def ca_deconvolution(ddf_trace, l0=False):
""" perform calcium image deconvolution
This function performs several calcium image
deconvolution approaches. Deconvolutions currently
supported:
1. OASIS (https://github.com/j-friedrich/OASIS)
2. Event detection script from Peter
3. AR-FPOP
input:
ddf_trace: a 1d-numpy array of length n (the number of
time steps in the calcium trace)
output:
a dictionary whose keys are the deconvolution method
used and values are a 1d-numpy array of length n with
the estimated spikes
TODO:
Add functionality for the following methods
4. ML Spike
5. One of the supervised methods?
"""
out = {}
# Method 1 OASIS (https://github.com/j-friedrich/OASIS)
c, s, b, g, lam = deconvolve(np.double(ddf_trace), penalty=1)
out['OASIS'] = s
# Method 2 event detection
yes_array, size_array = ed.get_events_derivative(ddf_trace)
times_new, heights_new = ed.concatenate_adjacent_events(yes_array,
size_array,
delta=3)
tmp = np.zeros_like(ddf_trace)
tmp[times_new] = heights_new
out['event_detection'] = tmp
# Method 3 FastLZeroSpikeInference
if (l0):
import arfpop_ctypes as af
# Some default parameters that need to be tuned!
gam = 0.99
penalty = 0.25
constraint = False
ar_fit = af.arfpop(ddf_trace, gam, penalty, constraint)
out['arfpop'] = ar_fit['pos_spike_mag']
return out
def ca_deconvolution(ddf_trace):
""" perform calcium image deconvolution
This function performs several calcium image
deconvolution approaches. Deconvolutions currently
supported:
1. OASIS (https://github.com/j-friedrich/OASIS)
2. Event detection script from Peter
input:
ddf_trace: a 1d-numpy array of length n (the number of
time steps in the calcium trace)
output:
a dictionary whose keys are the deconvolution method
used and values are a 1d-numpy array of length n with
the estimated spikes
TODO:
Add functionality for the following methods
3. AR-FPOP
4. ML Spike
5. One of the supervised methods?
"""
out = {}
# Method 1 OASIS (https://github.com/j-friedrich/OASIS)
c, s, b, g, lam = deconvolve(np.double(ddf_trace), penalty=1)
out['OASIS'] = s
# Method 2 event detection
yes_array, size_array = ed.get_events_derivative(ddf_trace)
times_new, heights_new = ed.concatenate_adjacent_events(yes_array,
size_array,
delta=3)
tmp = np.zeros_like(ddf_trace)
tmp[times_new] = heights_new
out['event_detection'] = tmp
return out
def PlotAll(SaveNames,params):
from numpy import min, max, percentile,asarray,ceil,sqrt
import numpy as np
import sys
from scipy.signal import welch
from pylab import load
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from matplotlib.backends.backend_pdf import PdfPages
# from scipy.ndimage.measurements import label
from AuxilaryFunctions import GetRandColors, max_intensity,SuperVoxelize,GetData,PruneComponents,SplitComponents,ThresholdShapes,MergeComponents,ThresholdData,make_sure_path_exists
# from BlockLocalNMF_AuxilaryFunctions import HALS4activity
from mpl_toolkits.axes_grid1 import make_axes_locatable
# makse sure relevant folders exist, and add to path
Results_folder='Results/'
make_sure_path_exists(Results_folder)
OASIS_path='OASIS/'
make_sure_path_exists(OASIS_path)
sys.path.append(OASIS_path)
from functions import deconvolve
## plotting params
# what to plot
plot_activities=True
plot_activities_PSD=False
plot_shapes_projections=True
plot_shapes_slices=False
plot_activityCorrs=False
plot_clustered_shape=False
plot_residual_slices=False
plot_residual_projections=False
# videos to generate
video_shapes=False
video_residual=True
video_slices=False
# what to save
save_video=True
save_plot=True
close_figs=True#close all figs right after saving (to avoid memory overload)
# PostProcessing
Split=False
Threshold=False #threshold shapes in the end and keep only connected components
Prune=False # Remove "Bad" components (where bad is defined within SplitComponent fucntion)
Merge=True # Merge highly correlated nearby components
FineTune=False # SHould we fine tune activity after post-processing? (mainly after merging)
IncludeBackground=False #should we include the background as an extracted component?
# how to plot
detrend=True #should we detrend the data (remove background component)?
scale=2 #scale colormap to enhance colors
satuartion_percentile=96 #saturate colormap ont this percentile, when ma=percentile is used
dpi=200 #for videos
restrict_support=True #in shape video, zero out data outside support of shapes
C=4 #number of components to show in shape videos (if larger then number of shapes L, then we automatically set C=L)
color_map='gray' #'gnuplot'
frame_rate=10.0 #Hz
# Fetch experimental 3D data
data=GetData(params.data_name)
if params.SuperVoxelize==True:
data=SuperVoxelize(data)
if params.ThresholdData==True:
data=ThresholdData(data)
dims=np.shape(data)
if len(dims)<4:
plot_Shapes2D=False
video_residual_2D=False
if plot_shapes_projections or plot_shapes_slices:
plot_Shapes2D=True
if video_residual==True:
video_residual_2D=True
plot_shapes_projections=False
plot_shapes_slices=False
plot_activityCorrs=False
plot_clustered_shape=False
plot_residual_slices=False
plot_residual_projections=False
video_shapes=False
video_residual=False
video_slices=False
print('2D data, ignoring 3D plots/video options')
else:
plot_Shapes2D=False
video_residual_2D=False
min_dim=np.argmin(dims[1:])
denoised_data=0
detrended_data=data
for rep in range(len(SaveNames)):
resultsName=SaveNames[rep]
try:
results=load('NMF_Results/'+SaveNames[rep])
except IOError:
if rep==0:
print('results file not found!!')
