def match(a, c):
matched_ROIs1, matched_ROIs2, non_matched1, non_matched2, performance, A2 = register_ROIs(
A, a, dims, align_flag=False, thresh_cost=.7)
cor = [
pearsonr(c1, c2)[0]
for (c1,
c2) in np.transpose([C[matched_ROIs1], c[matched_ROIs2]], (1, 0,
2))
]
print(np.mean(cor), np.median(cor))
return matched_ROIs2[[list(matched_ROIs1).index(i) for i in ids]]
max_thr = 0.1
A_forw, assign_forw, match_forw = register_multisession(A, dims, Cns, max_thr=max_thr)
A_back, assign_back, match_back = register_multisession(A[::-1], dims, Cns[::-1], max_thr=max_thr)
#%% align all pairs separately
N = len(A)
pairs = list(itertools.combinations(range(N), 2))
pairs_matches = []
for pair in pairs:
match_1, match_2, non_1, non_, perf_, _ = register_ROIs(A[pair[0]], A[pair[1]], dims,
template1=Cns[pair[0]],
template2=Cns[pair[1]],
plot_results=False,
max_thr=max_thr)
pairs_matches.append(np.vstack((match_1, match_2)))
#%% read all pairing through the union
pairs_forw = []
pairs_back = []
assign_back_rev = assign_back[:,::-1]
for pair in pairs:
match_f = extract_active_components(assign_forw, list(pair), only=False)
match_f1 = assign_forw[list(match_f)][:,list(pair)]
pairs_forw.append(match_f1.astype(int).T)
match_b = extract_active_components(assign_back_rev, list(pair), only=False)
match_b1 = assign_back_rev[list(match_b)][:,list(pair)]
#%% normalize matrices
A1 = csc_matrix(A1 / A1.sum(0))
A2 = csc_matrix(A2 / A2.sum(0))
A3 = csc_matrix(A3 / A3.sum(0))
#%% match consecutive pairs
from caiman.base.rois import register_ROIs
import matplotlib.pyplot as plt
import matplotlib.lines as mlines
import numpy as np
dims = 512, 512
match1_12, match2_12, mis1_12, mis2_12, perf_12 = register_ROIs(
A1, A2, dims, plot_results=True, template1=data1['Cn'], template2=data2['Cn'])
plt.figure()
match2_23, match3_23, mis2_23, mis3_23, perf_23 = register_ROIs(
A2, A3, dims, plot_results=True, template1=data2['Cn'], template2=data3['Cn'])
#%%
match2_12 = list(match2_12)
match2_23 = list(match2_23)
# ROIs in session 2 that are registered against both session 1 and session 3
ind_2 = list(set(match2_12).intersection(match2_23))
ind_1 = [match1_12[match2_12.index(x)] for x in ind_2]
ind_3 = [match3_23[match2_23.index(x)] for x in ind_2]
#%% make figure
dims,
Cns[::-1],
max_thr=max_thr)
#%% align all pairs separately
N = len(A)
pairs = list(itertools.combinations(range(N), 2))
pairs_matches = []
for pair in pairs:
match_1, match_2, non_1, non_, perf_, _ = register_ROIs(
A[pair[0]],
A[pair[1]],
dims,
template1=Cns[pair[0]],
template2=Cns[pair[1]],
plot_results=False,
max_thr=max_thr)
pairs_matches.append(np.vstack((match_1, match_2)))
#%% read all pairing through the union
pairs_forw = []
pairs_back = []
assign_back_rev = assign_back[:, ::-1]
for pair in pairs:
match_f = extract_active_components(assign_forw, list(pair), only=False)
match_f1 = assign_forw[list(match_f)][:, list(pair)]
pairs_forw.append(match_f1.astype(int).T)
match_b = extract_active_components(assign_back_rev,
max_thr=0)
#%%
N = len(A)
trip_forw = extract_active_components(assignments, list(range(N)), only=False)
