import matplotlib.cm
import numpy as np
import pyqtgraph as pg
from PyQt5 import QtGui, QtCore
from matplotlib.colors import hsv_to_rgb
import suite2p.gui.merge
from . import io
def make_buttons(parent, b0):
""" color buttons at row b0 """
# color buttons
parent.color_names = [
"A: random",
"S: skew",
"D: compact",
"F: footprint",
"G: aspect_ratio",
"H: chan2_prob",
"J: classifier, cell prob=",
"K: correlations, bin=",
"L: corr with 1D var, bin=^^^",
"M: rastermap / custom"
]
parent.colorbtns = QtGui.QButtonGroup(parent)
clabel = QtGui.QLabel(parent)
clabel.setText("<font color='white'>Colors</font>")
clabel.setFont(parent.boldfont)
parent.l0.addWidget(clabel, b0, 0, 1, 1)
iwid = 65
# add colormaps
parent.CmapChooser = QtGui.QComboBox()
cmaps = ['hsv', 'viridis', 'plasma', 'inferno', 'magma', 'cividis',
'viridis_r', 'plasma_r', 'inferno_r', 'magma_r', 'cividis_r']
parent.CmapChooser.addItems(cmaps)
parent.CmapChooser.setCurrentIndex(0)
parent.CmapChooser.activated.connect(lambda: cmap_change(parent))
parent.CmapChooser.setFont(QtGui.QFont("Arial", 8))
parent.CmapChooser.setFixedWidth(iwid)
parent.l0.addWidget(parent.CmapChooser, b0, 1, 1, 1)
nv = b0
b = 0
# colorbars for different statistics
colorsAll = parent.color_names.copy()
for names in colorsAll:
btn = ColorButton(b, "&" + names, parent)
parent.colorbtns.addButton(btn, b)
if b>4 and b<8:
parent.l0.addWidget(btn, nv + b + 1, 0, 1, 1)
else:
parent.l0.addWidget(btn, nv + b + 1, 0, 1, 2)
btn.setEnabled(False)
parent.color_names[b] = parent.color_names[b][3:]
b += 1
parent.chan2edit = QtGui.QLineEdit(parent)
parent.chan2edit.setText("0.6")
parent.chan2edit.setFixedWidth(iwid)
parent.chan2edit.setAlignment(QtCore.Qt.AlignRight)
parent.chan2edit.returnPressed.connect(lambda: chan2_prob(parent))
parent.l0.addWidget(parent.chan2edit, nv + b - 4, 1, 1, 1)
parent.probedit = QtGui.QLineEdit(parent)
parent.probedit.setText("0.5")
parent.probedit.setFixedWidth(iwid)
parent.probedit.setAlignment(QtCore.Qt.AlignRight)
parent.probedit.returnPressed.connect(
lambda: suite2p.gui.merge.apply(parent)
)
parent.l0.addWidget(parent.probedit, nv + b - 3, 1, 1, 1)
parent.binedit = QtGui.QLineEdit(parent)
parent.binedit.setValidator(QtGui.QIntValidator(0, 500))
parent.binedit.setText("1")
parent.binedit.setFixedWidth(iwid)
parent.binedit.setAlignment(QtCore.Qt.AlignRight)
parent.binedit.returnPressed.connect(
lambda: parent.mode_change(parent.activityMode)
)
parent.l0.addWidget(parent.binedit, nv + b - 2, 1, 1, 1)
b0 = nv+b+2
return b0
def cmap_change(parent):
index = parent.CmapChooser.currentIndex()
parent.ops_plot['colormap'] = parent.CmapChooser.itemText(index)
if parent.loaded:
print('colormap changed to %s, loading...'%parent.ops_plot['colormap'])
istat = parent.colors['istat']
for c in range(1, istat.shape[0]):
parent.colors['cols'][c] = istat_transform(istat[c], parent.ops_plot['colormap'])
rgb_masks(parent, parent.colors['cols'][c], c)
