def blob_from_hist(data, show=False, show_now=True, min_int=0, max_int=255):
data = tools.windowing(data, sliceId=0)
data = tools.resize_ND(data, scale=0.1).astype(np.uint8)
data = tools.smoothing(data, sliceId=0)
min_interval = int((0 + 125) / 350. * 255)
max_interval = int((100 + 125) / 350. * 255)
mask = np.ones_like(data)
mask = np.where(data < 5, 0, mask)
mask = np.where(data > 250, 0, mask)
pts = data[np.nonzero(mask)]
hist, bins = skiexp.histogram(pts)
hist = tools.hist_smoothing(bins, hist)
# bins = bins / 255 * 350 - 125
hist_rec = np.zeros_like(hist)
hist_rec[min_interval:max_interval] = hist[min_interval:max_interval]
peak_idx = np.argmax(hist_rec)
hist_ind = bins[peak_idx]
hist_hu = hist_ind / 255 * 350 - 125
print 'hist ind: %.1f = %i HU' % (hist_ind, int(hist_hu))
if show:
plt.figure()
plt.plot(bins, hist, 'b-', lw=4)
plt.plot((min_interval, min_interval), (0, 1.2 * hist_rec.max()),
'm-',
lw=4)
plt.plot((max_interval, max_interval), (0, 1.2 * hist_rec.max()),
'm-',
lw=4)
plt.plot(hist_ind, hist[peak_idx], 'go', markersize=12)
img = data[21, ...]
hist_diff = abs(img.astype(np.int64) - hist_ind)
plt.figure()
plt.subplot(121), plt.imshow(img, 'gray',
interpolation='nearest'), plt.axis('off')
plt.subplot(122), plt.imshow(hist_diff,
'gray',
interpolation='nearest'), plt.axis('off')
if show_now:
plt.show()
return hist_ind
def morph_hat_test(im):
im = tools.windowing(im)
im = tools.smoothing(im)
min_size = 11
max_size = 31
strels = []
for i in range(min_size, max_size + 1, 6):
strels.append(cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (i, i)))
tophats = tools.morph_hat(im, strels, show=True, show_now=False)
min_size = 41
max_size = 61
strels = []
for i in range(min_size, max_size + 1, 6):
strels.append(cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (i, i)))
blackhats = tools.morph_hat(im,
strels,
type='blackhat',
show=True,
show_now=True)
# hyper, 1 on the border -------------------
# arterial 0.6mm - not that bad
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_239_61293268_DE_Art_Abd_0.75_I26f_M_0.5-.pklz'
# venous 5mm - bad
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_239_61293268_DE_Ven_Abd_0.75_I26f_M_0.5-.pklz'
# shluk -----------------
# arterial 5mm
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_180_49509315_arterial_5.0_B30f-.pklz'
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_180_49509315_venous_5.0_B30f-.pklz'
# targeted
# arterial 0.6mm - bad
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_238_54280551_Abd_Arterial_0.75_I26f_3-.pklz'
# venous 0.6mm - bad
# fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_238_54280551_Abd_Venous_0.75_I26f_3-.pklz'
debug = True
verbose = True
data_fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_183_46324212_venous_5.0_B30f-.pklz'
data, mask, voxel_size = tools.load_pickle_data(data_fname)
slice_ind = 17
data_s = data[slice_ind, :, :]
data_s = tools.windowing(data_s)
mask_s = mask[slice_ind, :, :]
# data_s = tools.windowing(data)
# mask_s = mask
run(data_s, mask=mask_s, smoothing=True, save_fig=False, show=True, verbose=verbose)
plt.imshow(results[i], interpolation='nearest'), plt.title('result - %s' % t**s[i])
figname = 'unary_saliencies_alpha_%i_rad_%i.png' % (alpha, rad)
fig.savefig(os.path.join(figdir, figname), dpi=100, bbox_inches='tight', pad_inches=0)
plt.show()
#-----------------------------------------------------------------------------------------
#-----------------------------------------------------------------------------------------
if __name__ == '__main__':
