from pathlib import Path
import numpy as np
from scipy.sparse import hstack
import copy
import scipy.signal
import json
import lbm_mc as mc
import lbm_caiman_python
[docs]
def combine_z_planes(results: dict):
"""
Combines all z-planes in the results dictionary into a single estimates object.
Parameters
----------
results (dict): Dictionary with estimates for each z-plane.
Returns
-------
estimates.Estimates: Combined estimates for all z-planes.
"""
from caiman.source_extraction.cnmf import estimates
keys = sorted(results.keys())
e_list = [results[k].estimates for k in keys]
# Initialize lists to collect components
A_list = []
b_list = []
C_list = []
f_list = []
R_list = []
for e in e_list:
A_list.append(e.A)
b_list.append(e.b)
C_list.append(e.C)
f_list.append(e.f)
R_list.append(e.R)
# Combine the components
A_new = hstack(A_list).tocsr()
b_new = np.concatenate(b_list, axis=0)
C_new = np.concatenate(C_list, axis=0)
f_new = np.concatenate(f_list, axis=0)
R_new = np.concatenate(R_list, axis=0)
# Assuming all z-planes have the same spatial dimensions
dims_new = e_list[0].dims # e.g., (height, width)
# Create new estimates object
e_new = estimates.Estimates(
A=A_new,
C=C_new,
b=b_new,
f=f_new,
R=R_new,
dims=dims_new
)
return e_new
def calculate_interplane_shifts(volume, n_planes, params, json_logger=None):
"""
"""
interplane_shifts = np.zeros((n_planes - 1, 2), dtype=int)
accumulated_shifts = np.zeros((n_planes - 1, 2), dtype=int)
for i_plane in range(n_planes - 1):
im1_copy = copy.deepcopy(volume[0, :, :, i_plane])
im2_copy = copy.deepcopy(volume[0, :, :, i_plane + 1])
# Removing NaNs
nonan_mask = np.stack(
(~np.isnan(im1_copy), ~np.isnan(im2_copy)), axis=0
)
nonan_mask = np.all(nonan_mask, axis=0)
coord_nonan_pixels = np.where(nonan_mask)
min_x, max_x = np.min(coord_nonan_pixels[0]), np.max(coord_nonan_pixels[0])
min_y, max_y = np.min(coord_nonan_pixels[1]), np.max(coord_nonan_pixels[1])
im1_nonan = im1_copy[min_x:max_x + 1, min_y:max_y + 1]
im2_nonan = im2_copy[min_x:max_x + 1, min_y:max_y + 1]
# Normalize intensities
im1_nonan -= np.min(im1_nonan)
im2_nonan -= np.min(im2_nonan)
# Cross-correlation
cross_corr_img = scipy.signal.fftconvolve(
im1_nonan, im2_nonan[::-1, ::-1], mode="same"
)
# Find correlation peak
corr_img_peak_x, corr_img_peak_y = np.unravel_index(
np.argmax(cross_corr_img), cross_corr_img.shape
)
self_corr_peak_x, self_corr_peak_y = [
dim / 2 for dim in cross_corr_img.shape
]
interplane_shift = [
corr_img_peak_x - self_corr_peak_x,
corr_img_peak_y - self_corr_peak_y,
]
interplane_shifts[i_plane] = copy.deepcopy(interplane_shift)
accumulated_shifts[i_plane] = np.sum(
interplane_shifts, axis=0, dtype=int
)
# Normalize shifts to start at zero
min_accumulated_shift = np.min(accumulated_shifts, axis=0)
for xy in range(2):
accumulated_shifts[:, xy] -= min_accumulated_shift[xy]
# Optional JSON logging
if params.get("json_logging") and json_logger:
json_logger.info(
json.dumps({"accumulated_shifts": accumulated_shifts.tolist()})
)
return accumulated_shifts
if __name__ == "__main__":
path = Path().home() / 'lbm_data'
files = lbm_caiman_python.get_files_ext(path, ".tiff", 3)
print(files)