parent.colormat = draw_colorbar(parent.ops_plot['colormap'])
parent.update_plot()
def hsv2rgb(cols):
cols = cols[:,np.newaxis]
cols = np.concatenate((cols, np.ones_like(cols), np.ones_like(cols)), axis=-1)
cols = (255 * hsv_to_rgb(cols)).astype(np.uint8)
return cols
def make_colors(parent):
parent.colors['colorbar'] = []
ncells = len(parent.stat)
parent.colors['cols'] = np.zeros((len(parent.color_names), ncells, 3), np.uint8)
parent.colors['istat'] = np.zeros((len(parent.color_names), ncells), np.float32)
np.random.seed(seed=0)
allcols = np.random.random((ncells,))
if 'meanImg_chan2' in parent.ops:
allcols = allcols / 1.4
allcols = allcols + 0.1
print(parent.redcell.sum())
parent.randcols = allcols
allcols[parent.redcell] = 0
else:
parent.randcols = allcols
parent.colors['istat'][0] = parent.randcols
parent.colors['cols'][0] = hsv2rgb(parent.randcols)
b=0
for names in parent.color_names[:-3]:
if b > 0:
istat = np.zeros((ncells,1))
if b<len(parent.color_names)-2:
if names in parent.stat[0]:
for n in range(0,ncells):
istat[n] = parent.stat[n][names]
istat1 = np.percentile(istat,2)
istat99 = np.percentile(istat,98)
parent.colors['colorbar'].append([istat1,
(istat99-istat1)/2 + istat1,
istat99])
istat = istat - istat1
istat = istat / (istat99-istat1)
istat = np.maximum(0, np.minimum(1, istat))
else:
istat = np.expand_dims(parent.probcell, axis=1)
parent.parent.colors['colorbar'].append([0.0, .5, 1.0])
col = istat_transform(istat, parent.ops_plot['colormap'])
parent.colors['cols'][b] = col
parent.colors['istat'][b] = istat.flatten()
else:
parent.colors['colorbar'].append([0,0.5,1])
b+=1
parent.colors['colorbar'].append([0,0.5,1])
parent.colors['colorbar'].append([0,0.5,1])
parent.colors['colorbar'].append([0,0.5,1])
#parent.ops_plot[4] = corrcols
#parent.cc = cc
def flip_plot(parent):
parent.iflip = parent.ichosen
for n in parent.imerge:
iscell = int(parent.iscell[n])
parent.iscell[n] = ~parent.iscell[n]
parent.ichosen = n
flip_roi(parent)
if 'imerge' in parent.stat[n]:
for k in parent.stat[n]['imerge']:
parent.iscell[k] = ~parent.iscell[k]
parent.update_plot()
io.save_iscell(parent)
def chan2_prob(parent):
chan2prob = float(parent.chan2edit.text())
if abs(parent.chan2prob - chan2prob) > 1e-3:
parent.chan2prob = chan2prob
parent.redcell = parent.probredcell > parent.chan2prob
chan2_masks(parent)
parent.update_plot()
io.save_redcell(parent)
def make_colorbar(parent, b0):
colorbarW = pg.GraphicsLayoutWidget(parent)
colorbarW.setMaximumHeight(60)
colorbarW.setMaximumWidth(150)
colorbarW.ci.layout.setRowStretchFactor(0, 2)
colorbarW.ci.layout.setContentsMargins(0, 0, 0, 0)
parent.l0.addWidget(colorbarW, b0, 0, 1, 2)
parent.colorbar = pg.ImageItem()
cbar = colorbarW.addViewBox(row=0, col=0, colspan=3)
cbar.setMenuEnabled(False)
cbar.addItem(parent.colorbar)
parent.clabel = [
colorbarW.addLabel("0.0", color=[255, 255, 255], row=1, col=0),
