#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on May 13 2018
@author: kushal
Chatzigeorgiou Group
Sars International Centre for Marine Molecular Biology
GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007
"""
from ....viewer.modules.roi_manager_modules.roi_list import ROIList
from .roi_types import *
from ...core.common import ViewerUtils
from .... import pyqtgraphCore as pg
from copy import deepcopy
from .read_imagej import read_roi_zip as read_imagej
from ....common.configuration import HAS_CAIMAN
from matplotlib import cm as matplotlib_color_map
from tqdm import tqdm
import logging
if HAS_CAIMAN:
from caiman.utils.visualization import get_contours as caiman_get_contours
from caiman.source_extraction.cnmf.cnmf import load_CNMF
logger = logging.getLogger()
[docs]class AbstractBaseManager(metaclass=abc.ABCMeta):
"""Base ROI Manager"""
[docs] def __init__(self, parent, ui, viewer_interface: ViewerUtils):
"""
Set the common attributes
:param parent: The ModuleGUI QDockWidget instance
:param ui: The ui of the ModuleGUI QDockWidget instance,
:param viewer_interface: A ViewerUtils instance for accessing the Viewer the parent QDockWidget belongs to
"""
self.ui = ui
self.vi = viewer_interface
self.roi_list = None #: The ROIList instance that stores the list of ROIs
self.parent = parent
self.metadata = None
[docs] @abc.abstractmethod
def add_roi(self, *args, **kwargs):
"""Method for adding an ROI, must be implemented in subclass"""
pass
[docs] def is_empty(self) -> bool:
"""Return true if the ROI list is empty, else return False"""
if not hasattr(self, 'roi_list'):
return True
if self.roi_list is None:
return True
if len(self.roi_list) < 1:
return True
else:
return False
[docs] def get_all_states(self) -> dict:
"""
Get the ROI states for all ROIs in self.roi_list so that they can be restored.
The appropriate manager is instantiated based on the 'roi_type' key of the returned dict
"""
self.vi.viewer.status_bar_label.showMessage('Saving ROIs...')
# the key 'roi_type' determines which Manager subclass should be used, and 'states' are the actual ROI states
states = {
'roi_type': self.roi_list.roi_types.__name__,
'states': [],
'metadata': self.metadata
}
for roi in self.roi_list:
state = roi.to_state()
states['states'].append(state)
self.vi.viewer.status_bar_label.showMessage('ROIs saved!')
return states
def restore_from_states(self, states: dict):
if 'metadata' in states.keys():
self.metadata = states['metadata']
[docs] def get_plot_item(self) -> pg.PlotDataItem:
"""Get the viewer plot item that is associated to these ROIs"""
return self.vi.viewer.ui.roiPlot.plot()
[docs] def clear(self):
"""Cleanup of all ROIs in the list"""
if not hasattr(self, 'roi_list'):
return
self.roi_list.clear_()
del self.roi_list
[docs] def __del__(self):
"""Cleanup of all ROIs in the list and deletes the manager instance. Used when switching modes."""
self.clear()
# self.roi_list.list_widget.clear()
# self.roi_list.list_widget_tags.clear()
# self.roi_list.disconnect_all()
# for i in range(len(self.roi_list)):
[docs]class ManagerManual(AbstractBaseManager):
"""The Manager for the Manual mode"""
[docs] def __init__(self, parent, ui, viewer_interface):
super(ManagerManual, self).__init__(parent, ui, viewer_interface)
self.create_roi_list()
[docs] def create_roi_list(self):
"""Create a new empty ROI list instance for storing Manual ROIs"""
self.roi_list = ROIList(self.ui, ManualROI, self.vi)
[docs] def add_roi(self, shape: str) -> ManualROI:
"""
Add an ROI to the list
:param shape: either "PolyLineROI" or "EllipseROI"
"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
dims = self.vi.viewer.workEnv.imgdata.seq.shape
roi_graphics_object = ManualROI.get_generic_roi_graphics_object(shape, dims)
roi = ManualROI(self.get_plot_item(), roi_graphics_object, self.vi.viewer.getView())
self.roi_list.append(roi)
self.roi_list.reindex_colormap()
self.vi.workEnv_changed("ROIs imported")
return roi
[docs] def restore_from_states(self, states: dict):
"""Restore ROIs from states"""
super(ManagerManual, self).restore_from_states(states)
if not hasattr(self, 'roi_list'):
self.create_roi_list()
for state in states['states']:
roi = ManualROI.from_state(self.get_plot_item(), self.vi.viewer.getView(), state)
self.roi_list.append(roi)
# ix = 0
# for state in states['states']:
# self.roi_list[ix].set_roi_graphics_object_state(state['roi_graphics_object_state'])
# ix += 1
self.roi_list.reindex_colormap()
[docs] def get_all_states(self) -> dict:
"""Get the ROI states so that they can be restored later"""
self.vi.viewer.status_bar_label.showMessage('Saving ROIs...')
