jf-lab/chendoscope-minipipe

The JF-Lab CHEndoscope pipeline for preprocessing and analysis.

JF Lab Pipeline for CHEndoscope Analysis

The JF-Lab pipeline for preprocessing and analysis of CHEndoscope calcium imaging data.

1. Get your .mkv files to be merged, downsampled and/or motion corrected.
2. Run the following on the .mkv files:
   python minipipe.py file1.mkv file2.mkv -c 2000 --motion_corr --cores 8
   Flags:
   -d/--downsample: temporal downsample factor, defaut=4
   -c/--chunk_size: chunk_size, default=2000
   --motion_corr: if you want to correct motion, default=False
   -t/--threshold: if you want to indicate threshold for motion correction, default=1.0
   -target_frame: if you want to indicate frame of reference for motion correction, default=0
   --cores: number of threads to run in parallel, default=4
   --bigtiff: If .mkv(s) amount to > 12Gb, must use this mode or memory error will occur
   --merge: merge all the files instead of individually processing them
   -o/--output: If --merge, then the name for the merged .tiff file
   -f/--format: output format as tiff, avi or hdf5, default is tiffs
3. Run CNMF-E on the .tiff files, output is a .mat file.
4. Use review_traces.py to manually inspect the neurons to keep or exclude from analysis:
   python review_traces.py traces.mat

• Press 'k' to keep, 'j' to exclude, or the 'keep'/'exclude' buttons.

chendoscope-minipipe's Issues

Review Traces [Feature Request]

• zoom into traces
• which cell are you on,and total number of cells
• separate the noise from denoised more clearly
• keyboard shortcuts
• Acknowledgment at the end. TY JFLAB

Preprocessing all in mkv

PIMS works best with mkv, so if we do all preprocessing first, then before cnmf-e, we convert to TIFF, would work best.

Detect Cells

• - flatten frames taken from red channel to get working image
• - get ROIs via watershed or gaussian filter or other
• - plot/visualize ROIs
• - register to GCaMP video

mkv merge add pre-pre-cnmfe step

read and write motion correction/downsampling of large files in chunks

PIMS doesn't have ability to write to random access of video files.

pre_cnmfe.py

load target_frame(s) with PIMs
downsample target_frame
for chunk in chunks:
    load chunk with PIMS
    downsample
    motion correct to target_frame
    save as tiff chunk

Then use bash:
cat file1 file2 file3 > bigfile

make a cropping function

video registration pre-cnmfe, and neuronal registration post-cnmfe

pre to improve the alignment of cells post

bash pipeline script

$ bash minipipe.sh *.mkv -d 4

• take vids to process
• motion correct each
• convert to tiff
• concatenate tiffs if split
• delete temp tiffs