colorbarW.addLabel("0.5", color=[255, 255, 255], row=1, col=1),
colorbarW.addLabel("1.0", color=[255, 255, 255], row=1, col=2),
]
def init_masks(parent):
"""
creates RGB masks using stat and puts them in M0 or M1 depending on
whether or not iscell is True for a given ROI
args:
ops: mean_image, Vcorr
stat: xpix,ypix,xext,yext
iscell: vector with True if ROI is cell
ops_plot: plotROI, view, color, randcols
outputs:
M0: ROIs that are True in iscell
M1: ROIs that are False in iscell
"""
stat = parent.stat
iscell = parent.iscell
cols = parent.colors['cols']
ncells = len(stat)
Ly = parent.Ly
Lx = parent.Lx
parent.rois['Sroi'] = np.zeros((2,Ly,Lx), np.bool)
LamAll = np.zeros((Ly,Lx), np.float32)
# these have 3 layers
parent.rois['Lam'] = np.zeros((2,3,Ly,Lx), np.float32)
parent.rois['iROI'] = -1 * np.ones((2,3,Ly,Lx), np.int32)
# ignore merged cells
iignore = np.zeros(ncells, np.bool)
for n in np.arange(ncells-1,-1,-1,int):
ypix = stat[n]['ypix']
if ypix is not None and not iignore[n]:
if 'imerge' in stat[n]:
for k in stat[n]['imerge']:
iignore[k] = True
print(k)
xpix = stat[n]['xpix']
lam = stat[n]['lam']
lam = lam / lam.sum()
i = int(1-iscell[n])
# add cell on top
parent.rois['iROI'][i,2,ypix,xpix] = parent.rois['iROI'][i,1,ypix,xpix]
parent.rois['iROI'][i,1,ypix,xpix] = parent.rois['iROI'][i,0,ypix,xpix]
parent.rois['iROI'][i,0,ypix,xpix] = n
# add weighting to all layers
parent.rois['Lam'][i,2,ypix,xpix] = parent.rois['Lam'][i,1,ypix,xpix]
parent.rois['Lam'][i,1,ypix,xpix] = parent.rois['Lam'][i,0,ypix,xpix]
parent.rois['Lam'][i,0,ypix,xpix] = lam
parent.rois['Sroi'][i,ypix,xpix] = 1
LamAll[ypix,xpix] = lam
parent.rois['LamMean'] = LamAll[LamAll>1e-10].mean()
parent.rois['LamNorm'] = np.maximum(0, np.minimum(1, 0.75*parent.rois['Lam'][:,0]/parent.rois['LamMean']))
parent.colors['RGB'] = np.zeros((2,cols.shape[0],Ly,Lx,4), np.uint8)
for c in range(0, cols.shape[0]):
rgb_masks(parent, cols[c], c)
def rgb_masks(parent, col, c):
for i in range(2):
#S = np.expand_dims(parent.rois['Sroi'][i],axis=2)
H = col[parent.rois['iROI'][i,0], :]
#H = np.expand_dims(H,axis=2)
#hsv = np.concatenate((H,S,S),axis=2)
#rgb = (hsv_to_rgb(hsv)*255).astype(np.uint8)
parent.colors['RGB'][i,c,:,:,:3] = H
def draw_masks(parent): #ops, stat, ops_plot, iscell, ichosen):
'''
creates RGB masks using stat and puts them in M0 or M1 depending on
whether or not iscell is True for a given ROI
args:
ops: mean_image, Vcorr
stat: xpix,ypix
iscell: vector with True if ROI is cell
ops_plot: plotROI, view, color, randcols
outputs:
M0: ROIs that are True in iscell
M1: ROIs that are False in iscell
'''
ncells = parent.iscell.shape[0]
plotROI = parent.ops_plot['ROIs_on']
view = parent.ops_plot['view']
color = parent.ops_plot['color']
opacity = parent.ops_plot['opacity']
wplot = int(1-parent.iscell[parent.ichosen])
# reset transparency
for i in range(2):
parent.colors['RGB'][i,color,:,:,3] = (opacity[view==0] *
