def getTransformedDataset(self, imageCollections, onlyPatchzone=True):
dataset = []
namesObservations = []
coordinates = []
for imageCollection in imageCollections:
imageCount = 0
for image in imageCollection:
im = image
imageName = imageCollection.files[imageCount]
imageWorker = RGBImageWorker(im, convert=True)
(binaryImageWorkerCenters, HEDAVWorker, BRVLWorker,
binaryImageWorker) = imageWorker.getBinaryImage()
workerPack = [imageWorker, BRVLWorker, HEDAVWorker]
generalStatistics = imageWorker.getGeneralStatistics(
hara=True, zern=False)
centers = binaryImageWorkerCenters.getCenters()
print len(centers)
for center in centers:
print "center"
binaryPatch = binaryImageWorker.getPatch(
center, self.patchSize)
binaryStatistics = binaryPatch.getGeneralStatistics()
namesObservations.append(imageName)
coordinates.append([center[0], center[1]])
observation = binaryStatistics
observation.extend(copy.deepcopy(generalStatistics))
if onlyPatchzone:
for worker in workerPack:
patch = worker.getPatch(center, self.patchSize)
observation.extend(
patch.getGeneralStatistics(
hara=True,
zern=True,
tamura=False,
only1D=patch.image[binaryPatch.image]))
observation.extend(
patch.getGeneralStatistics(hara=True,
zern=True,
tamura=True))
else:
for patchsize in range(1, self.patchSize, 10):
for worker in workerPack:
patch = worker.getPatch(center, patchsize)
observation.extend(
patch.getGeneralStatistics())
patch = imageWorker.getPatch(center, self.patchSize)
observation.extend(
patch.getGeneralStatistics(hara=True,
zern=True,
tamura=True))
print len(observation)
dataset.append(observation)
print imageName
imageCount += 1
dataset = np.array(dataset)
return (namesObservations, coordinates, dataset)
def saveImages(self):
for index in range(self.length):
image = self.getImageFromName(self.namesObservations[index])
imageWorker = RGBImageWorker(image, self.rows, self.columns)
patch = imageWorker.getPatch(self.coordinates[index], self.patchSize)
reducedFileName = "%s_%d.tif" % (Utils.getFolderName(self.namesObservations[index]), index)
if self.targetVector is None:
fileName = os.path.join(data_io.get_testing_folder(), data_io.get_test_folder(), reducedFileName)
elif self.targetVector[index] == 1:
fileName = os.path.join(data_io.get_training_folder(), data_io.get_positive_folder(), reducedFileName)
else:
fileName = os.path.join(data_io.get_training_folder(), data_io.get_negative_folder(), reducedFileName)
io.imsave(fileName, patch.image)
def getTransformedDatasetChecking(self, imageCollections):
dataset = []
namesObservations = []
coordinates = []
totalMitoticPoints = 0
for imageCollection in imageCollections:
imageCount = 0
missedCount = 0
totalCount = 0
for image in imageCollection:
im = image
imageName = imageCollection.files[imageCount]
if not "10" in imageName:
imageCount += 1
continue
imageWorker = RGBImageWorker(im, convert=True)
(binaryImageWorkerCenters, _, _,
binaryImageWorker) = imageWorker.getBinaryImage()
centers = binaryImageWorkerCenters.getCenters()
count = 0
(incMissedCount, incTotalCount,
incTotalMitoticPoints) = self.checkMissedCount(
imageName, centers)
totalCount += incTotalCount
totalMitoticPoints += incTotalMitoticPoints
missedCount += incMissedCount
for center in centers:
binaryPatch = binaryImageWorker.getPatch(
center, self.patchSize)
binaryStatistics = binaryPatch.getGeneralStatistics()
count += 1
namesObservations.append(imageName)
coordinates.append([center[0], center[1]])
observation = binaryStatistics
dataset.append(observation)
print "Candidates: %d" % (count)
print "Missed this patient: %d" % (missedCount)
if totalCount != 0:
print "Missed ratio this patient: %f" % (
(missedCount + 0.0) / (totalCount + 0.0))
print "Total number of candidates: %d" % (len(dataset))
print "Total number of mitotic points: %d" % totalMitoticPoints
print "Total ratio mitotic/candidate: %f" % (
(0.0 + totalMitoticPoints) / (0.0 + len(dataset)))
print imageName
imageCount += 1
return (np.array(namesObservations), np.array(coordinates),
np.array(dataset))