data_fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_183_46324212_venous_5.0_B30f-.pklz'
data, mask, voxel_size = tools.load_pickle_data(data_fname)
slice_ind = 17
data_s = data[slice_ind, :, :]
data_s = tools.windowing(data_s)
mask_s = mask[slice_ind, :, :]
show = False
show_now = False
save_fig = True
run(data_s, mask_s, alpha=50, show=show, show_now=show_now, save_fig=save_fig)
# im_o, img, saliency = run(data_s, mask_s, show=False)
# im_o_s, img_s, saliency_s = run(data_s, mask_s, smoothing=True, show=False)
#
# plt.figure()
# plt.subplot(231), plt.imshow(im_o, 'gray', interpolation='nearest'), plt.title('input')
# plt.subplot(233), plt.imshow(saliency, 'gray', interpolation='nearest'), plt.title('saliency map')
#
dist_out = np.absolute(mean_out - int_new)
return dist_in, dist_out
#---------------------------------------------------------------------------------------------
#---------------------------------------------------------------------------------------------
if __name__ == '__main__':
data_fname = '/home/tomas/Data/medical/liver_segmentation/org-exp_183_46324212_venous_5.0_B30f-.pklz'
data, mask, voxel_size = tools.load_pickle_data(data_fname)
# 2D
# slice_ind = 17
slice_ind = 15
# data = data[slice_ind, :, :]
data = tools.windowing(data)
ball_r_out = 1
ball_r = 5
min_diff = 1
alpha = 0.9
scale = 0.5
max_iters = 100
smoothing = True
show = True
show_now = False
_, seeds = tools.initialize_graycom(data, slice=slice_ind, show=show, show_now=show_now)
# seeds = tools.eroding3D(seeds, np.ones((3, 3))).astype(np.uint8)
seeds = tools.morph_ND(seeds, 'erosion', selem=np.ones((3, 3)), slicewise=True, sliceId=0).astype(np.uint8)
def blob_from_glcm(data, show=False, show_now=True, min_int=0, max_int=255):
data = tools.windowing(data, sliceId=0)
data_f = tools.resize_ND(data, scale=0.1).astype(np.uint8)
data_f = tools.smoothing(data_f, sliceId=0)
min_interval = int((0 + 125) / 350. * 255)
max_interval = int((100 + 125) / 350. * 255)
# min_int = 20
# max_int = 80
print 'calculating glcm ...',
glcm = tools.graycomatrix_3D(data_f)
print 'done'
glcm_rec = glcm.copy()
mask = np.zeros_like(glcm_rec)
mask[min_interval:max_interval, min_interval:max_interval] = 1
glcm_rec *= mask
max_r, max_c = np.unravel_index(np.argmax(glcm_rec), glcm_rec.shape)
glcm_ind = (max_r + max_c) / 2
glcm_hu = glcm_ind / 255 * 350 - 125
print 'glcm ind: %.1f = %i HU' % (glcm_ind, int(glcm_hu))
img = data[21, ...]
if show:
plt.figure()
plt.subplot(121), plt.imshow(glcm,
interpolation='nearest',
vmax=5 * glcm.mean())
plt.subplot(122), plt.imshow(glcm,
interpolation='nearest',
vmax=5 * glcm.mean())
plt.plot((min_interval, min_interval), (min_interval, max_interval),
'm-',
lw=5)
plt.plot((min_interval, max_interval), (max_interval, max_interval),
'm-',
lw=5)
plt.plot((max_interval, max_interval), (max_interval, min_interval),
'm-',
lw=5)
plt.plot((max_interval, min_interval), (min_interval, min_interval),
'm-',
lw=5)
plt.plot(max_c, max_r, 'wo', markersize=12)
plt.axis('image')
glcm_diff = abs(img.astype(np.int64) - glcm_ind)
plt.figure()
plt.subplot(121), plt.imshow(img, 'gray',
interpolation='nearest'), plt.axis('off')
plt.subplot(122), plt.imshow(glcm_diff,
'gray',
interpolation='nearest'), plt.axis('off')
if show_now:
plt.show()
return glcm_ind
def segmentation_stats(data_struc, scale=1, smoothing=True):
# fnames = load_data(datadir)
# fnames = ['/home/tomas/Dropbox/Data/medical/dataset/180_arterial-.pklz',]
# creating workbook
workbook = xlsxwriter.Workbook('/home/tomas/Dropbox/Data/liver_segmentation/final_alg/liver_seg_stats.xlsx')
worksheet = workbook.add_worksheet()