states = {
'roi_type': self.roi_list.roi_types.__name__,
'states': [],
'metadata': self.metadata
}
for ix in range(len(self.roi_list)):
self.roi_list.set_pg_roi_plot(ix)
state = self.roi_list[ix].to_state()
states['states'].append(state)
self.vi.viewer.status_bar_label.showMessage('Finished saving ROIs!')
return states
[docs] def import_from_imagej(self, path: str):
"""
Uses read-roi package created by Hadrien Mary.
https://pypi.org/project/read-roi/
:param path: Full path to the ImageJ ROIs zip file
"""
ij_roi = read_imagej(path)
for k in ij_roi.keys():
xs = ij_roi[k]['x']
ys = ij_roi[k]['y']
ps = list(zip(xs, ys))
roi = ManualROI.from_positions(positions=ps,
curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView())
self.roi_list.append(roi)
self.roi_list.reindex_colormap()
def add_roi_from_points(self, xs, ys):
ps = list(zip(xs, ys))
roi = ManualROI.from_positions(
positions=ps,
curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView()
)
self.roi_list.append(roi)
self.roi_list.reindex_colormap()
[docs]class ManagerScatterROI(AbstractBaseManager):
"""Manager for unmoveable ROIs drawn using scatterplots"""
[docs] def __init__(self, parent, ui, viewer_interface: ViewerUtils):
super(ManagerScatterROI, self).__init__(parent, ui, viewer_interface)
self.create_roi_list()
self.list_widget = self.roi_list.list_widget
[docs] def add_roi(self, curve: np.ndarray,
xs: np.ndarray,
ys: np.ndarray,
metadata: dict = None,
dfof_data: np.ndarray = None,
spike_data: np.ndarray = None) \
-> ScatterROI:
"""
Add a single ROI
`xs` and `ys` arguments are 1D numpy arrays.
:param curve: curve data, 1-D array, y values/intensity values
:param xs: x-values for the scatter plot to spatially illustrate the ROI
:param ys: corresponding y-values for the scatter plot to spatially illustrate the ROI
:param metadata: Any metadata for this ROI
:return: ScatterROI object
"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
roi = ScatterROI(
curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView(),
curve_data=curve,
xs=xs,
ys=ys,
dfof_data=dfof_data,
spike_data=spike_data
)
roi.metadata = metadata
self.roi_list.append(roi)
self.roi_list.reindex_colormap()
self.vi.workEnv_changed("ROI Added")
return roi
[docs] def restore_from_states(self, states: dict):
"""Restore from states, such as when these ROIs are saved with a Project Sample"""
super(ManagerScatterROI, self).restore_from_states(states)
if not hasattr(self, 'roi_list'):
self.create_roi_list()
for state in states['states']:
roi = ScatterROI.from_state(self.get_plot_item(), self.vi.viewer.getView(), state)
self.roi_list.append(roi)
self.roi_list.reindex_colormap()
[docs] def create_roi_list(self):
"""Create empty ROI List"""
self.roi_list = ROIList(self.ui, ScatterROI, self.vi)
[docs] def set_spot_size(self, size: int):
"""Set the spot size for the scatter plot which illustrates the ROI"""
for roi in self.roi_list:
roi.get_roi_graphics_object().setSize(size)
roi.spot_size = size
[docs]class ManagerVolROI(ManagerScatterROI):
"""Manager for 3D ROIs"""
[docs] def __init__(self, parent, ui, viewer_interface: ViewerUtils):
super(ManagerVolROI, self).__init__(parent, ui, viewer_interface)
self.vi.viewer.sigZLevelChanged.connect(self.set_zlevel)