parent.rois['Sroi'][i] *
parent.rois['LamNorm'][i]).astype(np.uint8)
M = [np.array(parent.colors['RGB'][0,color]), np.array(parent.colors['RGB'][1,color])]
if view==0:
for n in parent.imerge:
ypix = parent.stat[n]['ypix'].flatten()
xpix = parent.stat[n]['xpix'].flatten()
v = (parent.rois['iROI'][wplot][:,ypix,xpix]>-1).sum(axis=0) - 1
v = 1 - v/3
M[wplot] = make_chosen_ROI(M[wplot], ypix, xpix, v)
else:
for n in parent.imerge:
ycirc = parent.stat[n]['ycirc']
xcirc = parent.stat[n]['xcirc']
ypix = parent.stat[n]['ypix'].flatten()
xpix = parent.stat[n]['xpix'].flatten()
M[wplot][ypix,xpix,3] = 0
col = parent.colors['cols'][color,n]
sat = 1
M[wplot] = make_chosen_circle(M[wplot], ycirc, xcirc, col, sat)
return M[0],M[1]
def make_chosen_ROI(M0, ypix, xpix, v):
M0[ypix,xpix,:] = np.tile((255*v[:,np.newaxis]).astype(np.uint8), (1,4))
return M0
def make_chosen_circle(M0, ycirc, xcirc, col, sat):
ncirc = ycirc.size
M0[ycirc,xcirc,:3] = col#[np.newaxis,:]
M0[ycirc,xcirc,3] = 255
return M0
def chan2_masks(parent):
c = 0
col = parent.randcols
col[parent.redcell] = 0
col = col.flatten()
parent.colors['cols'][c] = hsv2rgb(col)
rgb_masks(parent, col, c)
def custom_masks(parent):
c = 9
n = np.array(parent.imerge)
istat = parent.custom_mask
istat1 = istat.min()
istat99 = istat.max()
cl = [istat1, (istat99-istat1)/2 + istat1, istat99]
istat = istat - istat1
istat = istat / (istat99-istat1)
istat = np.maximum(0, np.minimum(1, istat))
parent.colors['colorbar'][c] = cl
istat = istat / istat.max()
col = istat_transform(istat, parent.ops_plot['colormap'])
parent.colors['cols'][c] = col
parent.colors['istat'][c] = istat.flatten()
rgb_masks(parent, col, c)
def rastermap_masks(parent):
c = 9
n = np.array(parent.imerge)
istat = parent.isort
# no 1D variable loaded -- leave blank
parent.colors['colorbar'][c] = ([0, istat.max()/2, istat.max()])
istat = istat / istat.max()
col = istat_transform(istat, parent.ops_plot['colormap'])
col[parent.isort==-1] = 0
parent.colors['cols'][c] = col
parent.colors['istat'][c] = istat.flatten()
rgb_masks(parent, col, c)
def beh_masks(parent):
c = 8
n = np.array(parent.imerge)
nb = int(np.floor(parent.beh_resampled.size/parent.bin))
sn = np.reshape(parent.beh_resampled[:nb*parent.bin], (nb,parent.bin)).mean(axis=1)
sn -= sn.mean()
snstd = (sn**2).mean()**0.5
cc = np.dot(parent.Fbin, sn.T) / parent.Fbin.shape[-1] / (parent.Fstd * snstd)
cc[n] = cc.mean()
istat = cc
inactive=False
istat_min = istat.min()
istat_max = istat.max()
istat = istat - istat.min()
istat = istat / istat.max()
col = istat_transform(istat, parent.ops_plot['colormap'])
parent.colors['cols'][c] = col
parent.colors['istat'][c] = istat.flatten()
parent.colors['colorbar'][c] = [istat_min,
(istat_max-istat_min)/2 + istat_min,
istat_max]
rgb_masks(parent, col, c)
def corr_masks(parent):
c = 7
n = np.array(parent.imerge)
sn = parent.Fbin[n].mean(axis=-2).squeeze()
snstd = (sn**2).mean()**0.5
