Extracting cellular data via tiling PCA

This procedure describes the few computational steps required to extract cellular data from large scale one photon movies.

Steps in order :

1. Load the raw data.

2. Use the Preprocessing_1P_MotCorr_Subsample app. This app allows use to motion correct the movie, bin in time and crop black borders Use the motion correction parameters so as to create a black and white image that focus on blood vessels higher contrast. You can subsample in time so as to create about 7Hz final frame rate (typically 3 times binning). Keep the raw full resolution in space. Crop the image 10 pixels down to remove black borders.

3. Use the Spatial High Pass App to remove large scale fluctuation. On full resolution movie, a value of 40 for sigma works well.

4. Apply DF/F using (F-F0)/F0 and calculating F0 on the entire time.

5. Tile the movie using "tile movie". Each tile should be 500 by 500 pixels with 10 pixels overlap

6. For each tile, Mask part of the movie that represent noisy backgrounds (like inside huge blood vessels or outside of the cannula). Use the Mask_Movie app for that. Don't replace with imfilter, just let the software use background value. This step is important to remove high variance backgrounds to avoid a bias with PCA_ICA

7. Run PCA/ICA on each tile. Select the number of output PCs to be 2 times the number of ICs. The number of ICs should grossly match the number of visible neurons plus some extras to account for blood vessels ICs. Typical number range from 200 to 500.

8. Sort each tile ICs using "Sort ICs" App.