def extract_super_pixels(mat_adj=None,
test_left=None,
test_right=None,
mat_cat=None,
num_neighbors=8,
cor_choice='mean',
connectivity=None,
min_pixels=50,
image=None,
plot=False,
use_mean_image=False):
if image is None:
if mat_cat is None:
mat_cat = torch.cat([m[test_left:test_right] for m in mat_adj],
dim=0)
cor_global = neighbor_cor(mat_cat,
neighbors=num_neighbors,
plot=plot,
choice=cor_choice,
title='correlation map')
if use_mean_image:
cor_global = mat_cat.mean(0) * cor_global
cor_global = cor_global / cor_global.max()
image = cor_global.detach().cpu().numpy()
else:
cor_global = image # for backward compatibility
if isinstance(image, torch.Tensor):
image = image.detach().cpu().numpy()
label_image, regions = get_label_image(image,
min_pixels=min_pixels,
connectivity=connectivity,
plot=False)
if plot:
plot_image_label_overlay(image, label_image)
return cor_global, label_image, regions
def refine_one_label(submat, min_pixels=50, return_traces=False, percentile=50):
soft_attention = attention_map(submat)
label_image, regions = get_label_image(soft_attention, min_pixels=min_pixels)
if return_traces:
submats, traces = extract_traces(submat, softmask=soft_attention, label_image=label_image, regions=regions,
percentile=percentile)
return submats, traces, soft_attention, label_image, regions
else:
return label_image
def get_high_conf_mask(cor_map,
low_percentile=25,
high_percentile=5,
min_cor=0.1,
min_pixels=20,
exclude_boundary_width=2):
"""Return a high-confidence mask
"""
label_image, regions = get_label_image(cor_map, min_pixels=min_pixels)
fg_mask = ((cor_map >= np.percentile(cor_map.cpu(), high_percentile)) &
(cor_map.new_tensor(label_image) > 0) & (cor_map > min_cor))
bg_mask = ((cor_map <= np.percentile(cor_map.cpu(), low_percentile)) &
(cor_map.new_tensor(label_image) == 0) & (cor_map < min_cor))
mask = cor_map.new_full(cor_map.size(), -1)
mask[bg_mask] = 0
mask[fg_mask] = 1
mask[:exclude_boundary_width] = -1
mask[-exclude_boundary_width:] = -1
mask[:, :exclude_boundary_width] = -1
mask[:, -exclude_boundary_width:] = -1
return mask
def refine_segmentation(submat,
label_mask,
label_image,
label_idx,
plot=False,
figsize=(15, 10)):
num_pcs, u, s, v = svd(submat[:, label_mask.bool()], plot=plot)
if num_pcs >= 2:
A, B, loss_history = step_decompose(submat.reshape(submat.size(0), -1),
num_components=2 * num_pcs)
mask = B[0].reshape(submat.size(1), submat.size(2))
X = mask.reshape(-1).unsqueeze(-1)
kmeans = KMeans(n_clusters=2 * num_pcs,
random_state=0).fit(X.detach().cpu())
split_label_segmentation = kmeans.labels_.reshape(mask.shape)
label_image, regions = get_label_image(
image=None,
label_image=label_image,
split_label=label_idx - 1,
split_label_segmentation=split_label_segmentation,
plot=plot,
figsize=figsize)
def basic_segmentation(mat, min_thresh=0.05, min_pixels=50, select_frames=True, show=True, median_detrend=False,
fft=False, fft_max_freq=200):
"""Basic segmentation
Args:
mat: torch.Tensor with shape (nframe, nrow, ncol) or (n_experiments, nframe, nrow, ncol)
min_thresh: float, used by get_label_image
min_pixels: int, used by get_label_image
select_frames: default True, only used when mat.ndim==4, selecting only frames with average mean beyond an otsu threshold
Returns:
cor_map: torch.Tensor with shape (nrow, ncol)
label_image: torch.Tensor with shape (nrow, ncol); 0 is background; label==i is mask for label i for i >= 1
regions: object returned by regionprops
"""
dtype = mat.dtype
if dtype == torch.float16:
mat = mat.float()
if median_detrend:
mat = mat - get_local_median(mat, window_size=50, dim=-3)
if fft:
if mat.ndim == 3:
mat = torch.rfft(mat.transpose(0, 2), signal_ndim=1, normalized=True)[..., :fft_max_freq, :].reshape(
mat.size(2), mat.size(1), -1).transpose(0, 2)
elif mat.ndim == 4:
mat = torch.rfft(mat.transpose(1, 3), signal_ndim=1, normalized=True)[..., :fft_max_freq, :].reshape(
mat.size(0), mat.size(3), mat.size(2), -1).transpose(1, 3)
if mat.ndim == 3:
cor_map = get_cor_map(mat)
elif mat.ndim == 4:
cor_map = get_cor_map_4d(mat, select_frames=select_frames, top_cor_map_percentage=20, padding=2, topk=5, shift_times=[0, 1, 2],
return_all=False, plot=False)
label_image, regions = get_label_image(cor_map, min_thresh=min_thresh, min_pixels=min_pixels)
label_image = torch.from_numpy(label_image).to(mat.device)
if show:
imshow(cor_map)
plot_image_label_overlay(cor_map, label_image=label_image, regions=regions)
return cor_map, label_image, regions