cc = np.dot(parent.Fbin, sn.T) / parent.Fbin.shape[-1] / (parent.Fstd * snstd)
cc[n] = cc.mean()
istat = cc
parent.colors['colorbar'][c] = [istat.min(),
(istat.max()-istat.min())/2 + istat.min(),
istat.max()]
istat = istat - istat.min()
istat = istat / istat.max()
col = istat_transform(istat, parent.ops_plot['colormap'])
parent.colors['cols'][c] = col
parent.colors['istat'][c] = istat.flatten()
rgb_masks(parent, col, c)
def flip_for_class(parent, iscell):
ncells = iscell.size
if (iscell==parent.iscell).sum() < 100:
for n in range(ncells):
if iscell[n] != parent.iscell[n]:
parent.iscell[n] = iscell[n]
parent.ichosen = n
flip_roi(parent)
else:
parent.iscell = iscell
init_masks(parent)
def plot_colorbar(parent):
bid = parent.ops_plot['color']
if bid==0:
parent.colorbar.setImage(np.zeros((20,100,3)))
else:
parent.colorbar.setImage(parent.colormat)
for k in range(3):
parent.clabel[k].setText('%1.2f'%parent.colors['colorbar'][bid][k])
def plot_masks(parent, M):
#M = parent.RGB[:,:,np.newaxis], parent.Alpha[]
parent.color1.setImage(M[0], levels=(0., 255.))
parent.color2.setImage(M[1], levels=(0., 255.))
parent.color1.show()
parent.color2.show()
def remove_roi(parent, n, i0):
"""
removes roi n from view i0
"""
ypix = parent.stat[n]['ypix']
xpix = parent.stat[n]['xpix']
# cell indices
ipix = np.array((parent.rois['iROI'][i0,0,:,:]==n).nonzero()).astype(np.int32)
ipix1 = np.array((parent.rois['iROI'][i0,1,:,:]==n).nonzero()).astype(np.int32)
ipix2 = np.array((parent.rois['iROI'][i0,2,:,:]==n).nonzero()).astype(np.int32)
# get rid of cell and push up overlaps on main views
parent.rois['Lam'][i0,0,ipix[0,:],ipix[1,:]] = parent.rois['Lam'][i0,1,ipix[0,:],ipix[1,:]]
parent.rois['Lam'][i0,1,ipix[0,:],ipix[1,:]] = 0
parent.rois['Lam'][i0,1,ipix1[0,:],ipix1[1,:]] = parent.rois['Lam'][i0,2,ipix1[0,:],ipix1[1,:]]
parent.rois['Lam'][i0,2,ipix1[0,:],ipix1[1,:]] = 0
parent.rois['Lam'][i0,2,ipix2[0,:],ipix2[1,:]] = 0
parent.rois['iROI'][i0,0,ipix[0,:],ipix[1,:]] = parent.rois['iROI'][i0,1,ipix[0,:],ipix[1,:]]
parent.rois['iROI'][i0,1,ipix[0,:],ipix[1,:]] = -1
parent.rois['iROI'][i0,1,ipix1[0,:],ipix1[1,:]] = parent.rois['iROI'][i0,2,ipix1[0,:],ipix1[1,:]]
parent.rois['iROI'][i0,2,ipix1[0,:],ipix1[1,:]] = -1
parent.rois['iROI'][i0,2,ipix2[0,:],ipix2[1,:]] = -1
# remove +/- 1 ROI exists
parent.rois['Sroi'][i0,ypix,xpix] = parent.rois['iROI'][i0,0,ypix,xpix] > 0
parent.rois['LamNorm'][i0,ypix,xpix] = np.maximum(0, np.minimum(1,
0.75*parent.rois['Lam'][i0,0,ypix,xpix]/parent.rois['LamMean']))
def add_roi(parent, n, i):
"""
add roi n to view i
"""
ypix = parent.stat[n]['ypix']
xpix = parent.stat[n]['xpix']
lam = parent.stat[n]['lam']
parent.rois['iROI'][i,2,ypix,xpix] = parent.rois['iROI'][i,1,ypix,xpix]
parent.rois['iROI'][i,1,ypix,xpix] = parent.rois['iROI'][i,0,ypix,xpix]
parent.rois['iROI'][i,0,ypix,xpix] = n
parent.rois['Lam'][i,2,ypix,xpix] = parent.rois['Lam'][i,1,ypix,xpix]