else:
break
SS=results['shapes']
AA=results['activity']
if rep>=params.Background_num:
adaptBias=False
else:
adaptBias=True
if IncludeBackground==True:
adaptBias=False
L=len(AA)-adaptBias
if L==0: #Stop if we encounter a file with zero components
break
S=SS[:-adaptBias]
b=SS[L:(L+adaptBias)]
A=AA[:-adaptBias]
f=AA[L:(L+adaptBias)]
if rep==0:
shapes=S
activity=A
background_shapes=b
background_activity=f
else:
shapes=np.append(shapes,S,axis=0)
activity=np.append(activity,A,axis=0)
background_shapes=np.append(background_shapes,b,axis=0)
background_activity=np.append(background_activity,f,axis=0)
L=len(shapes)
adaptBias=0
if Split==True:
shapes,activity,L,all_local_max=SplitComponents(shapes,activity,adaptBias)
if Merge==True:
shapes,activity,L=MergeComponents(shapes,activity,L,threshold=0.7,sig=10)
if Prune==True:
# deleted_indices=[5,9,11,14,15,17,24]+range(25,36)
shapes,activity,L=PruneComponents(shapes,activity,L,params.TargetAreaRatio)
activity_NonNegative=np.copy(activity)
activity_NonNegative[activity_NonNegative<0]=0
activity_noisy=np.copy(activity_NonNegative)
if FineTune==True:
for ll in range(L):
activity[ll], spikes, baseline, g, lam = deconvolve(activity_NonNegative[ll],optimize_g=10,penalty=0)
# activity,background_activity,S,bl,c1,sn,g,junk = update_temporal_components(data.reshape((len(data),-1)).transpose(), shapes.reshape((len(shapes),-1)).transpose(), background_shapes.reshape((len(background_shapes),-1)).transpose(), activity,background_activity,**options['temporal_params'])
activity_noisy=np.copy(activity_NonNegative)
activity_NonNegative=activity
print(str(L)+' shapes detected')
detrended_data= detrended_data - background_activity.T.dot(background_shapes.reshape((len(background_shapes), -1))).reshape(dims)
if Threshold==True:
shapes=ThresholdShapes(shapes,adaptBias,[],MaxRatio=[])
if plot_residual_projections or video_shapes or video_residual or video_slices or video_residual_2D:
colors=GetRandColors(L)
color_shapes=np.transpose(shapes.reshape(L, -1,1)*colors,[1,0,2]) #weird transpose for tensor dot product next line
denoised_data = denoised_data + (activity_NonNegative.T.dot(color_shapes)).reshape(tuple(dims)+(3,))
residual = detrended_data - activity_NonNegative.T.dot(shapes.reshape(L, -1)).reshape(dims)
if detrend==True:
data=detrended_data
#%% After loading loop - Normalize (colored) denoised data
if plot_residual_projections or video_shapes or video_residual or video_slices or video_residual_2D:
# denoised_data=denoised_data/np.max(denoised_data)
denoised_data=old_div(denoised_data,np.percentile(denoised_data[denoised_data>0],99.5)) #%% normalize denoised data range
denoised_data[denoised_data>1]=1
# plt.close('all')
#%% plotting params
ComponentsInFig=20
left = 0.05 # the left side of the subplots of the figure
right = 0.95 # the right side of the subplots of the figure
bottom = 0.05 # the bottom of the subplots of the figure
top = 0.95 # the top of the subplots of the figure
wspace = 0.1 # the amount of width reserved for blank space between subplots
hspace = 0.12 # the amount of height reserved for white space between subplots
#%% ###### Plot Individual neurons' activities
index=0 #component display index
sz=np.min([ComponentsInFig,L+adaptBias])
# a=ceil(sqrt(sz))
# b=ceil(sz/a)
a=sz
b=1
if plot_activities:
pp = PdfPages(Results_folder + 'Activities'+resultsName+'.pdf')
for ii in range(L+adaptBias):
if index==0:
# fig0=plt.figure(figsize=(dims[1] , dims[2]))
fig0=plt.figure(figsize=(11,18))
ax = plt.subplot(a,b,index+1)
# dt=1/30 # 30 Hz sample rate
time=list(range(len(activity[ii])))
plt.plot(time,activity_noisy[ii],linewidth=0.5,c='r')
plt.plot(time,activity[ii],linewidth=3,c='b')
ma=np.max([np.max(activity[ii]),np.max(activity_noisy[ii])])
plt.setp(ax, xticks=[],yticks=[0,ma])
# component number
ax.text(0.02, 0.8, str(ii),
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='black',weight='bold', fontsize=13)
index+=1
if ((ii%ComponentsInFig)==(ComponentsInFig-1)) or ii==(L+adaptBias-1):
index=0
if save_plot==True:
plt.subplots_adjust(left*2, bottom, right, top, wspace, hspace*2)
pp.savefig(fig0)
pp.close()
if close_figs:
plt.close('all')
#%% Plot activities` PSDs
index=0 #component display index
sz=np.min([ComponentsInFig,L+adaptBias])
a=ceil(sqrt(sz))
b=ceil(old_div(sz,a))
if plot_activities_PSD:
pp = PdfPages(Results_folder + 'ActivityPSDs'+resultsName+'.pdf')
for ii in range(L+adaptBias):
if index==0:
# fig0=plt.figure(figsize=(dims[1] , dims[2]))
fig0=plt.figure(figsize=(11,18))
ax = plt.subplot(a,b,index+1)
ff, psd_activity = welch(activity[ii], nperseg=round(old_div(len(activity[ii]), 64)))
plt.plot(ff,psd_activity,linewidth=3)
plt.setp(ax, xticks=[],yticks=[0])
# component number
ax.text(0.02, 0.8, str(ii),
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='black',weight='bold', fontsize=13)
index+=1
if ((ii%ComponentsInFig)==(ComponentsInFig-1)) or ii==(L+adaptBias-1):
index=0
if save_plot==True:
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
pp.savefig(fig0)
pp.close()
if close_figs:
plt.close('all')
#%% 2D shapes
index=0 #component display index
sz=np.min([ComponentsInFig,L+adaptBias])
a=ceil(0.5*sqrt(sz))
b=ceil(old_div(sz,a))
if plot_Shapes2D:
if save_plot==True:
pp = PdfPages(Results_folder + 'Shapes2D_'+resultsName+'.pdf')
for ll in range(L+adaptBias):
if index==0:
fig=plt.figure(figsize=(18 , 11))
ax = plt.subplot(a,b,index+1)
temp=shapes[ll]
mi=0
try:
ma=np.percentile(temp[temp>0],satuartion_percentile)
except IndexError:
ma=0
im=plt.imshow(temp,vmin=mi,vmax=ma,cmap=color_map)
plt.setp(ax,xticks=[],yticks=[])
mn=int(np.floor(mi)) # colorbar min value
mx=int(np.ceil(ma)) # colorbar max value
md=old_div((mx-mn),2)
# divider = make_axes_locatable(ax)
# cax = divider.append_axes("right", size="5%", pad=0.05)
# cb=plt.colorbar(im,cax=cax)
# cb.set_ticks([mn,md,mx])
# cb.set_ticklabels([mn,md,mx])
# component number
ax.text(0.02, 0.8, str(ll),
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
#sparsity
spar_str=str(np.round(np.mean(shapes[ll]>0)*100,2))+'%'
ax.text(0.02, 0.02, spar_str,
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
#L^p
for p in range(2,6,2):
Lp=old_div((np.sum(shapes[ll]**p))**(old_div(1,float(p))),np.sum(shapes[ll]))