# Add a bold format to use to highlight cells.
bold = workbook.add_format({'bold': True})
# green = workbook.add_format({'bg_color': '#C6EFCE'})
col_gc_fmea = 3
col_gc_prec = 6
col_gc_rec = 9
col_ls_fmea = 4
col_ls_prec = 7
col_ls_rec = 10
cols_gc = (3, 6, 9)
cols_ls = (4, 7, 10)
start_row = 2
worksheet.write(0, 0, 'DATA', bold)
worksheet.write(0, 1, 'slice', bold)
worksheet.write(0, 3, 'F-MEASURE', bold)
worksheet.write(0, 6, 'PRECISION', bold)
worksheet.write(0, 9, 'RECALL', bold)
worksheet.write(1, col_gc_fmea, 'Grow Cut', bold)
worksheet.write(1, col_ls_fmea, 'Level Sets', bold)
worksheet.write(1, col_gc_prec, 'Grow Cut', bold)
worksheet.write(1, col_ls_prec, 'Level Sets', bold)
worksheet.write(1, col_gc_rec, 'Grow Cut', bold)
worksheet.write(1, col_ls_rec, 'Level Sets', bold)
# loading data
data_fnames = []
data_slices = []
for name, num in data_struc:
for i in num:
data_fnames.append(name)
data_slices.append(i)
n_imgs = len(data_fnames)
for i, (data_fname, slice) in enumerate(zip(data_fnames, data_slices)):
fname = data_fname.split('/')[-1].split('.')[0]
print '#%i/%i, %s [%i]: ' % (i + 1, n_imgs, fname, slice),
data, gt_mask, voxel_size = tools.load_pickle_data(data_fname)
data = data[slice,...]
gt_mask = gt_mask[slice,...]
print 'windowing ...',
data = tools.windowing(data)
shape_orig = data.shape
data_o = data.copy()
# plt.figure()
# plt.imshow(data, 'gray', vmin=0, vmax=255, interpolation='nearest')
# plt.show()
if scale != 1:
print 'resizing ...',
data = tools.resize_ND(data, scale=scale).astype(np.uint8)
if smoothing:
print 'smoothing ...',
data = tools.smoothing(data)
print 'gc ...',
blob, seg_gc = coarse_segmentation(data, clear_border=False, show=False, verbose=False)
print 'ls ...',
seg_ls = segmentation_refinement(data, blob, show=False, verbose=False)
if seg_gc.shape != gt_mask.shape:
seg_gc = tools.resize_ND(seg_gc, shape=shape_orig)
if seg_ls.shape != gt_mask.shape:
seg_ls = tools.resize_ND(seg_ls, shape=shape_orig)
# computing statistics
print 'stats ...',
precision_gc, recall_gc, f_measure_gc = ac.calc_statistics(blob, gt_mask)
precision_ls, recall_ls, f_measure_ls = ac.calc_statistics(seg_ls, gt_mask)
# writing statistics
print 'writing ...',
items = (f_measure_gc, precision_gc, recall_gc, f_measure_ls, precision_ls, recall_ls)
worksheet.write(start_row + i, 0, fname)
worksheet.write(start_row + i, 1, slice)
for it, col in zip(items, cols_gc + cols_ls):
worksheet.write(start_row + i, col, it)
print 'done'
fig = tools.visualize_seg(data_o, seg_ls, mask=seg_gc, title=fname, show_now=False, for_save=True)
out_fname = '/home/tomas/Dropbox/Data/liver_segmentation/final_alg/%s_sl_%i.png' % (fname, slice)
fig.savefig(out_fname)
plt.close('all')
print '\t\tfmea = %.3f -> %.3f, prec = %.3f -> %.3f, rec = %.3f -> %.3f' %\
(f_measure_gc, f_measure_ls, precision_gc, precision_ls, recall_gc, recall_ls)
def register_dataset():
dirpath = '/home/tomas/Dropbox/Data/medical/dataset/gt/'
data_struc = [('180_venous-GT.pklz', '180_arterial-GT.pklz'),
('183a_venous-GT.pklz', '183a_arterial-GT.pklz'),
