def resp_map(norm_stack, n_baseline=30, n_stim=30):
"""
Compute a smooth response image
Parameters
----------
norm_stack: 3D np array
n_baseline: int
n_stim: int
Returns
-------
im_resp: numpy array
Response map
df: numpy array
Signal of all responding pixels
"""
# r = norm_stack[..., n_baseline:n_baseline+n_stim].mean(2)
r = find_resp(norm_stack)
# Keep only the top 5% pixels
z = r > np.percentile(r, 95)
im_resp = np.zeros(r.shape)
im_resp[z] = r[z]
mask = np.ones(r.shape)
# mask[100:-100,100:-100] = 1
im_resp = im_resp * mask
im_resp = gauss_filt(im_resp, 2)
##### Warning
im_resp = clean_response(im_resp)
df = norm_stack[im_resp > 0, :]
return im_resp, df
def overlay(ref, avg_stack=None, resp=None, cm=YlOrRd):
brain = np.float32(ref)
brain = img_to_uint8(brain)
brain = grey2rgb(brain)
if resp is None and avg_stack is not None:
resp = find_resp(avg_stack)
elif resp is None and avg_stack is None:
raise ValueError('No suitable data provided')
fresp = gauss_filt(resp, 2)
rgb_resp = cm(fresp)[..., :3]
rgb_resp = gauss_filt(rgb_resp, 2)
rgb_resp = img_to_uint8(rgb_resp)
rgb_resp[resp <= 0, :] = 0
overlaid = np.uint8(.6 * brain + .4 * rgb_resp)
return overlaid
def _make_frame(self, t):
full_time = np.linspace(0, self._duration, self._movie_stack.shape[2])
ix = np.intp(np.argmin(np.abs(full_time - t)))
c_frame = self._movie_stack[..., ix]
c_frame = gauss_filt(c_frame, 5)
return viridis(c_frame, alpha=None, bytes=True)[..., :3]
def norm_pic(self, pic):
b_pic = pic.copy()
b_pic = np.clip(b_pic, None, self.max_slider.value()/1000)
b_pic = gauss_filt(b_pic, 3)
divider = (b_pic.max() - b_pic.min())
if divider == 0:
divider = 1
b_pic = 255 * (b_pic - b_pic.min()) / divider
b_pic = np.uint8(b_pic)
return b_pic
def export_response(self):
# resp = np.mean(self.norm_stack[..., self.n_baseline:(self.n_baseline+30)], 2)
if 'resp_map' not in self.file['df']:
self.resp_mapping()
resp = self.file['df']['resp_map'][()]
# resp = np.clip(resp, 0, 2/100)
# resp = exposure.equalize_hist(resp)
resp = viridis(gauss_filt(resp, 3))[..., :3]
resp[self.max_project <= 0, :] = 0
# resp = 255 * (resp - resp.min()) / (resp.max() - resp.min())
resp = img_to_uint8(resp)
# resp = np.uint8(resp)
imsave(self.path.parent / f'response_{self.path.name}.png', resp)
def normalize_stack(stack, n_baseline=30):
start = 10
# Global average
y = np.nanmean(stack, (0, 1))[start:n_baseline]
y_min, y_max = y.min(), y.max()
# Exponential fit during baseline
t = np.arange(start, n_baseline)
z = 1 + (y - y_min) / (y_max - y_min)
p = np.polyfit(t, np.log(z), 1)
# Modeled decay
full_t = np.arange(stack.shape[2])
decay = np.exp(p[1]) * np.exp(full_t * p[0])
# Renormalized
decay = (decay - 1) * (y_max - y_min) + y_min
norm_stack = stack - decay
norm_stack = gauss_filt(norm_stack, 3, multichannel=False)
return norm_stack
def resp(self):
return gauss_filt(self.file['df']['max_project'][()], 3)
