Cellpose For Naparm

a little utility to generate cellpose rois for use with naparm

Usage

drag and drop images into the window
- multiple images corresponding to seperate z-plane can be dragged at once. multi-page tiffs can also be used.
adjust settings (optimise using the cellpose gui)
click run
load the generated '*_CELLPOSE.mat' file into naparm

Screenshots

The GUI:

The output MAT file:

Loaded into Naparm:

Output format

_CELLPOSE.m has the following fields:

'software' - flag to know genreated by cellpose
'img' - the images loaded into cellpose
'files' - the filepaths of the images
'model' cellpose parameter - the type of model
'cellsize' - cellpose paramter - roi size in pixels
'threshold' - cellpose paramter - threhsold
'masks' - cellpose output - array size of images with pixel coloured accoring to roi number

Make MATLAB ROIs

Load the *_CELLPOSE.m file into makeCellposeRois.m to generate ROI and halo masks to extract time-series data from movie (with extractTraces.m from 'simplePipeline')

filepath = 'C:\Users\User\Desktop\20200119_RN191_t-001_onlineREG_Plane1_CorrImg_CELLPOSE.mat';
watershed_width = 2;
halo_multiplier = 2;
show_plot = 1;
[roi, halo] = makeCellposeRois(filepath, watershed_width, halo_multiplier, show_plot);

Dependencies

cellpose
pyqt5
skimage

Improvements

figure out how to load cellpose.npy into matlab (readNPY.m does not work out the box)
modify cellpose to save MAT files, but also need to modify to allow multiple images at once
show output masks in this gui to allow for optimisation
add in image manipulation (blur, denoise, bg-removal etc) to improve results
or, run cellpose direct from matlab
- Python Error: MXNetError: Error in operator cpnet4_downsample0_resdown0_hybridsequential5_conv0_fwd: Shape inconsistent, Provided = [32,2,1,1], inferred shape=(32,3,1,1)