[docs] def set_zlevel(self, z: int):
"""Set the current z-level to be visible in the viewer"""
if not hasattr(self, 'roi_list'):
warn('roi list does not exist, probably empty work environment')
return
for roi in self.roi_list:
roi.set_zlevel(z)
[docs] def create_roi_list(self):
"""Create new empty ROI list"""
self.roi_list = ROIList(self.ui, VolCNMF, self.vi)
[docs]class ManagerVolCNMF(ManagerVolROI):
"""Manager for 3D CNMF based ROIs"""
[docs] def __init__(self, parent, ui, viewer_interface):
super(ManagerVolCNMF, self).__init__(parent, ui, viewer_interface)
self.roi_list = ROIList(self.ui, VolCNMF, self.vi)
self.input_params_dict = None
self.idx_components = None # Keep track of components if the user manually want to remove some
self.orig_idx_components = None # List of components prior to any manual deletion by the user
# cnmf data dict directly from the hdf5 file
self.cnmf_data_dict = None
# These correspond to the caiman.source_extraction.cnmf attributes
self.cnmA = None
self.cnmb = None
self.cnmC = None
self.cnm_f = None
self.cnmYrA = None
[docs] def create_roi_list(self):
self.roi_list = ROIList(self.ui, VolCNMF, self.vi)
[docs] def add_all_components(self, cnmf_data_dict: dict, input_params_dict: dict):
"""
Add all components from a CNMF(E) output. Arguments correspond to CNMF(E) outputs
:param cnmf_data_dict: CNMF results data directly from the HDF5 file
:param input_params_dict: dict of input params, from the batch manager
:param calc_raw_min_max: Calculate raw min & max for each ROI
:return:
"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
self.cnmf_data_dict = cnmf_data_dict
# self.cnmf_obj = load_CNMF(self.cnmf_data_dict)
self.cnmA = self.cnmf_data_dict['estimates']['A']
self.cnmb = self.cnmf_data_dict['estimates']['b']
self.cnm_f = self.cnmf_data_dict['estimates']['f']
self.cnmC = self.cnmf_data_dict['estimates']['C']
self.cnmYrA = self.cnmf_data_dict['estimates']['YrA']
self.dims = self.cnmf_data_dict['dims']
self.cnmS = self.cnmf_data_dict['estimates']['S']
self.cnm_dfof = self.cnmf_data_dict['estimates']['F_dff']
# components are already filtered from the output file
self.idx_components = np.arange(self.cnmC.shape[0])
self.orig_idx_components = deepcopy(self.idx_components)
self.input_params_dict = input_params_dict
# spatial components
contours = caiman_get_contours(self.cnmA[:, self.idx_components], self.dims, thr=0.9)
temporal_components = self.cnmC
self.input_params_dict = self.input_params_dict
num_components = len(temporal_components)
self.ui.radioButton_curve_data.setChecked(True)
for ix in range(num_components):
self.vi.viewer.status_bar_label.showMessage('Please wait, adding component #: '
+ str(ix) + ' / ' + str(num_components))
curve_data = temporal_components[self.idx_components[ix]]
contour = contours[ix]
roi = VolCNMF(curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView(),
cnmf_idx=self.idx_components[ix],
curve_data=curve_data,
contour=contour,
dfof_data=self.cnm_dfof[ix] if (self.cnm_dfof is not None) else None,
spike_data=self.cnmS[ix])
self.roi_list.append(roi)
self.vi.workEnv_changed("ROIs imported")
self.roi_list.reindex_colormap(random_shuffle=True)
self.vi.viewer.status_bar_label.showMessage('Finished adding all components!')