Lp_str=str(np.round(Lp*100,2))+'%' #'L'+str(p)+'='+
ax.text(0.02+p*0.2, 0.02, Lp_str,
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='yellow',weight='bold', fontsize=13)
index+=1
if (ll%ComponentsInFig==(ComponentsInFig-1)) or ll==L+adaptBias-1:
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
index=0
if save_plot==True:
pp.savefig(fig)
pp.close()
if close_figs:
plt.close('all')
#%% Re-write plot code from here, so that each figure has only ComponentsInFig components
#%% ###### Plot Individual neurons' area which is correlated with their activities
a=ceil(sqrt(L+adaptBias))
b=ceil(old_div((L+adaptBias),a))
if plot_activityCorrs:
if save_plot==True:
pp = PdfPages(Results_folder + 'CorrelationWithActivity'+resultsName+'.pdf')
for dd in range(len(shapes[0].shape)):
fig0=plt.figure(figsize=(11,18))
for ii in range(L+adaptBias):
ax = plt.subplot(a,b,ii+1)
corr_imag=old_div(np.dot(activity[ii],np.transpose(data,[1,2,0,3])),np.sqrt(np.sum(data**2,axis=0)*np.sum(activity[ii]**2)))
plt.imshow(np.abs(corr_imag).max(dd),cmap=color_map)
plt.setp(ax,xticks=[],yticks=[])
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
if save_plot==True:
pp.savefig(fig0)
pp.close()
if close_figs:
plt.close('all')
#%% All Shapes projections
a=ceil(sqrt(L+adaptBias))
b=ceil(old_div((L+adaptBias),a))
if plot_shapes_projections:
if save_plot==True:
pp = PdfPages(Results_folder + 'Shapes_projections'+resultsName+'.pdf')
for dd in range(len(shapes[0].shape)):
fig=plt.figure(figsize=(18 , 11))
for ll in range(L+adaptBias):
ax = plt.subplot(a,b,ll+1)
temp=shapes[ll].max(dd)
if dd==2:
temp=temp.T
mi=np.min(shapes[ll])
ma=np.max(shapes[ll])
im=plt.imshow(temp,vmin=mi,vmax=ma,cmap=color_map)
plt.setp(ax,xticks=[],yticks=[])
mn=int(np.floor(mi)) # colorbar min value
mx=int(np.ceil(ma)) # colorbar max value
md=old_div((mx-mn),2)
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.05)
cb=plt.colorbar(im,cax=cax)
# cb.set_ticks([mn,md,mx])
# cb.set_ticklabels([mn,md,mx])
# component number
ax.text(0.02, 0.8, str(ll),
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
# #sparsity
spar_str=str(np.round(np.mean(shapes[ll]>0)*100,2))+'%'
ax.text(0.02, 0.02, spar_str,
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
# #L^p
# for p in range(2,2,2):
# Lp=(np.sum(shapes[ll]**p))**(1/float(p))/np.sum(shapes[ll])
# Lp_str=str(np.round(Lp*100,2))+'%' #'L'+str(p)+'='+
# ax.text(0.02+p*0.2, 0.02, Lp_str,
# verticalalignment='bottom', horizontalalignment='left',
# transform=ax.transAxes,
# color='yellow',weight='bold', fontsize=13)
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
if save_plot==True:
pp.savefig(fig)
pp.close()
if close_figs:
plt.close('all')
#for ll in range(L+adaptBias):
# print 'Sparsity=',np.mean(shapes[ll]>0)
#%% All Shapes slices
transpose_shape= True # should we transpose shape
ComponentsInFig=3 # number of components in Figure
index=0 #component display index
# z_slices=[0,1,2,3,4,5,6,7,8] #which z slices to look at slice plots/videos
z_slices=list(range(dims[min_dim+1])) #which z slices to look at slice plots/videos
if plot_shapes_slices:
if save_plot==True:
pp = PdfPages(Results_folder + 'Shapes_slices'+resultsName+'.pdf')
for ll in range(L+adaptBias):
if index==0:
fig=plt.figure(figsize=(18, 11))
for dd in range(len(z_slices)):
ax = plt.subplot(ComponentsInFig,len(z_slices),index*len(z_slices)+dd+1)
temp=shapes[ll].take(dd,axis=min_dim)
if transpose_shape:
temp=np.transpose(temp)
mi=np.min(shapes[ll])
ma=np.max(shapes[ll])
im=plt.imshow(temp,vmin=mi,vmax=ma,cmap=color_map)
plt.setp(ax,xticks=[],yticks=[])
if dd==0:
# component number
ax.text(0.02, 0.8, str(ll),
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
#sparsity
spar_str=str(np.round(np.mean(shapes[ll]>0)*100,2))+'%'
ax.text(0.02, 0.02, spar_str,
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='white',weight='bold', fontsize=13)
mn=int(np.floor(mi)) # colorbar min value
mx=int(np.ceil(ma)) # colorbar max value
md=old_div((mx-mn),2)
divider = make_axes_locatable(ax)
cax = divider.append_axes("bottom", size="5%", pad=0.05)
cb=plt.colorbar(im,cax=cax,orientation="horizontal")
cb.set_ticks([mn,md,mx])
cb.set_ticklabels([mn,md,mx])
#L^p
for p in range(2,2,2):
Lp=old_div((np.sum(shapes[ll]**p))**(old_div(1,float(p))),np.sum(shapes[ll]))
Lp_str=str(np.round(Lp*100,2))+'%' #'L'+str(p)+'='+
ax.text(0.02+p*0.15, 0.02, Lp_str,
verticalalignment='bottom', horizontalalignment='left',
transform=ax.transAxes,
color='yellow',weight='bold', fontsize=13)
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
index+=1
if (ll%ComponentsInFig==(ComponentsInFig-1)) or ll==L+adaptBias-1:
if save_plot==True:
pp.savefig(fig)
index=0
pp.close()
if close_figs:
plt.close('all')
#for ll in range(L+adaptBias):
# print 'Sparsity=',np.mean(shapes[ll]>0)
#%% ###### Plot Individual neurons' shape projection with clustering
a=ceil(sqrt(L+adaptBias))
b=ceil(old_div((L+adaptBias),a))
if plot_clustered_shape:
from sklearn.cluster import spectral_clustering
pp = PdfPages(Results_folder + 'ClusteredShapes'+resultsName+'.pdf')
figs=[]
for dd in range(len(shapes[0].shape)):
figs.append(plt.figure(figsize=(18 , 11)))
for ll in range(L):
ind=np.reshape(shapes[ll],(1,)+tuple(dims[1:]))>0
temp=data[np.repeat(ind,dims[0],axis=0)].reshape(dims[0],-1)
delta=1 #affinity trasnformation parameter
clust=3 #number of cluster
similarity=np.exp(old_div(-np.corrcoef(temp.T),delta))
labels = spectral_clustering(similarity, n_clusters=clust, eigen_solver='arpack')
ind2=np.array(np.nonzero(ind.reshape(-1))).reshape(-1)
temp_shape=np.repeat(np.zeros_like(shapes[ll]).reshape(-1,1),clust,axis=1)
for cc in range(clust):
temp_shape[ind2[labels==cc],cc]=1
temp_shape=temp_shape.reshape(tuple(dims[1:])+(clust,))
for dd in range(len(shapes[0].shape)):
current_fig=figs[dd]
ax = current_fig.add_subplot(a,b,ll+1)
if dd==2:
temp_shape=np.transpose(temp_shape,axes=[1,0,2,3])
ax.imshow(temp_shape.max(dd))
plt.setp(ax,xticks=[],yticks=[])
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
if save_plot==True:
for dd in range(len(shapes[0].shape)):
current_fig=figs[dd]
pp.savefig(current_fig)
pp.close()
if close_figs:
plt.close('all')
