Identify adipocytes in bone marrow spatial omics data from segmentation masks.

Adipocytes are often lost in FFPE-processed bone marrow slides, leading to incomplete spatial proteomics or transcriptomics analysis. adipo_finder is a lightweight Python package designed to recover these regions by detecting adipocyte-shaped areas in segmentation masks, enabling more accurate tissue context reconstruction.

• 🔬 Detects adipocytes based on shape and absence of marker staining
• 🧠 Compatible with segmentation masks from spatial omics pipelines
• 🧩 Integrates with squidpy and AnnData objects
• ⚙️ Customizable parameters for blurring, object size, and morphology
• 🧪 Includes a toy dataset for demonstration and testing

From PyPI:

pip install adipo_finder

From GitHub:

pip install git+https://github.com/VergaJU/adipo_finder.git

from adipo_finder import utils as seg_utils
from adipo_finder import segmentation as seg
import matplotlib.pyplot as plt
import skimage.io as io


segmentation = io.imread("path_to_mask.tif")

old_seg,segmentation=seg_utils.Preprocessing.preprocess_image(image=segmentation)
new_img=seg.Segmentation.run_segmentation(image=img)
new_seg,df=seg_utils.Exporting.export_adipocytes(segmentation_image=old_seg,new_segmentation=new_img)
plt.imshow(new_seg)

from adipo_finder import utils as seg_utils
from adipo_finder import segmentation as seg
import matplotlib.pyplot as plt
import anndata as ad

adata = ad.read("path_to_anndata.h5ad")

old_seg,segmentation=seg_utils.Preprocessing.preprocess_image(adata=adata, library_id="library_identifier")
new_img=seg.Segmentation.run_segmentation(image=img)
new_seg,df=seg_utils.Exporting.export_adipocytes(segmentation_image=old_seg,new_segmentation=new_img)

df_exp= seg_utils.Exporting.expand_df(adata=adata,df=df, library_id="library_identifier")
adata_adipo = seg_utils.Exporting.create_adipo_adata(adata=adata,df=df_exp) 
adata_new=seg_utils.Exporting.merge_adatas(adata=adata,adata_tmp=adata_adipo,new_segmentation=new_seg,library_id="library_identifier")

A toy image is included in the repository to demonstrate how adipo_finder works. Use this for quick testing or to tune parameters before applying to real data.

Full documentation and examples are available in the GitHub Repository

If you use adipo_finder in your work, please cite the corresponding AppNote or preprint (coming soon).

This project is licensed under the MIT License.

✨ Contributions

Contributions and suggestions are welcome! Feel free to open issues or submit pull requests.