[docs] def add_roi(self):
"""Not implemented, uses add_all_components to import all ROIs instead"""
raise NotImplementedError('Not implemented for CNMFE ROIs')
[docs] def restore_from_states(self, states: dict):
"""Restore from states, such as when these ROIs are saved with a Project Sample"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
self.cnmf_data_dict = states['cnmf_data_dict']
for state in states['states']:
roi = VolCNMF.from_state(self.get_plot_item(), self.vi.viewer.getView(), state)
self.roi_list.append(roi)
self.input_params_dict = states['input_params_cnmfe']
self.cnmA = states['cnmf_output']['cnmA']
self.cnmb = states['cnmf_output']['cnmb']
self.cnmC = states['cnmf_output']['cnmC']
self.cnm_f = states['cnmf_output']['cnm_f']
self.cnmS = self.cnmf_data_dict['estimates']['S']
self.cnm_dfof = self.cnmf_data_dict['estimates']['F_dff']
self.cnmYrA = states['cnmf_output']['cnmYrA']
self.idx_components = states['cnmf_output']['idx_components']
self.orig_idx_components = states['cnmf_output']['orig_idx_components']
self.roi_list.reindex_colormap()
[docs] def get_all_states(self) -> dict:
"""Get all states so that they can be restored"""
states = super(ManagerVolROI, self).get_all_states()
# If the user has manually deleted some ROIs
new_idx_components = np.array([roi.cnmf_idx for roi in self.roi_list], dtype=np.int64)
# Make sure nothing weird happened
# l = [self.cnmA, self.cnmb, self.cnmC, self.cnm_f, self.cnmYrA, self.orig_idx_components, new_idx_components]
# if any(item is None for item in l):
# raise ValueError('One or more pieces of CNMF(E) data are missing')
# Store the actual cnmf attributes as well.
input_dict = {'input_params_cnmfe': self.input_params_dict,
'cnmf_data_dict': self.cnmf_data_dict,
'cnmf_output':
{
'cnmA': self.cnmA,
'cnmb': self.cnmb,
'cnmC': self.cnmC,
'cnm_f': self.cnm_f,
'cnmYrA': self.cnmYrA,
'idx_components': new_idx_components,
'orig_idx_components': self.orig_idx_components
}
}
states.update(input_dict)
return states
[docs] def update_idx_components(self, ix: int):
"""Update idx_components if the user manually delete an ROI"""
roi = self.roi_list[self.roi_list.current_index]
self.idx_components = np.delete(self.idx_components, np.where(self.idx_components == roi.cnmf_idx)[0])
[docs] def set_spot_size(self, size: int):
for roi in self.roi_list:
roi.get_roi_graphics_object().setSize(size)
roi.spot_size = size
class ManagerVolMultiCNMFROI(ManagerVolROI):
"""Manager for 3D data using one CNMF per plane"""
def __init__(self, parent, ui, viewer_interface):
super(ManagerVolMultiCNMFROI, self).__init__(parent, ui, viewer_interface)
self.roi_list = ROIList(self.ui, VolMultiCNMFROI, self.vi)
self.input_params_dict: dict = None
self.idx_components: List[np.ndarray] = []
self.orig_idx_components: List[np.ndarray] = []
self.cnmf_data_dicts: List[dict] = []
self.cnmA: List[np.ndarray] = []
self.cnmb: List[np.ndarray] = []
self.cnmC: List[np.ndarray] = []
self.cnm_f: List[np.ndarray] = []
self.cnmYrA: List[np.ndarray] = []
self.cnmS: List[np.ndarray] = []
self.cnm_dfof: List[np.ndarray] = []
self.dims: List[tuple] = []
self.roi_xys: List[np.ndarray] = [] # roi x-y coordinates
self.roi_ixs: List[np.ndarray] = [] # the roi index that each coordinate maps to
self.roi_crs: List[np.ndarray] = [] # the color that each roi index maps to
