This is a pipeline for converting dual-channel TEMPO microscopy recordings to dF/F movies with physiological and recording artifacts removed via a convolution unmixing procedure.
This pipeline was tested in MATLAB 2019b, 2021b, and 2023a. For MATLAB versions prior to 2021b, use "local" parallel pool instead of a threads-based one (parpool("local")).
Add all folders (except pipelines) to MATLAB path. Add all dependencies to the MATLAB path.
dependencies_toolboxes.txt. Available from MathWorks
dependencies_external.txt. Available from other developers through mathworks.com/matlabcentral or github.com
dependencies_binaries.txt. For Allen map alignment, allenmap.mat file with Allen map countours is required, available upon request. For .dcimg to .h5 conversion, binary files dcimgmex.mexw64 / dcimgmatlab.mexw64, dct_readtimestamps.exe, and sdk drives from Hamamatsu are required. For fast filtering (td convolution), compiled hdf5_movie_convolution.exe and open source linear algebra and fft libraries are required. Available upon request.
Browse the example pipelines in the pipelines folder.
pipelines\pipeline_preprocessing_2xmoco.m
Data preprocessing includes independent motion correction of both channels and registration of the reference channel to the signal channel.
pipelines\pipeline_unmixing.m
Unmixing of physiological and recording artifacts. Decrosstalking, high-pass filtering, convolutional unmixing, and F0 normalization.
This processing pipeline is described in Haziza et al., 2025. The convolutional unmixing procedure was first introduced in a talk Kruzhilin et al., 2023. Please cite us if you use this pipeline in your own work.