trip_back = extract_active_components(assignments_back,
list(range(N)),
only=True)
#%%
matched_ROIs1, matched_ROIs2, non_matched1, non_matched2, performan, _ = register_ROIs(
A_union,
A_back,
dims,
template1=Cns[-1],
template2=Cns[0],
plot_results=False,
thresh_cost=.8,
max_thr=0)
#%%
A_union_01, assignments_01, matchings_01 = register_multisession(A,
dims,
Cns,
max_thr=0.1)
A_back_01, assignments_back_01, matchings_back_01 = register_multisession(
A[::-1], dims, Cns[::-1], max_thr=0.1)
trip_forw_01 = extract_active_components(assignments_01,
list(range(N)),
for mm in masks_3[triplets[:,2]]]
# plt.legend(('Day1','Day2','Day3'))
plt.title('Matched components across multiple days')
plt.axis('off')
day1 = mlines.Line2D([], [], color='b', label='Day 1')
day2 = mlines.Line2D([], [], color='r', label='Day 2')
day3 = mlines.Line2D([], [], color='y', label='Day 3')
plt.legend(handles=[day1, day2, day3], loc=4)
plt.show()
#%%
dims = 512, 512
plt.figure()
match1_12, match2_12, mis1_12, mis2_12, perf_12, A2_12 = register_ROIs(
A1, A2, dims, plot_results = True, max_thr = 0.0, align_flag = True, template1=Cn1, template2=Cn2)
plt.figure()
match1b_12, match2b_12, mis1b_12, mis2b_12, perfb_12, A2b_12 = register_ROIs(
A1, A2, dims, plot_results=True, max_thr = 0.0, align_flag = True, template1=Cn1, template2=Cn2)
#plt.figure()
#%%
match2_23, match3_23, mis2_23, mis3_23, perf_23, _ = register_ROIs(
A3, A2, dims, plot_results=False, template1=Cn3, template2=Cn2)
match2b_23, match3b_23, mis2_23, mis3_23, perf_23, _ = register_ROIs(
A2, A3, dims, plot_results=False, template1=Cn2, template2=Cn3)
day1 = mlines.Line2D([], [], color='b', label='Day 1')
day2 = mlines.Line2D([], [], color='r', label='Day 2')
day3 = mlines.Line2D([], [], color='y', label='Day 3')
plt.legend(handles=[day1, day2, day3], loc=4)
plt.show()
#%%
dims = 512, 512
plt.figure()
match1_12, match2_12, mis1_12, mis2_12, perf_12, A2_12 = register_ROIs(
A1,
A2,
dims,
plot_results=True,
max_thr=0.0,
align_flag=True,
template1=Cn1,
template2=Cn2)
plt.figure()
match1b_12, match2b_12, mis1b_12, mis2b_12, perfb_12, A2b_12 = register_ROIs(
A1,
A2,
dims,
plot_results=True,
max_thr=0.0,
align_flag=True,
template1=Cn1,
template2=Cn2)
#cors[cell_i] = np.correlate(a, v)
cors[cell_i] = np.correlate(A1[:, cell_i], A2[:, cell_i])
return cors
pairs = list(itertools.combinations(range(len(templates)), 2))
stats = pd.DataFrame()
for pair in pairs:
match_1, match_2, non_1, non_2, perf_, A2 = register_ROIs(
spatial[pair[0]],
spatial[pair[1]],
dims,
template1=templates[pair[0]],
template2=templates[pair[1]],
plot_results=False,
max_thr=max_thr,
thresh_cost=thresh_cost,
max_dist=max_dist,
Cn=templates[pair[0]])
plt.suptitle(cnm_titles[pair[0]] + ' vs ' + cnm_titles[pair[1]])
my_roi_image(spatial[pair[0]].toarray(), A2, dims, match_1, match_2, non_1,
non_2, max_thr)
# Calculate centroid distances for the matched cells
cm_1 = com(spatial[pair[0]], dims[0], dims[1])[match_1]
cm_2 = com(spatial[pair[1]], dims[0], dims[1])[match_2]
cm_2_registered = com(A2, dims[0], dims[1])[match_2]
distances = [0] * len(cm_1)