self.roi_sps: List[pg.ScatterPlotItem] = [] # ROIs represented as scatterplots
self.num_zlevels: int = 0
def create_roi_list(self):
self.roi_list = ROIList(self.ui, VolMultiCNMFROI, self.vi)
def add_all_components(
self,
cnmf_data_dicts: List[dict],
input_params_dict: dict,
):
self.input_params_dict = input_params_dict
if not hasattr(self, 'roi_list'):
self.create_roi_list()
self.cnmf_data_dicts = cnmf_data_dicts
self.num_zlevels = len(self.cnmf_data_dicts)
for zcenter, cnmf_data_dict in enumerate(self.cnmf_data_dicts):
self.cnmA.append(cnmf_data_dict['estimates']['A'])
self.cnmb.append(cnmf_data_dict['estimates']['b'])
self.cnm_f.append(cnmf_data_dict['estimates']['f'])
self.cnmC.append(cnmf_data_dict['estimates']['C'])
self.cnmS.append(cnmf_data_dict['estimates']['S'])
self.cnm_dfof.append(cnmf_data_dict['estimates']['F_dff'])
self.cnmYrA.append(cnmf_data_dict['estimates']['YrA'])
self.dims.append(cnmf_data_dict['dims'])
self.idx_components.append(cnmf_data_dict['estimates']['idx_components'])
if self.idx_components[-1] is None:
self.idx_components[-1] = np.arange(self.cnmC[-1].shape[0])
self.orig_idx_components.append(
deepcopy(
self.idx_components[-1]
)
)
contours = caiman_get_contours(
self.cnmA[-1][:, self.idx_components[-1]],
self.dims[-1],
# swap_dim=True
)
num_components = len(self.idx_components[-1])
self.ui.radioButton_curve_data.setChecked(True)
roi_ixs = []
roi_xy = []
for ix in range(len(contours)):
coors = contours[ix]['coordinates']
coors = coors[~np.isnan(coors).any(axis=1)]
roi_xy += [coors]
roi_ixs += [ix] * coors.shape[0]
roi_xy = np.vstack(roi_xy)
roi_ixs = np.vstack(roi_ixs)
self.roi_xys.append(roi_xy)
self.roi_ixs.append(roi_ixs)
cm = matplotlib_color_map.get_cmap('hsv')
cm._init()
lut = (cm._lut * 255).view(np.ndarray)
cm_ixs = np.linspace(0, 210, np.unique(roi_ixs).size + 1, dtype=int)
roi_crs = []
for roi_ix, cm_ix in zip(np.unique(roi_ixs), cm_ixs):
c = lut[cm_ix]
roi_crs.append(
np.array([c] * roi_ixs[roi_ixs == roi_ix].size) # color for each spot
)
roi_crs = np.vstack(roi_crs)
self.roi_crs.append(roi_crs)
xy_coors = self.roi_xys[-1]
brushes = list(map(pg.mkBrush, roi_crs))
pens = list(map(pg.mkPen, roi_crs))
sp = pg.ScatterPlotItem(
xy_coors[:, 0],
xy_coors[:, 1],
symbol='s',
size=1,
pxMode=True,
brush=brushes,
pen=pens
)
self.vi.viewer.getView().addItem(sp)
sp.hide()
self.roi_sps.append(sp)
for ix in range(num_components):
self.vi.viewer.status_bar_label.showMessage(
f"Please wait, adding component {ix} / {num_components} "
f"on zlevel {zcenter} / {self.num_zlevels - 1}"
)
curve_data = self.cnmC[-1][self.idx_components[-1][ix]]
contour = contours[ix]
if self.cnm_dfof[-1] is not None:
dfof_data = self.cnm_dfof[-1][ix]
else:
dfof_data = None
roi = VolMultiCNMFROI(
curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView(),
cnmf_idx=self.idx_components[-1][ix],
curve_data=curve_data,
contour=contour,
spike_data=self.cnmS[-1][ix],
dfof_data=dfof_data,
zcenter=zcenter,
zlevel=self.vi.viewer.current_zlevel,
roi_ix=ix,
scatter_plot=sp,
parent_manager=self,
)
self.roi_list.append(roi, add_to_list_widget=False)
self.roi_list.list_widget.addItems(
list(map(str, range(len(self.roi_list))))
)
self.vi.workEnv_changed("ROIs imported")
# self.roi_list.reindex_colormap(random_shuffle=True)
self.roi_sps[self.vi.viewer.current_zlevel].show()
self.vi.viewer.status_bar_label.showMessage('Finished adding all components!')
