDir)
patient = patientDir + "/orig.nrrd"
scan = sitk.ReadImage(patient)
scan = sitk.GetArrayFromImage(scan)
shape = np.array(scan.shape)
# crop the image to fit desired --- NOTE WE ARE LOSING INFO (EDGE) HERE BE CAREFUL
excess = np.abs(desired - shape)
excess1 = excess / 2
excess2 = excess - excess1
scan = scan[excess1[0]:-excess2[0], excess1[1]:-excess2[1],
excess1[2]:-excess2[2]]
scan = pad(scan, desired)
sitk.WriteImage(sitk.GetImageFromArray(scan), sliceDir + "orig.nrrd")
scan = cuber.cubify(scan)
nCubes = scan.shape[0]
for arrNo in range(nCubes):
wp = sliceDir + "sliced_{0}.bin".format(arrNo)
scan[arrNo].tofile(wp)
paths = glob.glob(sliceDir + "sliced_[0-9].bin")
paths = [os.path.abspath(path) for path in paths]
y = ["dummy.bin" for x in paths]
csv = pd.DataFrame({"x": paths, "y": y})
csv.to_csv(sliceDir + "csv.csv", index=0)
def grouper(patientDir):
multiple = 4
desired = np.multiply(IN_SIZE, multiple)
newshape = np.array(IN_SIZE)
cuber = Cubify(oldshape=desired, newshape=newshape)
patientDir += "sliced/"
nCubes = len(glob.glob(patientDir + "sliced_*_yPred.bin"))
scan = np.empty((nCubes, IN_SIZE[0], IN_SIZE[1], IN_SIZE[2]))
for i in xrange(nCubes):
path = patientDir + "sliced_{0}_yPred.bin".format(i)
img = np.fromfile(path, dtype=np.float32).reshape(IN_SIZE)
scan[i] = img
scan = np.array(scan)
scan = cuber.uncubify(scan)
mass = scan.sum()
mass.tofile(patientDir + "predictedMass.bin")
sitk.WriteImage(sitk.GetImageFromArray(scan),
patientDir + "predicted.nrrd")
# Original image
orig = sitk.ReadImage(patientDir + "../orig.nrrd")
orig = sitk.GetArrayFromImage(orig)
desired = np.array(orig.shape)
difference = desired - np.array(scan.shape)
difference = difference / 2
try:
padding = ((difference[0], difference[0]),
(difference[1], difference[1]), (difference[2],
difference[2]))
scanPad = np.pad(scan, padding, "constant")
sitk.WriteImage(sitk.GetImageFromArray(scanPad),
patientDir + "predictedPad.nrrd")
except ValueError:
print("Couldn't pad this one")
#
# Cubify Tutorial Part 2
#
# This tutorial shows you how to use CubeSets
#
from cubify import Cubify
import json
# Instantiate Cubify
cubify= Cubify()
#
# Do cleanup from previous runs of this tutorial
#
cubify.deleteCubeSet('purchasesCubeSet')
cubify.deleteCubeSet('purchasesCubeSet2')
# Create a cube set called 'purchasesCubeSet' (with automatic binning)
cubeSet = cubify.createCubeSet('tutorial', 'purchasesCubeSet', 'purchases.csv')
print ""
print "CubeSet purchasesCubeSet created successfully"
print ""
cubeRows = cubify.getSourceCubeRows(cubeSet)
binnedCubeRows = cubify.getBinnedCubeRows(cubeSet)
print ""
print "Cube rows in purchasesCubeSet's source cube:"
for cubeRow in cubeRows:
#
# This tutorial shows you how to use Cubify to:
#
# 1. Create a cube
# 2. Export a cube
# 3. Query cube rows
# 4. Add columns to a cube
# 5. Bin a cube
# 6. Aggregate a cube
#
from cubify import Cubify
import json
# Instantiate Cubify
cubify= Cubify()
#
# Do cleanup from previous runs of this tutorial
#
cubify.deleteCube('purchases')
cubify.deleteCube('purchases_autobinned_1')
cubify.deleteCube('purchases_autobinned_2')
cubify.deleteCube('purchases_binned_1')
cubify.deleteCube('purchases_binned_2')
cubify.deleteCube('purchases_binned_2_CustomerId')
cubify.deleteCube('purchases_binned_2_CustomerState-ProductCategory')
cubify.deleteCube('purchases_binned_2_ProductId')
cubify.deleteCube('purchases_binned_2_CustomerId-TransactionDate')
cubify.deleteCube('purchases_binned_2_agg1')
cubify.deleteCube('purchases_binned_2_agg2')