def getTransformedDataset(self, imageCollections, onlyPatchzone=True):
dataset = []
namesObservations = []
coordinates = []
for imageCollection in imageCollections:
imageCount = 0
for image in imageCollection:
im = image
imageName = imageCollection.files[imageCount]
imageWorker = RGBImageWorker(im, convert=True)
(binaryImageWorkerCenters, HEDAVWorker, BRVLWorker, binaryImageWorker) = imageWorker.getBinaryImage()
workerPack = [imageWorker, BRVLWorker, HEDAVWorker]
generalStatistics = imageWorker.getGeneralStatistics(hara=True, zern=False)
centers = binaryImageWorkerCenters.getCenters()
print len(centers)
for center in centers:
print "center"
binaryPatch = binaryImageWorker.getPatch(center, self.patchSize)
binaryStatistics = binaryPatch.getGeneralStatistics()
namesObservations.append(imageName)
coordinates.append([center[0], center[1]])
observation = binaryStatistics
observation.extend(copy.deepcopy(generalStatistics))
if onlyPatchzone:
for worker in workerPack:
patch = worker.getPatch(center, self.patchSize)
observation.extend(patch.getGeneralStatistics(hara=True, zern=True, tamura=False, only1D=patch.image[binaryPatch.image]))
observation.extend(patch.getGeneralStatistics(hara=True, zern=True, tamura=True))
else:
for patchsize in range(1, self.patchSize, 10):
for worker in workerPack:
patch = worker.getPatch(center, patchsize)
observation.extend(patch.getGeneralStatistics())
patch = imageWorker.getPatch(center, self.patchSize)
observation.extend(patch.getGeneralStatistics(hara=True, zern=True, tamura=True))
print len(observation)
dataset.append(observation)
print imageName
imageCount += 1
dataset = np.array(dataset)
return (namesObservations, coordinates, dataset)
def getTransformedDatasetChecking(self, imageCollections):
dataset = []
namesObservations = []
coordinates = []
totalMitoticPoints = 0
for imageCollection in imageCollections:
imageCount = 0
missedCount = 0
totalCount = 0
for image in imageCollection:
im = image
imageName = imageCollection.files[imageCount]
if not "10" in imageName:
imageCount += 1
continue
imageWorker = RGBImageWorker(im, convert=True)
(binaryImageWorkerCenters, _, _, binaryImageWorker) = imageWorker.getBinaryImage()
centers = binaryImageWorkerCenters.getCenters()
count = 0
(incMissedCount, incTotalCount, incTotalMitoticPoints) = self.checkMissedCount(imageName, centers)
totalCount += incTotalCount
totalMitoticPoints += incTotalMitoticPoints
missedCount += incMissedCount
for center in centers:
binaryPatch = binaryImageWorker.getPatch(center, self.patchSize)
binaryStatistics = binaryPatch.getGeneralStatistics()
count += 1
namesObservations.append(imageName)
coordinates.append([center[0], center[1]])
observation = binaryStatistics
dataset.append(observation)
print "Candidates: %d" % (count)
print "Missed this patient: %d" % (missedCount)
if totalCount != 0:
print "Missed ratio this patient: %f" % ((missedCount + 0.0) / (totalCount + 0.0))
print "Total number of candidates: %d" % (len(dataset))
print "Total number of mitotic points: %d" % totalMitoticPoints
print "Total ratio mitotic/candidate: %f" % ((0.0 + totalMitoticPoints) / (0.0 + len(dataset)))
print imageName
imageCount += 1
return (np.array(namesObservations), np.array(coordinates), np.array(dataset))
def saveImages(self):
for index in range(self.length):
image = self.getImageFromName(self.namesObservations[index])
imageWorker = RGBImageWorker(image, self.rows, self.columns)
patch = imageWorker.getPatch(self.coordinates[index],
self.patchSize)
reducedFileName = "%s_%d.tif" % (Utils.getFolderName(
self.namesObservations[index]), index)
if self.targetVector is None:
fileName = os.path.join(data_io.get_testing_folder(),
data_io.get_test_folder(),
reducedFileName)
elif self.targetVector[index] == 1:
fileName = os.path.join(data_io.get_training_folder(),
data_io.get_positive_folder(),
reducedFileName)
else:
fileName = os.path.join(data_io.get_training_folder(),
data_io.get_negative_folder(),
reducedFileName)
io.imsave(fileName, patch.image)