def set_zlevel(self, z: int):
"""Set the current z-level to be visible in the viewer"""
super(ManagerVolMultiCNMFROI, self).set_zlevel(z)
for i in range(len(self.roi_sps)):
if i == z:
self.roi_sps[i].show()
else:
self.roi_sps[i].hide()
def add_roi(self):
"""Not implemented, uses add_all_components to import all ROIs instead"""
raise NotImplementedError('Not implemented for CNMFE ROIs')
def restore_from_states(self, states: dict):
if 'metadata' in states.keys():
self.metadata = states['metadata']
if not hasattr(self, 'roi_list'):
self.create_roi_list()
self.cnmf_data_dicts = states['cnmf_data_dicts']
# for state in states['states']:
# roi = VolMultiCNMFROI.from_state(
# self.get_plot_item(),
# self.vi.viewer.getView(),
# state
# )
#
# self.roi_list.append(roi)
self.input_params_dict = states['input_params_cnmf']
self.num_zlevels = states['num_zlevels']
self.cnmA = states['cnmf_output']['cnmA']
self.cnmb = states['cnmf_output']['cnmb']
self.cnmC = states['cnmf_output']['cnmC']
self.cnm_f = states['cnmf_output']['cnm_f']
self.cnmYrA = states['cnmf_output']['cnmYrA']
self.idx_components = states['cnmf_output']['idx_components']
self.orig_idx_components = states['cnmf_output']['orig_idx_components']
self.dims = states['cnmf_output']['dims']
if not hasattr(self, 'roi_list'):
self.create_roi_list()
for zcenter in range(self.num_zlevels):
logger(f"Loading z-level {zcenter}")
contours = caiman_get_contours(
self.cnmA[zcenter][:, self.idx_components[zcenter]],
self.dims[zcenter],
# swap_dim=True
)
num_components = len(self.idx_components[zcenter])
self.ui.radioButton_curve_data.setChecked(True)
roi_ixs = []
roi_xy = []
for ix in range(len(contours)):
coors = contours[ix]['coordinates']
coors = coors[~np.isnan(coors).any(axis=1)]
roi_xy += [coors]
roi_ixs += [ix] * coors.shape[0]
roi_xy = np.vstack(roi_xy)
roi_ixs = np.vstack(roi_ixs)
self.roi_xys.append(roi_xy)
self.roi_ixs.append(roi_ixs)
cm = matplotlib_color_map.get_cmap('hsv')
cm._init()
lut = (cm._lut * 255).view(np.ndarray)
cm_ixs = np.linspace(0, 210, np.unique(roi_ixs).size + 1, dtype=int)
roi_crs = []
for roi_ix, cm_ix in zip(np.unique(roi_ixs), cm_ixs):
c = lut[cm_ix]
roi_crs.append(
np.array([c] * roi_ixs[roi_ixs == roi_ix].size) # color for each spot
)
roi_crs = np.vstack(roi_crs)
self.roi_crs.append(roi_crs)
xy_coors = self.roi_xys[-1]
brushes = list(map(pg.mkBrush, roi_crs))
pens = list(map(pg.mkPen, roi_crs))
sp = pg.ScatterPlotItem(
xy_coors[:, 0],
xy_coors[:, 1],
symbol='s',
size=1,
pxMode=True,
brush=brushes,
pen=pens
)
self.vi.viewer.getView().addItem(sp)
sp.hide()
self.roi_sps.append(sp)
for ix in tqdm(range(len(self.idx_components[zcenter]))):
self.vi.viewer.status_bar_label.showMessage(
f"Please wait, adding component {ix} / {num_components} "
f"on zlevel {zcenter} / {self.num_zlevels - 1}"
)
curve_data = self.cnmC[zcenter][self.idx_components[zcenter][ix]]
contour = contours[ix]
cnmf_idx = self.idx_components[zcenter][ix]
roi = VolMultiCNMFROI(
curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView(),
cnmf_idx=cnmf_idx,
curve_data=curve_data,
contour=contour,
zcenter=zcenter,
zlevel=self.vi.viewer.current_zlevel,
roi_ix=ix,
scatter_plot=sp,
parent_manager=self,
)
roi_state = list(
filter(
lambda r: r['cnmf_idx'] == cnmf_idx,
states['states'][zcenter]
)
)[0]
for k in roi_state['tags'].keys():
roi.set_tag(k, roi_state['tags'][k])
roi.dfof_data = roi_state['dfof_data']
roi.spike_data = roi_state['spike_data']
self.roi_list.append(roi)
self.roi_list.list_widget.addItems(
list(map(str, range(len(self.roi_list))))
)
self.vi.workEnv_changed("ROIs imported")
# self.roi_list.reindex_colormap(random_shuffle=True)
self.roi_sps[self.vi.viewer.current_zlevel].show()
self.vi.viewer.status_bar_label.showMessage('Finished adding all components!')