#%% ##### Video Shapes
if video_shapes:
components=list(range(min(asarray([C,L]))))
C=len(components)
if restrict_support==True:
shape_support=shapes[components[0]]>0
for cc in range(C):
shape_support=np.logical_or(shape_support,shapes[components[cc]]>0)
detrended_data=shape_support.reshape((1,)+tuple(dims[1:]))*detrended_data
fig = plt.figure(figsize=(16,7))
mi = 0
ma = max(data)*scale
#mi2 = 0
#ma2 = max(shapes[ll])*max(activity[ll])
ii=0
#import colormaps as cmaps
#cmap=cmaps.viridis
cmap=color_map
a=3
b=1+C
ax1 = plt.subplot(a,b,1)
im1 = ax1.imshow(data[ii].max(0), vmin=mi, vmax=ma,cmap=cmap)
title=ax1.set_title('Data')
#plt.colorbar(im1)
ax2=[]
ax4=[]
ax6=[]
im2=[]
im4=[]
im6=[]
for cc in range(C):
ax2.append(plt.subplot(a,b,2+cc))
comp=shapes[components[cc]].max(0)*activity_NonNegative[components[cc],ii]
ma2=max(shapes[components[cc]].max(0))*max(activity_NonNegative[components[cc]])*scale
im2.append(ax2[cc].imshow(comp,vmin=0,vmax=ma2,cmap=cmap))
#ax2[0].set_title('Shape')
# plt.colorbar(im2)
ax3 = plt.subplot(a,b,1+b)
im3 = ax3.imshow(data[ii].max(1), vmin=mi, vmax=ma,cmap=cmap)
#plt.colorbar(im3)
for cc in range(C):
ax4.append(plt.subplot(a,b,2+b+cc))
comp=shapes[components[cc]].max(1)*activity_NonNegative[components[cc],ii]
ma2=max(shapes[components[cc]].max(1))*max(activity_NonNegative[components[cc]])*scale
im4.append(ax4[cc].imshow(comp,vmin=0,vmax=ma2,cmap=cmap))
#plt.colorbar(im4)
ax5 = plt.subplot(a,b,1+2*b)
im5 = ax5.imshow(np.transpose(detrended_data[ii].max(2)), vmin=mi, vmax=ma,cmap=cmap)
#plt.colorbar(im5)
for cc in range(C):
ax6.append(plt.subplot(a,b,2+2*b+cc))
comp=np.transpose(shapes[components[cc]].max(2))*activity_NonNegative[components[cc],ii]
ma2=max(shapes[components[cc]].max(2))*max(activity_NonNegative[components[cc]])*scale
im6.append(ax6[cc].imshow(comp,vmin=0,vmax=ma2,cmap=cmap))
#plt.colorbar(im6)
fig.tight_layout()
ComponentsActive=np.array([])
for cc in range(C):
ComponentsActive=np.append(ComponentsActive,np.nonzero(activity_NonNegative[components[cc]]))
ComponentsActive=np.unique(ComponentsActive)
def update(tt):
ii=ComponentsActive[tt]
im1.set_data(data[ii].max(0))
im3.set_data(data[ii].max(1))
im5.set_data(np.transpose(data[ii].max(2)))
for cc in range(C):
im2[cc].set_data(shapes[components[cc]].max(0)*activity_NonNegative[components[cc],ii])
im4[cc].set_data(shapes[components[cc]].max(1)*activity_NonNegative[components[cc],ii])
im6[cc].set_data(np.transpose(shapes[components[cc]].max(2))*activity_NonNegative[components[cc],ii])
title.set_text('Data, time = %.1f' % ii)
if save_video==True:
writer = animation.writers['ffmpeg'](fps=10)
ani = animation.FuncAnimation(fig, update, frames=len(ComponentsActive), blit=True, repeat=False)
if restrict_support==True:
ani.save(Results_folder + 'Shapes_Restricted'+resultsName+'.mp4',dpi=dpi,writer=writer)
else:
ani.save(Results_folder + 'Shapes_'+resultsName+'.mp4',dpi=dpi,writer=writer)
else:
ani = animation.FuncAnimation(fig, update, frames=len(ComponentsActive), blit=True, repeat=False)
plt.show()
#%% ##### Plot denoised projection - Results
if plot_residual_projections==True:
dims=data.shape
cmap=color_map
pic_residual=percentile(residual, 95, axis=0)
pic_denoised = max_intensity(denoised_data, axis=0)
pic_data=percentile(data, 95, axis=0)
left = 0.05 # the left side of the subplots of the figure
right = 0.95 # the right side of the subplots of the figure
bottom = 0.05 # the bottom of the subplots of the figure
top = 0.95 # the top of the subplots of the figure
wspace = 0.05 # the amount of width reserved for blank space between subplots
hspace = 0.05 # the amount of height reserved for white space between subplots
fig1=plt.figure(figsize=(11,18))
mi=min(pic_data)
ma=max(pic_data)
ax = plt.subplot(311)
im=ax.imshow(pic_data.max(0),vmin=mi,vmax=ma,cmap=cmap)
ax.set_title('Data')
plt.colorbar(im)
plt.setp(ax,xticks=[],yticks=[])
ax2 = plt.subplot(312)
im2=ax2.imshow(max_intensity(pic_denoised,0),interpolation='None')
ax2.set_title('Denoised')
plt.setp(ax,xticks=[],yticks=[])
plt.colorbar(im2)
ax3 = plt.subplot(313)
im3=ax3.imshow(pic_residual.max(0),cmap=cmap)
ax3.set_title('Residual')
plt.setp(ax,xticks=[],yticks=[])
plt.colorbar(im3)
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
fig2=plt.figure(figsize=(11,18))
mi=min(pic_data)
ma=max(pic_data)
ax = plt.subplot(311)
im=ax.imshow(pic_data.max(1),vmin=mi,vmax=ma,cmap=cmap)
ax.set_title('Data')
plt.colorbar(im)
plt.setp(ax,xticks=[],yticks=[])
ax2 = plt.subplot(312)
im2=ax2.imshow(max_intensity(pic_denoised,1),interpolation='None')
ax2.set_title('Denoised')
plt.colorbar(im2)
plt.setp(ax,xticks=[],yticks=[])
ax3 = plt.subplot(313)
im3=ax3.imshow(pic_residual.max(1),cmap=cmap)
ax3.set_title('Residual')
plt.colorbar(im3)
plt.setp(ax,xticks=[],yticks=[])
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
fig3=plt.figure(figsize=(11,18))
mi=min(pic_data)
ma=max(pic_data)
ax = plt.subplot(311)
im=ax.imshow(pic_data.max(2).T,vmin=mi,vmax=ma,cmap=cmap)
ax.set_title('Data')
plt.colorbar(im)
plt.setp(ax,xticks=[],yticks=[])
ax2 = plt.subplot(312)
im2=ax2.imshow(np.transpose(max_intensity(pic_denoised,2),[1,0,2]),interpolation='None')
ax2.set_title('denoised')
plt.setp(ax,xticks=[],yticks=[])
plt.colorbar(im2)
ax3 = plt.subplot(313)
im3=ax3.imshow(np.transpose(pic_residual.max(2)),cmap=cmap)
ax3.set_title('Residual')
plt.colorbar(im3)
plt.setp(ax,xticks=[],yticks=[])
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
if save_plot==True:
pp = PdfPages(Results_folder + 'Data_Denoised_Residual_Projections'+resultsName+'.pdf')
pp.savefig(fig1)
pp.savefig(fig2)
pp.savefig(fig3)
pp.close()
#fig = plt.figure()
#plt.plot(MSE_array)
#plt.xlabel('Iteration')
#plt.ylabel('MSE')
#plt.show()
#%% ##### Plot denoised slices - Results
# z_slices=[0,2,4,6,8] #which z slices to look at slice plots/videos
z_slices=list(range(dims[min_dim+1])) #which z slices to look at slice plots/videos
D=len(z_slices)
if plot_residual_slices==True:
dims=data.shape
cmap=color_map
pic_residual=percentile(residual, 95, axis=0)
pic_denoised = max_intensity(denoised_data, axis=0)
pic_data=percentile(data, 95, axis=0)
a=3 #number of rows
fig1=plt.figure(figsize=(18,11))