parent.rois['Lam'][i,1,ypix,xpix] = parent.rois['Lam'][i,0,ypix,xpix]
parent.rois['Lam'][i,0,ypix,xpix] = lam #/ lam.sum()
# set whether or not an ROI + weighting of pixels
parent.rois['Sroi'][i,ypix,xpix] = 1
parent.rois['LamNorm'][:,ypix,xpix] = np.maximum(0, np.minimum(1, 0.75*parent.rois['Lam'][:,0,ypix,xpix]/parent.rois['LamMean']))
def redraw_masks(parent, ypix, xpix):
"""
redraw masks after roi added/removed
"""
for c in range(parent.colors['cols'].shape[0]):
for i in range(2):
col = parent.colors['cols'][c]
rgb = col[parent.rois['iROI'][i,0,ypix,xpix],:]
parent.colors['RGB'][i,c,ypix,xpix,:3] = rgb
def flip_roi(parent):
"""
flips roi to other plot
there are 3 levels of overlap so this may be buggy if more than 3 cells are on
top of each other
"""
cols = parent.ops_plot['color']
n = parent.ichosen
i = int(1-parent.iscell[n])
i0 = 1-i
# remove ROI
remove_roi(parent, n, i0)
# add cell to other side (on top) and push down overlaps
add_roi(parent, n, i)
# redraw colors
ypix = parent.stat[n]['ypix']
xpix = parent.stat[n]['xpix']
redraw_masks(parent, ypix, xpix)
def draw_colorbar(colormap='hsv'):
H = np.linspace(0,1,101).astype(np.float32)
rgb = istat_transform(H, colormap)
colormat = np.expand_dims(rgb, axis=0)
colormat = np.tile(colormat,(20,1,1))
return colormat
def istat_hsv(istat):
istat = istat / 1.4
istat = istat + (0.4/1.4)
icols = 1 - istat
icols = hsv2rgb(icols.flatten())
return icols
def istat_transform(istat, colormap='hsv'):
if colormap=='hsv':
icols = istat_hsv(istat)
else:
try:
cmap = matplotlib.cm.get_cmap(colormap)
icols = istat
icols = cmap(icols)[:,:3]
icols *= 255
icols = icols.astype(np.uint8)
except:
print('bad colormap, using hsv')
icols = istat_hsv(istat)
return icols
### Changes colors of ROIs
# button group is exclusive (at least one color is always chosen)
class ColorButton(QtGui.QPushButton):
def __init__(self, bid, Text, parent=None):
super(ColorButton,self).__init__(parent)
self.setText(Text)
self.setCheckable(True)
self.setStyleSheet(parent.styleInactive)
self.setFont(QtGui.QFont("Arial", 8, QtGui.QFont.Bold))
self.resize(self.minimumSizeHint())
self.clicked.connect(lambda: self.press(parent, bid))
self.show()
def press(self, parent, bid):
for b in range(len(parent.color_names)):
if parent.colorbtns.button(b).isEnabled():
parent.colorbtns.button(b).setStyleSheet(parent.styleUnpressed)
self.setStyleSheet(parent.stylePressed)
parent.ops_plot['color'] = bid
if not parent.sizebtns.button(1).isChecked():
if bid==0:
for b in [1,2]:
parent.topbtns.button(b).setEnabled(False)
parent.topbtns.button(b).setStyleSheet(parent.styleInactive)
else:
for b in [1,2]:
parent.topbtns.button(b).setEnabled(True)
parent.topbtns.button(b).setStyleSheet(parent.styleUnpressed)
else:
for b in range(3):
parent.topbtns.button(b).setEnabled(False)
parent.topbtns.button(b).setStyleSheet(parent.styleInactive)
parent.update_plot()
parent.show()