def get_all_states(self) -> dict:
roi_list_sorted = \
[
[
roi for roi in self.roi_list if roi.zcenter == zlevel
] for zlevel in range(self.num_zlevels)
]
states = \
{
'roi_type': self.roi_list.roi_types.__name__,
'states': [],
'metadata': self.metadata
}
roi_states = \
[
[
roi.to_state() for roi in rois_zlevel
] for rois_zlevel in roi_list_sorted
]
states['states'] = roi_states
# make a new idx_components list in case the user has manually deleted some ROIs
new_idx_components: List[np.ndarray] = []
for zlevel in range(len(roi_list_sorted)):
roi_cnmf_idxs = [roi.cnmf_idx for roi in roi_list_sorted[zlevel]]
roi_cnmf_idxs.sort()
new_idx_components.append(
np.array(roi_cnmf_idxs, dtype=np.uint64)
)
# store the cnmf attributes as well
input_dict = \
{
'input_params_cnmf': self.input_params_dict,
'cnmf_data_dicts': self.cnmf_data_dicts,
'num_zlevels': self.num_zlevels,
'cnmf_output':
{
'cnmA': self.cnmA,
'cnmb': self.cnmb,
'cnmC': self.cnmC,
'cnm_f': self.cnm_f,
'cnmYrA': self.cnmYrA,
'idx_components': new_idx_components,
'orig_idx_components': self.orig_idx_components,
'dims': self.dims
}
}
states.update(input_dict)
return states
[docs]class ManagerCNMFROI(AbstractBaseManager):
"""Manager for ROIs imported from CNMF or CNMFE outputs"""
[docs] def __init__(self, parent, ui, viewer_interface):
"""Instantiate necessary attributes"""
super(ManagerCNMFROI, self).__init__(parent, ui, viewer_interface)
self.create_roi_list()
self.list_widget = self.roi_list.list_widget
self.input_params_dict = None
self.idx_components = None # Keep track of components if the user manually want to remove some
self.orig_idx_components = None # List of components prior to any manual deletion by the user
# cnmf data dict directly from the hdf5 file
self.cnmf_data_dict = None
# These correspond to the caiman.source_extraction.cnmf attributes
self.cnmA = None
self.cnmb = None
self.cnmC = None
self.cnm_f = None
self.cnmYrA = None
self.raw_normalization_choices = ['top_5', 'top_10', 'top_5p', 'top_10p', 'top_25p']
[docs] def create_roi_list(self):
"""Create empty CNMFROI list"""
self.roi_list = ROIList(self.ui, CNMFROI, self.vi)
[docs] def add_all_components(self, cnmf_data_dict, input_params_dict, calc_raw_min_max=False):
"""
Add all components from a CNMF(E) output. Arguments correspond to CNMF(E) outputs
:param cnmf_data_dict: CNMF results data directly from the HDF5 file
:param input_params_dict: dict of input params, from the batch manager
:param calc_raw_min_max: Calculate raw min & max for each ROI
:return:
"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
self.cnmf_data_dict = cnmf_data_dict
self.cnmA = self.cnmf_data_dict['estimates']['A']
self.cnmb = self.cnmf_data_dict['estimates']['b']
self.cnm_f = self.cnmf_data_dict['estimates']['f']
self.cnmC = self.cnmf_data_dict['estimates']['C']
self.cnmS = self.cnmf_data_dict['estimates']['S']
# # h5py is doing a weird thing where ``None`` gets stored as a byte string
# if isinstance(self.cnmf_data_dict['estimates']['F_dff'][()], np.ndarray):
# self.cnm_dfof = self.cnmf_data_dict['estimates']['F_dff']
# else:
# self.cnm_dfof = None
self.cnm_dfof = self.cnmf_data_dict['estimates']['F_dff']
self.cnmYrA = self.cnmf_data_dict['estimates']['YrA']
self.dims = self.cnmf_data_dict['dims']
self.idx_components = cnmf_data_dict['estimates']['idx_components']
if self.idx_components is None:
self.idx_components = np.arange(self.cnmC.shape[0])
self.orig_idx_components = deepcopy(self.idx_components)
self.input_params_dict = input_params_dict
# spatial components
contours = caiman_get_contours(self.cnmA[:, self.idx_components], self.dims)
# if dfof:
# temporal_components = cnmC
# else:
# temporal_components = cnmC[idx_components]
self.input_params_dict = self.input_params_dict
num_components = len(self.idx_components)
if calc_raw_min_max:
img = self.vi.viewer.workEnv.imgdata.seq.T
self.ui.radioButton_curve_data.setChecked(True)
for ix in range(num_components):
self.vi.viewer.status_bar_label.showMessage('Please wait, adding component #: '
+ str(ix) + ' / ' + str(num_components))
curve_data = self.cnmC[self.idx_components[ix]]
contour = contours[ix]
if calc_raw_min_max:
# Get a binary mask
mask = self.cnmA[:, self.idx_components[ix]].toarray().reshape(self.dims, order='F') > 0
# mask3d = np.array((mask,) * curve_data.shape[0])
max_ix = curve_data.argmax()
min_ix = curve_data.argmin()
array_at_max = img[max_ix, :, :].copy()
array_at_max = array_at_max[mask]
array_at_min = img[min_ix, :, :].copy()
array_at_min = array_at_min[mask]
raw_min_max = self.get_raw_min_max(array_at_max=array_at_max,
array_at_min=array_at_min)
else:
raw_min_max = None
roi = CNMFROI(curve_plot_item=self.get_plot_item(),
view_box=self.vi.viewer.getView(),
cnmf_idx=self.idx_components[ix],
curve_data=curve_data,
contour=contour,
raw_min_max=raw_min_max,
dfof_data=self.cnm_dfof[ix] if (self.cnm_dfof is not None) else None,
spike_data=self.cnmS[ix])
self.roi_list.append(roi)
if calc_raw_min_max:
del img
self.roi_list.reindex_colormap()
self.vi.viewer.status_bar_label.showMessage('Finished adding all components!')