mi=min(pic_data)
ma=max(pic_data)
for kk in range(D):
ax2 = plt.subplot(a,D,kk+1)
temp=np.squeeze(np.take(pic_denoised,(z_slices[kk],),axis=min_dim))
im2=ax2.imshow(temp,interpolation='None')
ax2.set_title('Denoised')
plt.setp(ax2,xticks=[],yticks=[])
plt.colorbar(im2)
ax = plt.subplot(a,D,kk+D+1)
temp=np.squeeze(np.take(pic_data,(z_slices[kk],),axis=min_dim))
im=ax.imshow(temp,vmin=mi,vmax=ma,cmap=cmap)
ax.set_title('Data')
plt.colorbar(im)
plt.setp(ax,xticks=[],yticks=[])
ax3 = plt.subplot(a,D,kk+2*D+1)
temp=np.squeeze(np.take(pic_residual,(z_slices[kk],),axis=min_dim))
im3=ax3.imshow(temp,cmap=cmap)
ax3.set_title('Residual')
plt.setp(ax3,xticks=[],yticks=[])
plt.colorbar(im3)
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
if save_plot==True:
pp = PdfPages(Results_folder + 'Data_Denoised_Residual_Slice_'+resultsName+'.pdf')
pp.savefig(fig1)
pp.close()
#fig = plt.figure()
#plt.plot(MSE_array)
#plt.xlabel('Iteration')
#plt.ylabel('MSE')
#plt.show()
#%% ##### 2D Video Residual
if video_residual_2D:
fig = plt.figure(figsize=(16,7))
mi = 0
ma = np.percentile(data,satuartion_percentile)
mi3 = 0
ma3 = old_div(ma,np.max([np.floor(old_div(ma,np.percentile(residual[residual>0],satuartion_percentile))),1]))
ii=0
#import colormaps as cmaps
#cmap=cmaps.viridis
cmap=color_map
a=1
b=3
im_array=[]
temp=np.shape(data[ii])
ax1 = plt.subplot(a,b,1)
pic=denoised_data[ii]
im_array += [ax1.imshow(pic,interpolation='None')]
ax1.set_title('Denoised')
plt.setp(ax1,xticks=[],yticks=[])
ax2 = plt.subplot(a,b,2)
pic=data[ii]
im_array += [ax2.imshow(pic, vmin=mi, vmax=ma,cmap=cmap)]
title=ax2.set_title('Data')
plt.setp(ax2,xticks=[],yticks=[])
divider = make_axes_locatable(ax2)
cax2 = divider.append_axes("right", size="5%", pad=0.05)
plt.colorbar(im_array[-1], cax=cax2)
ax3 = plt.subplot(a,b,3)
pic=residual[ii]
im_array += [ax3.imshow(pic, vmin=mi3, vmax=ma3,cmap=cmap)]
ax3.set_title('Residual x' + '%.1f' % (old_div(ma,ma3)))
plt.setp(ax3,xticks=[],yticks=[])
divider = make_axes_locatable(ax3)
cax3 = divider.append_axes("right", size="5%", pad=0.05)
plt.colorbar(im_array[-1], cax=cax3)
# fig.tight_layout()
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
def update(ii):
im_array[0].set_data(denoised_data[ii])
im_array[1].set_data(data[ii])
im_array[2].set_data(residual[ii])
if frame_rate!=[]:
title.set_text('Data, time = %.2f sec' % (old_div(ii,frame_rate)))
else:
title.set_text('Data, time = %.1f' % ii)
if save_video==True:
writer = animation.writers['ffmpeg'](fps=10)
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
ani.save(Results_folder + 'Data_Denoised_Residual_2D_' +resultsName+'.avi',dpi=dpi,writer=writer)
else:
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
plt.show()
#%% ##### Video Projections Residual
if video_residual:
fig = plt.figure(figsize=(16,7))
mi = 0
ma = max(data)*scale
mi3 = 0
ma3 = max(residual)*scale
ii=0
#import colormaps as cmaps
#cmap=cmaps.viridis
cmap=color_map
spatial_dims_ind=list(range(len(dims)-1))
D=len(spatial_dims_ind)
a=D
b=3
im_array=[]
transpose_flags=[]
for kk in range(D):
transpose_flags+= [False]
temp=np.shape(data[ii].max(spatial_dims_ind[kk]))
if temp[0]>temp[1]:
transpose_flags[kk]=True
for kk in range(D):
ax1 = plt.subplot(a,b,D*kk+1)
if transpose_flags[kk]==False:
pic=max_intensity(denoised_data[ii],spatial_dims_ind[kk])
else:
pic=np.transpose(max_intensity(denoised_data[ii],spatial_dims_ind[kk]),[1,0,2])
im_array += [ax1.imshow(pic,interpolation='None')]
ax1.set_title('Denoised')
plt.colorbar(im_array[-1])
plt.setp(ax1,xticks=[],yticks=[])
ax2 = plt.subplot(a,b,D*kk+2)
if transpose_flags[kk]==False:
pic=data[ii].max(spatial_dims_ind[kk])
else:
pic=np.transpose(data[ii].max(spatial_dims_ind[kk]))
im_array += [ax2.imshow(pic, vmin=mi, vmax=ma,cmap=cmap)]
title=ax2.set_title('Data')
plt.colorbar(im_array[-1])
plt.setp(ax2,xticks=[],yticks=[])
ax3 = plt.subplot(a,b,D*kk+3)
if transpose_flags[kk]==False:
pic=residual[ii].max(spatial_dims_ind[kk])
else:
pic=np.transpose(residual[ii].max(spatial_dims_ind[kk]))
im_array += [ax3.imshow(pic, vmin=mi3, vmax=ma3,cmap=cmap)]
ax3.set_title('Residual')
plt.colorbar(im_array[-1])
plt.setp(ax3,xticks=[],yticks=[])
# fig.tight_layout()
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
def update(ii):
for kk in range(D):
if transpose_flags[kk]==False:
im_array[kk*D].set_data(max_intensity(denoised_data[ii],spatial_dims_ind[kk]))
im_array[kk*D+1].set_data(data[ii].max(spatial_dims_ind[kk]))
im_array[kk*D+2].set_data(residual[ii].max(spatial_dims_ind[kk]))
else:
im_array[kk*D].set_data(np.transpose(max_intensity(denoised_data[ii],spatial_dims_ind[kk]),[1,0,2]))
im_array[kk*D+1].set_data(np.transpose(data[ii].max(spatial_dims_ind[kk])))
im_array[kk*D+2].set_data(np.transpose(residual[ii].max(spatial_dims_ind[kk])))
title.set_text('Data, time = %.1f' % ii)
if save_video==True:
writer = animation.writers['ffmpeg'](fps=10)
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
ani.save(Results_folder + 'Data_Denoised_Residual_Projections'+resultsName+'.mp4',dpi=dpi,writer=writer)
else:
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
plt.show()
#%% ##### Video Slices Residual
# z_slices=[0,2,4,6,8] #which z slices to look at slice plots/videos
z_slices=list(range(dims[min_dim+1])) #which z slices to look at slice plots/videos
if video_slices:
fig = plt.figure(figsize=(16,7))
mi = 0
ma = np.percentil(data[data>0],satuartion_percentile)
mi3 = 0
ma3 = np.percentil(data[data>0],satuartion_percentile)
ii=0
#import colormaps as cmaps
#cmap=cmaps.viridis
cmap=color_map
a=3
D=len(z_slices) #number of spatial dimensions
im_array=[]
transpose_flag= True
for kk in range(D):
ax1 = plt.subplot(a,D,kk+1)
temp=np.squeeze(np.take(denoised_data[ii],(z_slices[kk],),axis=min_dim))
if transpose_flag==False:
pic=temp
else:
pic=np.transpose(temp,[1,0,2])
im_array += [ax1.imshow(pic,interpolation='None')]
ax1.set_title('Denoised, z='+ str(z_slices[kk]+1))
plt.colorbar(im_array[-1])
plt.setp(ax1,xticks=[],yticks=[])
ax2 = plt.subplot(a,D,kk+D+1)
temp=np.squeeze(np.take(data[ii],(z_slices[kk],),axis=min_dim))
if transpose_flag==False:
pic=temp
else:
pic=np.transpose(temp)
im_array += [ax2.imshow(pic, vmin=mi, vmax=ma,cmap=cmap)]
