def test_model_transform_single(eng):
regions = many([[[0, 0], [0, 1]]])
model = ExtractionModel(regions=regions)
im0 = [[0, 1], [1, 2]]
im1 = [[3, 4], [5, 6]]
im2 = [[7, 8], [9, 10]]
data = fromarray([im0, im1, im2], engine=eng)
transformed = model.transform(data)
assert allclose(transformed.toarray(), [[0.5, 3.5, 7.5]])
def filter_shape(model, min_diameter=7, max_diameter=13, min_eccentricity=0.2):
"""
Filter extraction model regions based on shape criterion.
Parameters
----------
model : ExtractionModel.
min_diameter : float, default 7.
Minimum allowed diameter of regions
max_diameter : float, default 13.
Maxium allowed diameter of regions
min_eccentricity : float, default 0.2.
Minimum allowed eccentricity of regions
"""
regions = []
for region in model.regions:
mask = region.mask(fill=[1, 1, 1], background=[0, 0, 0])[:, :, 0]
props = regionprops(mask.astype('int'))[0]
if (props.eccentricity > min_eccentricity) and (
props.equivalent_diameter >
min_diameter) and (props.equivalent_diameter < max_diameter):
regions.append(region)
return ExtractionModel(regions)
def fit(self, images):
# compute local correlation of each pixel after bluring a cell bodies worth
localcorr = images.localcorr(self.diameter)
# detect blobs in local correlations corresponding to size of a cell
centers = findcenters(localcorr[:, self.boundary[0]:-self.boundary[1]],
diameter=self.diameter,
clip_limit=self.clip_limit,
threshold=self.theshold)
centers = array([[x[0], x[1] + self.boundary[0]] for x in centers])
# reshape the data into blocks around each center coordinate
reshaped = images.map(
lambda x: selectCenters(x, centers, self.diameter))
# compute cross correlation with central pixels timeseries in each block
stack = centercorr(reshaped, sigma_blur=self.sigma_blur)
# detect boundaries of each putative cell and reshift coordinates back to original image
masks = [
detect(img,
dilationSize=2,
radial=True,
filterSize=int(5 * self.diameter)) for img in stack
]
regions = [
mask_to_coords(masks[ii], centers[ii])
for ii in range(len(centers))
]
return ExtractionModel(regions)
def neuropilModel(model, inner=3, outer=8, mask=True, shape=(512, 512)):
regionsNeuropil = model.regions.outline(inner, outer)
regionsNeuropil = regionsNeuropil.crop(zeros(len(shape)), shape)
if mask:
mask = model.regions.mask(fill=[1, 1, 1], base=zeros(shape))[:, :, 0]
regionsNeuropil = regionsNeuropil.exclude(mask)
return ExtractionModel(regionsNeuropil)
def test_load():
model = ExtractionModel.load('test/resources/regions.json')
assert allclose(model.regions.coordinates, [[[0, 1], [0, 2]], [[1, 2], [1, 3]]])
