def test_cdog(self):
im_nuclei_stain_data = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_nuclei_stain.npz'))
im_nuclei_stain = im_nuclei_stain_data['Easy1_nuclei_stain']
im_nuclei_fgnd_mask_data = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_nuclei_fgnd_mask.npz'))
im_nuclei_fgnd_mask = im_nuclei_fgnd_mask_data[
'Easy1_nuclei_fgnd_mask']
sigma_min = 10.0 / np.sqrt(2.0)
sigma_max = 40.0 / np.sqrt(2.0)
im_dog_max, im_sigma_max = cdog(im_nuclei_stain, im_nuclei_fgnd_mask,
sigma_min, sigma_max)
im_dog_max_gtruth_data = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_cdog_max.npz'))
im_dog_max_gtruth = im_dog_max_gtruth_data['Easy1_cdog_max']
assert_array_almost_equal_neighborhood_lines(im_dog_max,
im_dog_max_gtruth,
decimal=4)
compare_maxima(im_dog_max, im_dog_max_gtruth)
im_sigma_max_gtruth_data = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_cdog_sigma_max.npz'))
im_sigma_max_gtruth = im_sigma_max_gtruth_data['Easy1_cdog_sigma_max']
assert_array_almost_equal_neighborhood_lines(im_sigma_max,
im_sigma_max_gtruth,
decimal=4)
def test_cdog(self):
im_nuclei_stain = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_nuclei_stain.npy'))
im_nuclei_fgnd_mask = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_nuclei_fgnd_mask.npy'))
sigma_min = 10.0 / np.sqrt(2.0)
sigma_max = 40.0 / np.sqrt(2.0)
im_dog_max, im_sigma_max = cdog(im_nuclei_stain, im_nuclei_fgnd_mask,
sigma_min, sigma_max)
im_dog_max_gtruth = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_cdog_max.npy'))
np.testing.assert_array_almost_equal(im_dog_max.astype(np.float16),
im_dog_max_gtruth,
decimal=4)
im_sigma_max_gtruth = np.load(
os.path.join(TEST_DATA_DIR, 'Easy1_cdog_sigma_max.npy'))
np.testing.assert_array_almost_equal(im_sigma_max.astype(np.float16),
im_sigma_max_gtruth,
decimal=4)
def detect_nuclei_kofahi(im_nuclei_stain, args):
# segment nuclear foreground mask
# (assumes nuclei are darker on a bright background)
im_nuclei_fgnd_mask = im_nuclei_stain < args.foreground_threshold
# smooth foreground mask with closing and opening
im_nuclei_fgnd_mask = skimage.morphology.closing(
im_nuclei_fgnd_mask, skimage.morphology.disk(3))
im_nuclei_fgnd_mask = skimage.morphology.opening(
im_nuclei_fgnd_mask, skimage.morphology.disk(3))
im_nuclei_fgnd_mask = sp.ndimage.morphology.binary_fill_holes(
im_nuclei_fgnd_mask)
# run adaptive multi-scale LoG filter
im_log_max, im_sigma_max = htk_shape_filters.cdog(
im_nuclei_stain, im_nuclei_fgnd_mask,
sigma_min=args.min_radius / np.sqrt(2),
sigma_max=args.max_radius / np.sqrt(2)
)
# apply local maximum clustering
im_nuclei_seg_mask, seeds, maxima = htk_seg.nuclear.max_clustering(
im_log_max, im_nuclei_fgnd_mask, args.local_max_search_radius)
# split any objects with disconnected fragments
im_nuclei_seg_mask = htk_seg.label.split(im_nuclei_seg_mask, conn=8)
# filter out small objects
im_nuclei_seg_mask = htk_seg.label.area_open(
im_nuclei_seg_mask, args.min_nucleus_area).astype(np.int)
return im_nuclei_seg_mask
def detect_nuclei_kofahi(im_nuclei_stain, args):
# segment foreground (assumes nuclei are darker on a bright background)
im_nuclei_fgnd_mask = sp.ndimage.morphology.binary_fill_holes(
im_nuclei_stain < args.foreground_threshold)
# run adaptive multi-scale LoG filter
im_log_max, im_sigma_max = htk_shape_filters.cdog(
im_nuclei_stain,
im_nuclei_fgnd_mask,
sigma_min=args.min_radius / np.sqrt(2),
sigma_max=args.max_radius / np.sqrt(2))
# apply local maximum clustering
im_nuclei_seg_mask, seeds, maxima = htk_seg.nuclear.max_clustering(
im_log_max, im_nuclei_fgnd_mask, args.local_max_search_radius)
# filter out small objects
im_nuclei_seg_mask = htk_seg.label.area_open(
im_nuclei_seg_mask, args.min_nucleus_area).astype(np.int)
return im_nuclei_seg_mask
def detect_nuclei(im_input, min_radius=6, max_radius=10, display_result=False):
# color normalization
ref_mu_lab = (8.63234435, -0.11501964, 0.03868433)
ref_std_lab = (0.57506023, 0.10403329, 0.01364062)
im_nmzd = htk_cnorm.reinhard(im_input, ref_mu_lab, ref_std_lab)
# color deconvolution
w_est = htk_cdeconv.rgb_separate_stains_macenko_pca(im_nmzd, 255)
nuclear_chid = htk_cdeconv.find_stain_index(
htk_cdeconv.stain_color_map['hematoxylin'], w_est)
im_nuclei_stain = htk_cdeconv.color_deconvolution(im_nmzd, w_est,
