General imports

In [1]:
import numpy as np
import pydoc
import simcalc
In [2]:
%load_ext autoreload
%autoreload 2

Plotting

In [3]:
import holoviews as hv
hv.extension('bokeh', 'matplotlib')
%opts Raster (cmap='Greys_r') Image (cmap='Greys_r')

Generate basic spikes and calcium traces

In [4]:
#gcamp6f stats
taur = 0.1
taud = 0.76
A = 1
p = [0.85, -0.006]

# define other inputs
nCells = 5
h = 50 # height and width of field of view
T = 7.5 # final time
dt= 0.01 # timestep
cycles = 2 # how many stimulus cycles
rates = np.linspace(0,1,nCells)
rates[0] = 0.2
rates[np.arange(1,nCells,3)] = 0.5
rates[np.arange(2,nCells,3)] = 0.5

# run model
model = simcalc.SimCalc(h, nCells, T, cycles, dt)
model.gen_calcium(taur,taud,A,p, rates=rates)
In [5]:
%%opts Curve [width=900] Curve.spikes (color='k') Curve.F (color='g')
sp = model.sp # spikes
c = model.c # inherent calcium level without nonlinearities
F = model.F # calcium levels with measurement nonlinearities taken into account
times = np.arange(F.shape[1])*dt # time points

# plot spikes and calcium traces
calc_fig = hv.Overlay()
offset=6
for n in range(nCells): # loop over trials
    calc_fig *= hv.Curve(zip(times, sp[n, :]*1.5+offset*n), group='spikes', label='spikes')
    calc_fig *= hv.Curve(zip(times, F[n, :]+offset*n), group='F', label='calcium ground truth')
calc_fig
Out[5]:

Generate neuropil signal and spatial footprint

In [6]:
model.gen_npil()

Generate spatial kernels for the cells

In [7]:
model.gen_spat_kernels()

Show generated shapes

In [8]:
fig = hv.Layout()
for i in range(nCells):
    fig += hv.Image(model.spat_kernels[i])
fig.cols(3)
Out[8]:

Generate video data

In [9]:
model.gen_data(p)

Of 5 cells:
0
1
2
3
4

Final video

In [10]:
%%output holomap='scrubber'
subt = 10 # downsampling
startT = 0
endT = model.F.shape[1]
frames = {f: hv.Image(model.video[int(f-subt/2):int(f+subt/2),:,:].mean(axis=0)) for f in np.arange(startT+subt/2,endT,subt)}
hv.HoloMap(frames)
Out[10]:


Once Loop Reflect

Generate masks for the neurons, and measure and display traces

Because of neuropil and possible cell overlap, these won't match perfectly

In [11]:
%%opts Curve.Measured (color='b')
masks = model.gen_masks(threshold=0.5) 
traces = [np.mean(model.video[:, masks[i]], axis=1) for i in range(nCells)]
traces = [traces[i]-traces[i].min() for i in range(nCells)] # remove minima
for n in range(nCells): # loop over trials
    # update figure made before with spikes and calcium traces
    calc_fig *= hv.Curve(zip(times, traces[n]+offset*n), group='Measured', label='calcium measured')
calc_fig
Out[11]:

Some other useful things that can be found in the model object

In [12]:
# the average background image
hv.Image(model.Bg_img)
Out[12]:

The average background trace

In [13]:
%%opts Curve [width=900]
hv.Curve(model.Bg_trace)
Out[13]:

Export data as tiff file

In [14]:
model.save_tiff(fname='data.tif')