title=ax2.set_title('Data')
plt.colorbar(im_array[-1])
plt.setp(ax2,xticks=[],yticks=[])
ax3 = plt.subplot(a,D,kk+2*D+1)
temp=np.squeeze(np.take(residual[ii],(z_slices[kk],),axis=min_dim))
if transpose_flag==False:
pic=temp
else:
pic=np.transpose(temp)
im_array += [ax3.imshow(pic, vmin=mi3, vmax=ma3,cmap=cmap)]
ax3.set_title('Residual')
plt.colorbar(im_array[-1])
plt.setp(ax3,xticks=[],yticks=[])
# fig.tight_layout()
plt.subplots_adjust(left, bottom, right, top, wspace, hspace)
def update(ii):
for kk in range(D):
temp1=np.squeeze(np.take(denoised_data[ii],(z_slices[kk],),axis=min_dim))
temp2=np.squeeze(np.take(data[ii],(z_slices[kk],),axis=min_dim))
temp3=np.squeeze(np.take(residual[ii],(z_slices[kk],),axis=min_dim))
if transpose_flag==False:
im_array[a*kk].set_data(temp1)
im_array[a*kk+1].set_data(temp2)
im_array[a*kk+2].set_data(temp3)
else:
im_array[a*kk].set_data(np.transpose(temp1,[1,0,2]))
im_array[a*kk+1].set_data(np.transpose(temp2))
im_array[a*kk+2].set_data(np.transpose(temp3))
title.set_text('Data, time = %.1f' % ii)
if save_video==True:
writer = animation.writers['ffmpeg'](fps=10)
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
ani.save(Results_folder + 'Data_Denoised_Residual_Slices'+resultsName+'.mp4',dpi=dpi,writer=writer)
else:
ani = animation.FuncAnimation(fig, update, frames=len(data), blit=False, repeat=False)
plt.show()
def LocalNMF(data, centers, sig, NonNegative=True,FinalNonNegative=True,verbose=False,adaptBias=True,TargetAreaRatio=[],estimateNoise=False,
PositiveError=False,MedianFilt=False,Connected=False,FixSupport=False, WaterShed=False,SmoothBkg=False,FineTune=True,Deconvolve=False,
SigmaMask=[],updateLambdaIntervals=2,updateRhoIntervals=2,addComponentsIntervals=1,bkg_per=20,SigmaBlur=[],
iters=10,iters0=[30], mbs=[1], ds=1,lam1_s=0,lam1_t=0,lam2_s=0,lam2_t=0):
"""
Parameters
----------
data : array, shape (T, X, Y[, Z])
block of the data
centers : array, shape (L, D)
L centers of suspected neurons where D is spatial dimension (2 or 3)
sig : array, shape (D,)
size of the gaussian kernel in different spatial directions
NonNegative : boolean
if True, neurons activity should be considered as non-negative
FinalNonNegative : boolean
if False, last activity iteration is done without non-negativity constraint, even if NonNegative==True
verbose : boolean
print progress and record MSE if true (about 2x slower)
adaptBias : boolean
subtract rank 1 estimate of bias (background)
TargetAreaRatio : list of length 2
Lower and upper bounds on sparsity of non-background components
estimateNoise : boolean
estimate noise variance and use it determine if to add components, and to modify sparsity by affecting lam1_s (does not work very well)
PositiveError : boolean
do not allow pixels in which the residual (summed over time) becomes negative, by increasing lam1_s in these pixels
MedianFilt : boolean
do median filter of spatial components
Connected: boolean
impose connectedness of spatial component by keeping only the largest non-zero connected component in each iteration of HALS
WaterShed: boolean
impose that each spatial component has a single watershed region
SmoothBkg: boolean
Remove local peaks from background component
FixSupport : boolean
do not allow spatial components to be non-zero where sub-sampled spatial components are zero
FineTune : boolean
fine tune main iterations on full data, if not, use (last) downsampled data
Deconvolve : boolean
Deconvolve activity to get smoothed (denoised) calcium trace. This is done only on the main itreations, and if FineTune=True
SigmaMask : scalar or empty
if not [], then update masks so that they are SigmaMasks around non-zero support of shapes
SigmaBlur : scalar
if not [], then de-blur spatial components using Gaussian Kernel of this width
updateLambdaIntervals : int
update lam1_s every this number of HALS iterations, to match contraints
updateRhoIntervals : int
decrease rho, update rate of lam1_s, every this number of updateLambdaIntervals HALS iterations (only active during main iterations)
addComponentsIntervals : int
add new component, if possible, every this number of updateLambdaIntervals HALS iterations (only active during sub-sampled iterations)
bkg_per : float
the background is intialized at this height (percentrilce image)
iters : int
number of final iterations on whole data
iters0 : list
numbers of initial iterations on subset
mbs : list
minibatchsizes for temporal downsampling
ds : int or list
factor for spatial downsampling, can be an integer or a list of the size of spatial dimensions
lam1_s : float
L_1 regularization constant for sparsity of shapes
lam2_s : float
L_2 regularization constant for sparsity of shapes
lam_t : float
L_1 regularization constant for sparsity of activity
lam2_t : float
L_2 regularization constant for sparsity of activity
Returns
-------
MSE_array : list (empty if verbose is False)
Mean square error during algorithm operation
shapes : array, shape (L+adaptBias, X, Y (,Z))
the neuronal shape vectors (empty if no components found)
activity : array, shape (L+adaptBias, T)
the neuronal activity for each shape (empty if no components found)
boxes : array, shape (L, D, 2)
edges of the boxes in which each neuronal shapes lie (empty if no components found)
"""
# Catch Errors
if ds!=1 and SigmaBlur!=[]:
raise NameError('case ds!=1 and SigmaBlur!=[] no yet written in NMF code')
# Initialize Parameters
dims = data.shape # data dimensions
D = len(dims) #number of data dimensions
R = 3 * asarray(sig) # size of bounding box is 3 times size of neuron
L = len(centers) # number of components (not including background)
inner_iterations=10 # number of iterations in inners loops
shapes = [] #array of spatial components
mask = [] # binary array, support of spatial components
boxes = zeros((L, D - 1, 2), dtype=int) #initial support of spatial components
MSE_array = [] #CNMF residual error
mb = mbs[0] if iters0[0] > 0 else 1
activity = zeros((L, old_div(dims[0], mb))) #array of temporal components
lam1_s0=np.copy(lam1_s) #intial spatial sparsity (l1) parameters
if TargetAreaRatio!=[]:
if TargetAreaRatio[0]>TargetAreaRatio[1]:
print('WARNING - TargetAreaRatio[0]>TargetAreaRatio[1] !!!')