('185a_venous-GT.pklz', '185a_arterial-GT.pklz'),
('186a_venous-GT.pklz', '186a_arterial-GT.pklz'),
('189a_venous-GT.pklz', '189a_arterial-GT.pklz'),
('221_venous-GT.pklz', '221_arterial-GT.pklz'),
('222a_venous-GT.pklz', '222a_arterial-GT.pklz'),
('232_venous-GT.pklz', '232_arterial-GT.pklz'),
# ('234_venous-GT.pklz', '234_arterial-GT.pklz', range(0, 0, s)),
('235_venous-GT.pklz', '235_arterial-GT.pklz')]
data_struc = [(dirpath + x[0], dirpath + x[1]) for x in data_struc]
n_data = len(data_struc)
for i, (fixed_fn, moving_fn) in enumerate(data_struc):
if i < 5:
continue
print 'data #%i/%i: ' % (i + 1, n_data),
data_v, gt_mask, voxel_size = tools.load_pickle_data(fixed_fn)
data_v = tools.windowing(data_v)
data_a, gt_mask, voxel_size = tools.load_pickle_data(moving_fn)
# data_a = tools.windowing(data_a)
moving_arr = tools.resize_ND(data_a, shape=data_v.shape).astype(data_a.dtype)
moving_arr = tools.match_size(moving_arr, data_v.shape, verbose=False)
data_ao = moving_arr.copy()
moving_arr = tools.windowing(moving_arr).astype(np.float)
# moving_arr = tools.windowing(data_a).astype(np.float)
fixed_arr = data_v.astype(np.float)
# moving_arr = tools.resize_ND(data_a, shape=data_v.shape).astype(fixed_arr.dtype)
# moving_arr = tools.match_size(moving_arr, fixed_arr.shape, verbose=False)
# print fixed_arr.shape, moving_arr.shape
# plt.figure()
# plt.subplot(121), plt.imshow(fixed_arr[15,...], 'gray')
# plt.subplot(122), plt.imshow(moving_arr[15,...], 'gray')
# plt.show()
moving_mask = tools.resize_ND(gt_mask, shape=data_v.shape).astype(gt_mask.dtype)
moving_mask = tools.match_size(moving_mask, data_v.shape)
n_slices = fixed_arr.shape[0]
print 'slices: %i - ' % n_slices,
data_reg = np.zeros_like(moving_arr)
mask_reg = np.zeros_like(moving_mask)
for j, (fixed, moving, mask) in enumerate(zip(fixed_arr, moving_arr, moving_mask)):
if (j % 5) == 0:
print ' ',
# print '#',
print j
fixed = sitk.GetImageFromArray(fixed)
moving = sitk.GetImageFromArray(moving)
# mask = sitk.GetImageFromArray(mask)
setups = (('simil', 'ms'), ('affine', 'ms'), ('spline', 'mi'))
for ttype, mtype in setups:
data_r, tx = register(fixed, moving, mask=mask, transform_type=ttype, metric=mtype, max_iters=300, return_resampler=True, show=False)
if (data_r == 0).sum() < (0.8 * np.prod(data_r.shape)):
print ttype, mtype
data_r = transform_data(data_ao[j,...], tx)
mask_r = transform_data(mask, tx)
break
data_reg[j, ...] = data_r
mask_reg[j,...] = mask_r
# plt.figure()
# plt.subplot(131), plt.imshow(fixed_arr[j,...], 'gray')
# plt.subplot(132), plt.imshow(moving_arr[j,...], 'gray')
# plt.subplot(133), plt.imshow(tools.windowing(data_r), 'gray')
# plt.show()
print ' - OK,',
datap_reg = {'data3d': data_reg, 'segmentation': mask_reg, 'voxelsize_mm': voxel_size, 'slab': {'none': 0, 'liver': 1, 'lesions': 6}}
outname = dirpath + 'registered/' + moving_fn.split('/')[-1].replace('.pklz', '-registered.pklz')
# if not os.path.exists(outname):
# os.mkdir(outname)
f = gzip.open(outname, 'wb')
pickle.dump(datap_reg, f)
f.close()
