BUG: neuropil masks missing for manually added ROIs

[A-Courtney]

Describe the issue:

Hello!

First of all, thanks for creating a great tool! I have been using suite2p to analyse my 1P data. I use cellpose to find cells and then I manually add any cells that cellpose missed. I then wanted to look at the neuropil masks for all my ROIs but all the manually added ROIs are 'nan' while the cell pose ROIs have the mask coordinates. This seems strange as I can see neuropil fluorescence traces for all ROIs. Have I done something wrong or is this a bug?

Many thanks in advance for your help!

Amy

Reproduce the code example:

#this is the code I use to visualise the neuropil masks in jupyter lab. It works fine with ROIs that #were generated using the suite2p gui/cell pose ROI detection but I get an error when I try to run data #that was generated with manual ROIs, which is due to the nan values.


neumask = np.zeros((ops['Ly'], ops['Lx']))
for c in range(len(iscell)):
    if iscell[c,0] == 0:
         continue

    neuy, neux = np.unravel_index(stats[c]['neuropil_mask'], (ops['Ly'], ops['Lx']))
    
    neumask[neuy, neux] = 1
    neumask[stats[c]['ypix'],stats[c]['xpix']] = 2
plt.imshow(neumask)

sns.despine(left=True, bottom=True)

Error message:

No response

Version information:

suite2p v0.14.4

Context for the issue:

No response

[carsen-stringer]

thanks we'll look into this

[chriski777]

Hi @A-Courtney, thanks for letting us know about this. There was an issue with manual labeling in the current version of the code. Could you try out the version of the code in #1219 and let us know if the issue persists?

[chriski777]

HI @A-Courtney, just following up on this.

The updated outputs of suite2p don't contain the neuropil mask key anymore. Neuropil masks can be determined by using the function create_neuropil_masks. I've tested this, and this works for cells added manually via the GUI (new cells will be appended to the front of the stat.npy. I'll close this issue for now but feel free to reopen/comment if you have any other questions.

Here is a script that you can use to look at the neuropil mask for your cells:

import numpy as np
import matplotlib.pyplot as plt
from suite2p.extraction.masks import create_neuropil_masks, create_cell_pix

plane_path = "<INSERT_PATH_TO_PLANE_DIRECTORY>"

stat   = np.load(f"{plane_path}/stat.npy",   allow_pickle=True)
ops    = np.load(f"{plane_path}/ops.npy",    allow_pickle=True).item()
iscell = np.load(f"{plane_path}/iscell.npy", allow_pickle=True)

Ly, Lx = ops["Ly"], ops["Lx"]

cell_indices = np.where(iscell[:, 0].astype(bool))[0]
cell_stat = stat[cell_indices]

print(f'Number of cells: {cell_stat.shape[0]}.')
cell_idx = 0 # index into cells only, not all ROIs

# Build cell_pix map from all ROIs
cell_pix = create_cell_pix(stat, Ly=Ly, Lx=Lx)

# Compute neuropil mask for the chosen cell
neuropil_masks = create_neuropil_masks(
    ypixs=[cell_stat[cell_idx]["ypix"]],
    xpixs=[cell_stat[cell_idx]["xpix"]],
    cell_pix=cell_pix,
)
nm = neuropil_masks[0]
ny, nx = np.unravel_index(nm, (Ly, Lx))

# Plot
fig, ax = plt.subplots(figsize=(6, 6))
ax.imshow(ops["meanImg"], cmap="gray")
ax.scatter(nx, ny, s=1, c="red", alpha=0.3, label="neuropil")
ax.scatter(cell_stat[cell_idx]["xpix"], cell_stat[cell_idx]["ypix"], s=2, c="cyan", label="cell ROI")
ax.legend()
ax.set_title(f"Cell {cell_indices[cell_idx]} (cell index {cell_idx})")
ax.axis("off")
plt.show()