if iters0[0] == 0:
ds = 1
### Initialize shapes, activity, and residual ###
data0,dims0=DownScale(data,mb,ds) #downscaled data and dimensions
if isinstance(ds,int):
ds=ds*np.ones(D-1)
if D == 4: #downscale activity
activity = data0[:, list(map(int, old_div(centers[:, 0], ds[0]))), list(map(int, old_div(centers[:, 1], ds[1]))),
list(map(int, old_div(centers[:, 2], ds[2])))].T
else:
activity = data0[:, list(map(int, old_div(centers[:, 0], ds[0]))), list(map(int, old_div(centers[:, 1], ds[1])))].T
data0 = data0.reshape(dims0[0], -1) #reshape data0 to more convient timexspace form
Energy0=np.sum(data0**2,axis=0) #data0 energy per pixel
data0sum=np.sum(data0,axis=0) # for sign check later
data = data.astype('float').reshape(dims[0], -1) #reshape data to more convient timexspace form
datasum=np.sum(data,axis=0)# for sign check later
# float is faster than float32, presumable float32 gets converted later on
# to float again and again
Energy=np.sum((data**2),axis=0) #data energy per pixel
# extract shapes and activity from given centers
for ll in range(L):
boxes[ll] = GetBox(old_div(centers[ll], ds), old_div(R, ds), dims0[1:])
temp = zeros(dims0[1:])
temp[[slice(*a) for a in boxes[ll]]]=1
mask += np.where(temp.ravel())
temp = [old_div((arange(int(old_div(dims[i + 1], ds[i]))) -int( old_div(centers[ll][i], ds[i]))) ** 2, (2 * (old_div(sig[i], ds[i])) ** 2))
for i in range(D - 1)]
temp = exp(-sum(ix_(*temp)))
temp.shape = (1,) + dims0[1:]
temp = RegionCut(temp, boxes[ll])
shapes.append(temp[0])
S = zeros((L + adaptBias, prod(dims0[1:]))) #shape component
for ll in range(L):
S[ll] = RegionAdd(
zeros((1,) + dims0[1:]), shapes[ll].reshape(1, -1), boxes[ll]).ravel()
if adaptBias:
# Initialize background as bkg_per percentile
S[-1] = percentile(data0, bkg_per, 0)
activity = np.r_[activity, ones((1, dims0[0]))]
lam1_s=lam1_s0*np.ones_like(S)*mbs[0] #intialize sparsity parameters
### Get shape estimates on subset of data ###
if iters0[0] > 0:
for it in range(len(iters0)):
if estimateNoise:
sn_target,sn_std= GetSnPSDArray(data0)#target noise level
else:
sn_target=np.zeros(prod(dims0[1:]))
sn_std=sn_target
MSE_target = np.mean(sn_target**2)
ES=ExponentialSearch(lam1_s) #object to update sparsity parameters
lam1_s=ES.lam
for kk in range(iters0[it]):
# update sparisty parameters
if kk%updateLambdaIntervals==0:
sn=old_div(np.sqrt(Energy0-2*np.sum(np.dot(activity,data0)*S,axis=0)+np.sum(np.dot(np.dot(activity,activity.T),S)*S,axis=0)),dims0[0]) # efficient way to calcuate MSE per pixel
delta_sn=sn-sn_target # noise margin
signcheck=(data0sum-np.dot(np.sum(activity.T,axis=0),S))<0
if PositiveError: #obsolete
delta_sn[signcheck]=-float("inf") # residual should not have negative pixels, so we increase lambda for these pixels
if len(S)==0:
spars=0
else:
spars=np.mean(S>0,axis=1)
temp=repeat(delta_sn.reshape(1,-1),L+adaptBias,axis=0)
if TargetAreaRatio==[]:
cond_decrease=temp>sn_std
cond_increase=temp<-sn_std
else:
if adaptBias:
spars[-1]=old_div((TargetAreaRatio[1]+TargetAreaRatio[0]),2) # ignore sparsity target for background (bias) component
temp2=repeat(spars.reshape(-1,1),len(S[0]),axis=1)
cond_increase=np.logical_or(temp2>TargetAreaRatio[1],temp<-sn_std)
cond_decrease=np.logical_and(temp2<TargetAreaRatio[0],temp>sn_std)
ES.update(cond_decrease,cond_increase)
lam1_s=ES.lam
#Print residual error and additional information
MSE = np.mean(sn**2)
if verbose and L>0:
print(' MSE = {0:.6f}, Target MSE={1:.6f},Sparsity={2:.4f},lam1_s={3:.6f}'.format(MSE,MSE_target,np.mean(spars[:L]),np.mean(lam1_s)))
#add a new component
if (kk%addComponentsIntervals==0) and (kk!=iters0[it]-1):
delta_sn[signcheck]=-float("inf") # residual should not have negative pixels
new_cent=np.argmax(delta_sn) #should I smooth the data a bit first?
MSE_std=np.mean(sn_std**2)
checkNoZero= not((0 in np.sum(activity,axis=1)) and (0 in np.sum(S,axis=1)))
if ((MSE-MSE_target>2*MSE_std) and checkNoZero and (delta_sn[new_cent]>sn_std[new_cent])):
S, activity, mask,centers,boxes,L=addComponent(new_cent,data0,dims0,old_div(R,ds),S, activity, mask,centers,boxes,adaptBias)
new_lam=lam1_s0*np.ones_like(data0[0,:]).reshape(1,-1)
lam1_s=np.insert(lam1_s,0,values=new_lam,axis=0)
ES=ExponentialSearch(lam1_s) #we need to restart exponential search each time we add a component
#apply additional constraints/processing
if SigmaBlur==[]:
S = HALS4shape(data0, S, activity,mask,lam1_s,lam2_s,adaptBias,inner_iterations)
else: #obsolete
S=FISTA4shape(data0, S, activity,mask,lam1_s,adaptBias,SigmaBlur,dims0)
if Connected==True:
S=LargestConnectedComponent(S,dims0,adaptBias)
if WaterShed==True:
S=LargestWatershedRegion(S,dims0,adaptBias)
activity = HALS4activity(data0, S, activity,NonNegative,lam1_t,lam2_t,dims0,SigmaBlur,inner_iterations)
if SigmaMask!=[]:
mask=GrowMasks(S,mask,boxes,dims0,adaptBias,SigmaMask)
S, activity, mask,centers,boxes,ES,L=RenormalizeDeleteSort(S, activity, mask,centers,boxes,ES,adaptBias,MedianFilt)
lam1_s=ES.lam
if SmoothBkg==True:
S=SmoothBackground(S,dims0,adaptBias,tuple(old_div(np.array(sig),np.array(ds))))
print('Subsampled iteration',kk,'it=',it,'L=',L)
# use next (smaller) value for temporal downscaling
if it < len(iters0) - 1:
mb = mbs[it + 1]
data0 = data[:len(data) / mb * mb].reshape(-1, mb, prod(dims[1:])).mean(1)
if D==4:
data0 = data0.reshape(len(data0), int(old_div(dims[1], ds[0])), ds[0], int(old_div(dims[2], ds[1])), ds[1],
int(old_div(dims[3], ds[2])), ds[2]).mean(-1).mean(-2).mean(-3)
else:
data0 = data0.reshape(len(data0), int(old_div(dims[1], ds[0])), ds[0], int(old_div(dims[2], ds[1])),
ds[1]).mean(-1).mean(-2)
data0.shape = (len(data0), -1)
activity = ones((L + adaptBias, len(data0))) * activity.mean(1).reshape(-1, 1)
lam1_s=lam1_s*mbs[it+1]/mbs[it]
activity = HALS4activity(data0, S, activity,NonNegative,lam1_t,lam2_t,dims0,SigmaBlur,30)