# break
print 'done'
def run(data,
mask,
voxel_size,
is_2D=False,
bboxing=False,
return_vis=False,
show=False,
show_now=True,
save_fig=False):
data_w = tools.windowing(data).astype(np.uint8)
data_s = tools.smoothing(data_w, sliceId=0) * (mask > 0)
data_rgb = np.zeros((data_s.shape[1], data_s.shape[2], 3, data_s.shape[0]))
for i, s in enumerate(data_s):
data_rgb[:, :, :, i] = cv2.cvtColor(s, cv2.COLOR_BAYER_GR2RGB)
data_rgb = np.swapaxes(data_rgb, 2, 3)
slics = skiseg.slic(data_rgb,
spacing=np.roll(voxel_size, 2),
multichannel=True,
n_segments=10)
# slics = skiseg.slic(data_rgb, spacing=np.roll(voxel_size, 2), multichannel=True, n_segments=100,
# slic_zero=True, enforce_connectivity=True)
slics = np.swapaxes(np.swapaxes(slics, 1, 2), 0, 1) * (mask > 0)
# tools.show_3d(slics)
slics_big = np.zeros_like(slics)
if is_2D:
areaT = 500
else:
areaT = 6000
for i in np.unique(slics):
slic = tools.opening3D(slics == i, selem=skimor.disk(3))
area = slic.sum()
if area > areaT:
slics_big = np.where(slic, slics, slics_big)
# tools.show_3d((data, slics, slics_big))
if not return_vis:
return slics_big
vis = []
for i, (im, slic, slic_b,
mask_s) in enumerate(zip(data_w, slics, slics_big, mask)):
vis_s = []
if bboxing:
im, mask_bb = tools.crop_to_bbox(im, mask_s)
slic, mask_bb = tools.crop_to_bbox(slic, mask_s)
slic_b, mask_bb = tools.crop_to_bbox(slic_b, mask_s)
vis_s.append(('input', im))
vis_s.append(('slic', slic))
vis_s.append(('thresholded slic', slic_b))
vis_s.append(
('boundaries', skiseg.mark_boundaries(im, slic_b,
color=(1, 0, 1))))
vis.append(vis_s)
# ranges for visualization
ranges = [(0, 255), (slics.min(), slics.max()),
(slics_big.min(), slics_big.max()), (0, 255)]
# colormaps
cmaps = ['gray', 'gray', 'jet', 'gray']
if show or save_fig:
n_imgs = len(vis[0])
# pbar = progressbar.ProgressBar(maxval=len(vis), widgets=[progressbar.Bar('=', '[', ']'), ' ', progressbar.Percentage()])
# widgets = ["Seting up figures: ", progressbar.Percentage(), " ", progressbar.Bar(), " ", progressbar.ETA()]
# pbar = progressbar.ProgressBar(maxval=len(vis), widgets=widgets)
# pbar.start()
for slice_id, (v, mask_s) in enumerate(zip(vis, mask)):
fig = plt.figure(figsize=(24, 14))
for i, ((tit, im), rg, cm) in enumerate(zip(v, ranges, cmaps)):
# tit = v[slice_id][i][0]
# im = v[slice_id][i][1]
# im_bb, mask_bb = tools.crop_to_bbox(im, mask_s)
plt.subplot(1, n_imgs, i + 1)
plt.imshow(im,
cmap=cm,
vmin=rg[0],
vmax=rg[1],
interpolation='nearest')
plt.title(tit)
if save_fig:
fig_dir = '/home/tomas/Dropbox/Work/Dizertace/figures/slics/'
dirs = fig_dir.split(os.sep)
for i in range(2, len(dirs)):
subdir = os.sep.join(dirs[:i])
if not os.path.exists(subdir):
os.mkdir(subdir)
if is_2D:
dirname = os.path.join(
fig_dir, 'slics_vis_2D_slice_%i.png' % (slice_id + 1))
else:
dirname = os.path.join(
fig_dir, 'slics_vis_slice_%i.png' % (slice_id + 1))
fig.savefig(dirname,
dpi=100,
bbox_inches='tight',
pad_inches=0)
plt.close('all')
# pbar.update(slice_id)
# pbar.finish()
if show_now:
plt.show()
return slics_big, vis, ranges, cmaps