4. Suite2p Grid Search

Demonstrates parameter grid-search on threshold_scaling and max_overlap in suite2p.

These are roi detection parameters.

• threshold_scaling (default: 1.0): controls how much you want your neurons to “stand out” from the noise to count as a cell, lower will detect more cells.

• max_overlap (default: 0.75): set to 1 will allow all overlapping ROI’s, 0.5 means if half of the cell is overlapping the roi will be discarded.

from pathlib import Path
import os
import matplotlib.pyplot as plt
import numpy as np
import suite2p
import mbo_utilities as mbo
import lbm_suite2p_python as lsp

4.1. Data path setup

data_dir = Path(r"D:\W2_DATA\kbarber\2025_03_01\mk301")
assembled_path = data_dir.joinpath("assembled")  # where the assembled tiffs are saved
save_path = data_dir.joinpath("results_gridsearch")
print(f"Saving raw tiffs to:        {assembled_path}")
print(f"Saving suite2p results to:  {save_path}")

Saving raw tiffs to:        D:\W2_DATA\kbarber\2025_03_01\mk301\assembled
Saving suite2p results to:  D:\W2_DATA\kbarber\2025_03_01\mk301\results_gridsearch

input_files = mbo.get_files(assembled_path, str_contains='tif', max_depth=3)
input_files = [Path(x) for x in input_files]
input_files[:3]

[WindowsPath('D:/W2_DATA/kbarber/2025_03_01/mk301/assembled/plane_01_mk301.tiff'),
 WindowsPath('D:/W2_DATA/kbarber/2025_03_01/mk301/assembled/plane_02_mk301.tiff'),
 WindowsPath('D:/W2_DATA/kbarber/2025_03_01/mk301/assembled/plane_03_mk301.tiff')]

4.2. Define parameter values to change

from itertools import product

scaling_variants = [1.2, 1, .8]
overlap_variants = [.9, .75, .6]
parameter_grid = list(product(scaling_variants, overlap_variants))

base_dir = save_path.parent / 'grid_search'
base_dir.mkdir(exist_ok=True)

mdata = mbo.get_metadata(input_files[9])
ops = mbo.params_from_metadata(mdata, suite2p.default_ops())

Ops provided. Setting pipeline to suite2p

from copy import deepcopy

for scaling, overlap in parameter_grid:
    dir_name = f"thrscale_{scaling}_overlap_{overlap}".replace(".", "_")
    path = base_dir.joinpath(dir_name)
    path.mkdir(exist_ok=True)
    new_ops = deepcopy(ops)
    new_ops['threshold_scaling'] = scaling
    new_ops['max_overlap'] = overlap
    ops['save_path0'] = str(path)
    opsEnd=suite2p.run_s2p(ops=new_ops,db={'data_path': [str(input_files[9].parent)], 'save_path': [path]})

ops_files = mbo.get_files(base_dir, 'ops', 4)
ops_files

['D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_0_8_overlap_0_75\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_0_8_overlap_0_9\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_2_overlap_0_6\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_2_overlap_0_75\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_2_overlap_0_9\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_overlap_0_6\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_overlap_0_75\\suite2p\\plane0\\ops.npy',
 'D:\\W2_DATA\\kbarber\\2025-02-27\\mk301\\grid_search\\thrscale_1_overlap_0_9\\suite2p\\plane0\\ops.npy']

for op in ops_files:
    ops_dict = lsp.load_ops(op)
    root_dir = Path(ops_dict['save_path'])
    plane_name = Path(ops_dict["tiff_list"][0]).stem
    prepend_str = root_dir.parent.parent.name
    save_name = base_dir / 'grid_figs'
    save_name.mkdir(exist_ok=True)
    full_savename = save_name / f"{plane_name}_{prepend_str}"
    # lsp.plot_segmentation(ops_dict, f"{full_savename}_segmentation.png", overlay=True, fig_label = prepend_str)
    lsp.plot_segmentation(ops_dict, f"{full_savename}_segmentation_overlay.png", overlay=False, fig_label=prepend_str)