S, activity, mask,centers,boxes,ES,L=RenormalizeDeleteSort(S, activity, mask,centers,boxes,ES,adaptBias,MedianFilt)
lam1_s=ES.lam
### Stop adding components ###
if L==0: #if no non-background components found, return empty arrays
return [], [], [], []
if FineTune: ### Upscale Back to full data ##
activity = ones((L + adaptBias, dims[0])) * activity.mean(1).reshape(-1, 1)
data0=data
dims0=dims
if D==4:
S = repeat(repeat(repeat(S.reshape((-1,) + dims0[1:]), ds[0], 1), ds[1], 2), ds[2], 3)
lam1_s= repeat(repeat(repeat(lam1_s.reshape((-1,) + dims0[1:]), ds[0], 1), ds[1], 2), ds[2], 3)
else:
S = repeat(repeat(S.reshape((-1,) + dims0[1:]), ds[0], 1), ds[1], 2)
lam1_s= repeat(repeat(lam1_s.reshape((-1,) + dims0[1:]), ds[0], 1), ds[1], 2)
for dd in range(1,D):
while S.shape[dd]<dims[dd]:
shape_append=np.array(S.shape)
shape_append[dd]=1
S=np.append(S,values=np.take(S,-1,axis=dd).reshape(shape_append),axis=dd)
lam1_s=np.append(lam1_s,values=np.take(lam1_s,-1,axis=dd).reshape(shape_append),axis=dd)
S=S.reshape(L + adaptBias, -1)
lam1_s=lam1_s.reshape(L+ adaptBias,-1)
for ll in range(L):
boxes[ll] = GetBox(centers[ll], R, dims[1:])
temp = zeros(dims[1:])
temp[[slice(*a) for a in boxes[ll]]] = 1
mask[ll] = np.where(temp.ravel())[0]
if FixSupport: #obsolete
for ll in range(L):
lam1_s[ll,S[ll]==0]=float("inf")
ES=ExponentialSearch(lam1_s)
activity = HALS4activity(data0, S, activity,NonNegative,lam1_t,lam2_t,dims0,SigmaBlur, 30)
S, activity, mask,centers,boxes,ES,L=RenormalizeDeleteSort(S, activity, mask,centers,boxes,ES,adaptBias,MedianFilt)
lam1_s=ES.lam
if estimateNoise:
sn_target,sn_std= GetSnPSDArray(data0)#target noise level
else:
sn_target=np.zeros(prod(dims0[1:]))
sn_std=sn_target
MSE_target = np.mean(sn_target**2)
MSE_std=np.mean(sn_std**2)
# MSE = np.mean((data0-np.dot(activity.T,S))**2)
#### Main Loop ####
print('starting main NMF loop')
for kk in range(iters):
lam1_s=ES.lam #update sparsity parameters
if SigmaBlur==[]:
S = HALS4shape(data0, S, activity,mask,lam1_s,lam2_s,adaptBias,inner_iterations)
else: #obsolete
S = FISTA4shape(data0, S, activity,mask,lam1_s,adaptBias,SigmaBlur,dims0)
#apply additional constraints/processing
if Connected==True:
S=LargestConnectedComponent(S,dims0,adaptBias)
if WaterShed==True:
S=LargestWatershedRegion(S,dims0,adaptBias)
if kk==iters-1:
if FinalNonNegative==False:
NonNegative=False
activity = HALS4activity(data0, S, activity,NonNegative,lam1_t,lam2_t,dims0,SigmaBlur,inner_iterations)
if FineTune and Deconvolve:
for ll in range(L):
if np.sum(np.abs(activity[ll])>0)>30: #make sure there is enough signal before we try to deconvolve
activity[ll], _, _, _, _ = deconvolve(activity[ll], penalty=0)
if SigmaMask!=[]:
mask=GrowMasks(S,mask,boxes,dims0,adaptBias,SigmaMask)
S, activity, mask,centers,boxes,ES,L=RenormalizeDeleteSort(S, activity, mask,centers,boxes,ES,adaptBias,MedianFilt)
# Measure MSE and update sparsity parameters
print('main iteration kk=',kk,'L=',L)
if (kk+1)%updateLambdaIntervals==0:
sn=np.sqrt(old_div((Energy-2*np.sum(np.dot(activity,data0)*S,axis=0)+np.sum(np.dot(np.dot(activity,activity.T),S)*S,axis=0)),dims0[0]))
delta_sn=sn-sn_target
MSE = np.mean(sn**2)
signcheck=(datasum-np.dot(np.sum(activity.T,axis=0),S))<0
if PositiveError: #obsolete
delta_sn[signcheck]=-float("inf") # residual should not have negative pixels, so we increase lambda for these pixels
if S==[]:
spars=0
else:
spars=np.mean(S>0,axis=1)
temp=repeat(delta_sn.reshape(1,-1),L+adaptBias,axis=0)
if TargetAreaRatio==[]:
cond_decrease=temp>sn_std
cond_increase=temp<-sn_std
else:
if adaptBias:
spars[-1]=old_div((TargetAreaRatio[1]+TargetAreaRatio[0]),2) # ignore sparsity target for background (bias) component
temp2=repeat(spars.reshape(-1,1),len(S[0]),axis=1)
cond_increase=np.logical_or(temp2>TargetAreaRatio[1],temp<-sn_std)
cond_decrease=np.logical_and(temp2<TargetAreaRatio[0],temp>sn_std)
ES.update(cond_decrease,cond_increase)
lam1_s=ES.lam
if kk<old_div(iters,3): #restart exponential search unless enough iterations have passed
ES=ExponentialSearch(lam1_s)
else:
if not(np.any(cond_increase) or np.any(cond_decrease)):
print('sparsity target reached')
break
if L+adaptBias>1: # if we have more then one component just keep exponitiated grad descent instead
if (kk+1)%updateRhoIntervals==0: #update rho every updateRhoIntervals if we are still not converged
if np.any(spars[:L]<TargetAreaRatio[0]) or np.any(spars[:L]>TargetAreaRatio[1]):
ES.rho=2-old_div(1,(ES.rho))
print('rho=',ES.rho)
ES=ExponentialSearch(lam1_s,rho=ES.rho)
# prinst MSE and other information
if verbose:
print(' MSE = {0:.6f}, Target MSE={1:.6f},Sparsity={2:.4f},lam1_s={3:.6f}'.format(MSE,MSE_target,np.mean(spars[:L]),np.mean(lam1_s)))
if kk == (iters - 1):
print('Maximum iteration limit reached')
MSE_array.append(MSE)
# Some post-processing
S=S.reshape((-1,) + dims[1:])
S,activity,L=PruneComponents(S,activity,L) #prune "bad" components
if len(S)>1:
S,activity,L=MergeComponents(S,activity,L,threshold=0.9,sig=10) #merge very similar components
if not FineTune:
activity = ones((L + adaptBias, dims[0])) * activity.mean(1).reshape(-1, 1) #extract activity from full data
activity=HALS4activity(data, S.reshape((len(S),-1)), activity,NonNegative,lam1_t,lam2_t,dims0,SigmaBlur,iters=30)
return asarray(MSE_array), S, activity
# example to check code works
#T = 1000
#X = 201
#Y = 101
#data = np.random.randn(T, X, Y)
#centers = asarray([[40, 30]])
#data[:, 30:45, 25:33] += 2*np.sin(np.array(range(T))/200).reshape(-1,1,1)*np.ones([T,15,8])
#sig = [300, 300]
#
#MSE_array, shapes, activity, boxes = LocalNMF(
# data, centers, sig, NonNegative=True, verbose=True,lam1_s=0.1,adaptBias=True)
#
#
#import matplotlib.pyplot as plt
#plt.imshow(shapes[0])
#
#for ll in range(len(shapes)):
# print np.mean(shapes[ll]>0)
