sphero-vem

sphero-vem is a Python library for end-to-end quantitative analysis of volume electron microscopy (vEM) data, with a focus on 3D cell and nucleus morphology and nanoparticle (NP) localization and quantification in tumor spheroid models. It covers the full analysis pipeline from raw image ingestion to per-label quantitative property extraction, with GPU acceleration throughout: automatic via CuPy and CuCIM for array operations, and PyTorch-native for registration and segmentation. Cell and nucleus segmentation builds on Cellpose-SAM, self-supervised denoising on CAREamics, and image registration on PyTorch.

Originally developed for SBF-SEM imaging of gold nanoparticle (AuNP) distribution in FaDu head-and-neck tumor spheroids, the pipeline is designed to generalize to other vEM modalities and biological specimens.

sphero-vem accompanies the paper:

Bottone et al., 3D Reconstruction of Nanoparticle Distribution in Tumor Spheroids with Volume Electron Microscopy, 2026. https://doi.org/10.64898/2026.04.17.719153

The annotated dataset is publicly available at BioImage Archive (S-BIAD3263), and finetuned model weights for cell and nucleus segmentation are available on Zenodo.

Contents

• Installation
  • Requirements
  • Development install
  • GPU acceleration
  • Dataset and model weights
• Analysis Pipeline
  • Stage 1: Ingest
  • Stage 2: Denoising
  • Stage 3: Registration
  • Stage 4: Cell and nucleus segmentation
  • Stage 5: Nanoparticle segmentation
  • Stage 6: Shape analysis
  • Stage 7: Spatial analysis
  • Stage 8: Output
  • GPU dispatch
• API Reference
  • IO (sphero_vem.io)
  • Preprocessing (sphero_vem.preprocessing)
  • Denoising (sphero_vem.denoising)
  • Metrics (sphero_vem.metrics)
  • Registration (sphero_vem.registration)
  • Segmentation (sphero_vem.segmentation)
  • Measure (sphero_vem.measure)
  • Utils (sphero_vem.utils)