2P conversion notes

[weiglszonja]

This is an ongoing data collection, based on the examples that we received from Shiva we have the following source data:

Source data

• Raw imaging acquired with ScanImage
  • single plane, single channel
  • single plane, dual channel
• Behavior video recordings in multiple TIF files
• DLC output (not sure what version)
• Motion corrected imaging data in TIF format and the xy shifts in .mat file
• Experiment logs in DAT and MAT format: velocity, break_history, day_time etc.

ScanImage data

The imaging session 2586269R_230817 consists of 8 TIF files in ScanImage format. In each TIF the video shape is (4000, 512, 512), based on the ScanImage metadata it is single channel and single plane data.

Example snippet:

I had some issues with ScanImageTiffSinglePlaneImagingExtractor when trying to load the video from the TIF files (issue opened here). The _frames_per_slice attribute is determined from the ScanImage metadata which reflects the total number of frames over the 8 TIF files, and this results in an IndexingError when trying to load the video of a single TIF which only contains 4000 frames.

This is fixed by https://github.com/catalystneuro/roiextractors/pull/294.

Dual-channel ScanImage data

The imaging session 2620749R2_231211 consists of multiple TIF files as well. In each TIF the video shape is (8000, 512, 512), based on ScanImage metadata it is dual channel and single plane data.

Example frame from "Channel 1" and "Channel 2":

Schematic representation of NWB mapping:

TODO:

• What is the version of DLC that they are using? They are not compatible with the interface in Neuroconv
• Need an example with lick signal

[weiglszonja]

When running full conversion make sure to check this note from @alessandratrapani:

check the deconvolved traces once you do the whole conversion because for the stub it's only zeros.