Skip to content

tools._segmentation

def predict_cell_num(st_adata_path,
                     diameter,
                     save_png_result=False,
                     model_type='cyto3',
                     out_path='./',
                     cellprob_threshold=1,
                     save=True)

Predicts cell numbers from spatial transcriptomics data using Cellpose.

Parameters:

Name Type Description Default
st_adata AnnData

Spatial transcriptomics AnnData object.

 required
library_ids string

Identifier for the library in the AnnData object.

 'library_ids'
diameter int

Estimated diameter of the cells to be detected.

 None
save_png_result bool

If True, saves the segmentation results as PNG files.

 False
model_type string

Type of Cellpose model to use. Default is 'cyto3'.

 'cyto3'
out_path string

Directory to save the output files. Default is './'.

 required
cellprob_threshold float

Cell probability threshold for Cellpose. Default is 1.

 0.4
save bool

If True, saves the results to an h5ad file. Default is True.

 True

Returns:

Name Type Description
st_adata AnnData

Updated AnnData object with predicted cell numbers.
cell_pos DataFrame

DataFrame containing cell positions.
Source code in cytobulk\tools\_segmentation.py 
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
def predict_cell_num(st_adata,
                     library_ids="library_ids",
                     project="test",
                     diameter=None,
                     save_png_result=False,
                     model_type='cyto3',
                     out_dir='.',
                     cellprob_threshold=0.4,
                     save=True):

    '''
    Predicts cell numbers from spatial transcriptomics data using Cellpose.

    Parameters
    ----------
    st_adata : AnnData
        Spatial transcriptomics AnnData object.

    library_ids : string
        Identifier for the library in the AnnData object.

    diameter : int, optional
        Estimated diameter of the cells to be detected.

    save_png_result : bool, optional
        If True, saves the segmentation results as PNG files.

    model_type : string, optional
        Type of Cellpose model to use. Default is 'cyto3'.

    out_path : string, optional
        Directory to save the output files. Default is './'.

    cellprob_threshold : float, optional
        Cell probability threshold for Cellpose. Default is 1.

    save : bool, optional
        If True, saves the results to an h5ad file. Default is True.

    Returns
    -------
    st_adata : AnnData
        Updated AnnData object with predicted cell numbers.

    cell_pos : pandas.DataFrame
        DataFrame containing cell positions.
    '''
    from cellpose import models, io, plot
    from tqdm import tqdm

    print('-----Initializing model...')
    model = models.Cellpose(model_type=model_type)
    ch = [0, 0] # NOTE: here we set all images to greyscale

    print('-----Reading files...')
    img = rgb2grey(st_adata.uns['spatial'][library_ids]['images']['hires'])


    coord = st_adata.obsm['spatial']*st_adata.uns['spatial'][library_ids]['scalefactors']['tissue_hires_scalef']
    spots = pd.DataFrame(coord, columns=["X", "Y"])
    crop_r = int(st_adata.uns['spatial'][library_ids]['scalefactors']['spot_diameter_fullres']*st_adata.uns['spatial'][library_ids]['scalefactors']['tissue_hires_scalef'])
    #crop_r = int(st_adata.uns['spatial'][library_ids]['scalefactors']['spot_diameter_fullres'])
    half_r = crop_r // 2 + 5
    print(half_r)
    print('-----Predicting cell number...')
    ret = pd.DataFrame(data={'X':[], 'Y':[], 'cell_num':[]})
    cell_pos = pd.DataFrame(data={'id':[], 'X':[], 'Y':[]})
    for _, row in tqdm(spots.iterrows()):
        x = int(row[Const.PIX_X]); y = int(row[Const.PIX_Y])
        x_max = min(x+half_r, img.shape[0]-1)
        x_min = max(x-half_r, 0)
        y_max = min(y+half_r, img.shape[1]-1)
        y_min = max(y-half_r, 0)

        tile = img[x_min:x_max, y_min:y_max]
        masks, flows, styles, diams = model.eval(tile, diameter=diameter, channels=ch,  cellprob_threshold=cellprob_threshold)
        cell_num = len(np.unique(masks))
        ret.loc[len(ret.index)] = [x, y, cell_num]
        for i in range(cell_num):
            xi = np.where(masks == i)[0].mean()
            yi = np.where(masks == i)[1].mean()
            cell_pos.loc[len(cell_pos.index)] = [f"spot{_}_cell{i}", xi, yi]

        if save_png_result:
            fig = plt.figure()
            plot.show_segmentation(fig, tile, masks, flows[0], channels=ch)
            plt.tight_layout()
            #plt.savefig(save_png_result.replace('.', f'_{x}x{y}.'))
            plt.savefig(f"{out_dir}/figures/{_}_segmentation_result.png")

    st_adata.obsm["cell_num"] = (ret["cell_num"]).to_numpy()
    st_adata.uns["seg_cell_pos"] = cell_pos
    ret["cell_num"].to_csv(f"{out_dir}/output/{project}_cell_num.csv")
    cell_pos.to_csv(f"{out_dir}/output/{project}_cell_pos.csv")
    st_adata.write_h5ad(f"{out_dir}/output/segmentation_adata.h5ad")
    return st_adata, cell_pos