def get_raw_min_max(self, array_at_max, array_at_min):
a_size = array_at_max.size
p5 = int(a_size * 0.05)
p10 = p5 * 2
p25 = p5 * 5
out = {}
for a, r in zip((array_at_max, array_at_min), ('raw_max', 'raw_min')):
out[r] = {'top_5': self.get_raw_mean(a, min(5, a_size)),
'top_10': self.get_raw_mean(a, min(10, a_size)),
'top_5p': self.get_raw_mean(a, p5),
'top_10p': self.get_raw_mean(a, p10),
'top_25p': self.get_raw_mean(a, p25),
'full_mean': a.mean()
}
return out
def get_raw_mean(self, array, num_items):
return np.partition(array, -num_items)[-num_items:].mean()
[docs] def add_roi(self):
"""Not implemented, uses add_all_components to import all ROIs instead"""
raise NotImplementedError('Not implemented for CNMFE ROIs')
[docs] def restore_from_states(self, states: dict):
"""Restore from states, such as when these ROIs are saved with a Project Sample"""
super(ManagerCNMFROI, self).restore_from_states(states)
if not hasattr(self, 'roi_list'):
self.create_roi_list()
if 'cnmf_data_dict' in states.keys():
self.cnmf_data_dict = states['cnmf_data_dict']
else:
self.cnmf_data_dict = None
for state in states['states']:
roi = CNMFROI.from_state(self.get_plot_item(), self.vi.viewer.getView(), state)
self.roi_list.append(roi)
self.input_params_dict = states['input_params_cnmfe']
self.cnmA = states['cnmf_output']['cnmA']
self.cnmb = states['cnmf_output']['cnmb']
self.cnmC = states['cnmf_output']['cnmC']
self.cnm_f = states['cnmf_output']['cnm_f']
self.cnmYrA = states['cnmf_output']['cnmYrA']
self.idx_components = states['cnmf_output']['idx_components']
self.orig_idx_components = states['cnmf_output']['orig_idx_components']
self.roi_list.reindex_colormap()
[docs] def get_all_states(self) -> dict:
"""Get all states so that they can be restored"""
if not hasattr(self, 'roi_list'):
self.create_roi_list()
states = super(ManagerCNMFROI, self).get_all_states()
# If the user has manually deleted some ROIs
new_idx_components = np.array([roi.cnmf_idx for roi in self.roi_list], dtype=np.int64)
# Store the actual cnmf attributes as well.
input_dict = {'input_params_cnmfe': self.input_params_dict,
'cnmf_data_dict': self.cnmf_data_dict,
'cnmf_output':
{
'cnmA': self.cnmA,
'cnmb': self.cnmb,
'cnmC': self.cnmC,
'cnm_f': self.cnm_f,
'cnmYrA': self.cnmYrA,
'idx_components': new_idx_components,
'orig_idx_components': self.orig_idx_components
}
}
states.update(input_dict)
return states
[docs] def update_idx_components(self, ix: int):
"""Update idx_components if the user manually delete an ROI"""
roi = self.roi_list[self.roi_list.current_index]
self.idx_components = np.delete(self.idx_components, np.where(self.idx_components == roi.cnmf_idx)[0])
def set_spot_size(self, size: int):
for roi in self.roi_list:
roi.get_roi_graphics_object().setSize(size)
roi.spot_size = size