255).Stains[:, :,
nuclear_chid]
# segment nuclei foreground
th = skimage.filters.threshold_li(im_nuclei_stain) * 0.8
# th = skimage.filters.threshold_otsu(im_nuclei_stain)
im_fgnd_mask = im_nuclei_stain < th
im_fgnd_mask = skimage.morphology.opening(im_fgnd_mask,
skimage.morphology.disk(2))
im_fgnd_mask = skimage.morphology.closing(im_fgnd_mask,
skimage.morphology.disk(1))
# detect nuclei
im_dog, im_dog_sigma = htk_shape_filters.cdog(
im_nuclei_stain,
im_fgnd_mask,
sigma_min=min_radius / np.sqrt(2),
sigma_max=max_radius / np.sqrt(2))
nuclei_coord = skimage.feature.peak_local_max(im_dog,
min_distance=min_radius / 2,
threshold_rel=0.1)
nuclei_coord = nuclei_coord[im_fgnd_mask[nuclei_coord[:, 0],
nuclei_coord[:, 1]], :]
nuclei_rad = np.array([
im_dog_sigma[nuclei_coord[i, 0], nuclei_coord[i, 1]] * np.sqrt(2)
for i in range(nuclei_coord.shape[0])
])
# display result
if display_result:
print 'Number of nuclei = ', nuclei_coord.shape[0]
plt.figure(figsize=(30, 20))
plt.subplot(2, 2, 1)
plt.imshow(im_input)
plt.title('Input', fontsize=labelsize)
plt.axis('off')
plt.subplot(2, 2, 2)
plt.imshow(im_nuclei_stain)
plt.title('Deconv nuclei stain', fontsize=labelsize)
plt.axis('off')
plt.subplot(2, 2, 3)
plt.imshow(im_fgnd_mask)
plt.title('Foreground mask', fontsize=labelsize)
plt.axis('off')
plt.subplot(2, 2, 4)
plt.imshow(im_nmzd)
plt.plot(nuclei_coord[:, 1], nuclei_coord[:, 0], 'k+')
for i in range(nuclei_coord.shape[0]):
cx = nuclei_coord[i, 1]
cy = nuclei_coord[i, 0]
r = nuclei_rad[i]
mcircle = mpatches.Circle((cx, cy), r, color='g', fill=False)
plt.gca().add_patch(mcircle)
plt.title('Nuclei detection', fontsize=labelsize)
plt.axis('off')
plt.tight_layout()
return nuclei_coord, nuclei_rad
def detect_nuclei_kofahi(im_nuclei_stain, im_nuclei_fgnd_mask, min_radius,
max_radius, min_nucleus_area,
local_max_search_radius):
"""Performs a nuclear segmentation using kofahi's method.
This method uses scale-adaptive multi-scale Laplacian-of-Gaussian filtering
for blob enhancement and a local maximum clustering for segmentation. The
original implementation described by Al-Kofahi et al. uses Laplacian of Gaussian
but this function replaces it with Difference of Gaussian to improve speed.
Parameters
----------
im_nuclei_stain : array_like
A hematoxylin intensity image obtained from ColorDeconvolution.
im_nuclei_fgnd_mask: array_like
A binary mask of the nuclear foreground typically obtained by applying
a threshold on the hematoxylin/nuclei stain image
min_radius : float
Minimum nuclear radius (used to set min sigma of the multiscale LoG filter)
max_radius : float
Maximum nuclear radius (used to set max sigma of the multiscale LoG filter)
min_nucleus_area : int
Minimum area that each nucleus should have
local_max_search_radius : float
Local max search radius used for detection seed points in nuclei
Returns
-------
im_nuclei_seg_mask : array_like
A 2D array mask of the nuclei segmentation.
References
----------
.. [#] Y. Al-Kofahi et al "Improved Automatic Detection and Segmentation of
Cell Nuclei in Histopathology Images" in IEEE Transactions on Biomedical
Engineering, Volume: 57, Issue: 4, doi: 10.1109/TBME.2009.2035102,
April 2010.
"""
# smooth foreground mask with closing and opening
im_nuclei_fgnd_mask = skimage.morphology.closing(
im_nuclei_fgnd_mask, skimage.morphology.disk(3))
im_nuclei_fgnd_mask = skimage.morphology.opening(
im_nuclei_fgnd_mask, skimage.morphology.disk(3))
im_nuclei_fgnd_mask = sp.ndimage.morphology.binary_fill_holes(
im_nuclei_fgnd_mask)
if not np.any(im_nuclei_fgnd_mask):
return im_nuclei_fgnd_mask
# run adaptive multi-scale LoG filter
im_log_max, im_sigma_max = htk_shape_filters.cdog(
im_nuclei_stain,
im_nuclei_fgnd_mask,
sigma_min=min_radius / np.sqrt(2),
sigma_max=max_radius / np.sqrt(2))
# apply local maximum clustering
im_nuclei_seg_mask, seeds, maxima = htk.segmentation.nuclear.max_clustering(
im_log_max, im_nuclei_fgnd_mask, local_max_search_radius)
if seeds is None:
return im_nuclei_seg_mask
# split any objects with disconnected fragments
im_nuclei_seg_mask = htk.segmentation.label.split(im_nuclei_seg_mask,
conn=8)
# filter out small objects
im_nuclei_seg_mask = htk.segmentation.label.area_open(
im_nuclei_seg_mask, min_nucleus_area).astype(np.int)
return im_nuclei_seg_mask