Skip to content

Welcome to segger

segger is a cutting-edge tool for cell segmentation in single-molecule spatial omics datasets. By leveraging graph neural networks (GNNs) and heterogeneous graphs, segger offers unmatched accuracy and scalability.

Quick Links

Why segger?

  • Highly parallelizable – Optimized for multi-GPU environments
  • Fast and efficient – Trains in a fraction of the time compared to alternatives
  • Transfer learning – Easily adaptable to new datasets and technologies

Challenges in Segmentation

Spatial omics segmentation faces issues like:

  • Over/Under-segmentation
  • Transcript contamination
  • Scalability limitations

segger tackles these with a graph-based approach, achieving superior segmentation accuracy.

How segger Works

Segger Model

Powered by

  • PyTorch Lightning & PyTorch Geometric: Enables fast, efficient graph neural network (GNN) implementation for heterogeneous graphs.
  • Dask: Scalable parallel processing and distributed task scheduling, ideal for handling large transcriptomic datasets.
  • Shapely & Geopandas: Utilized for spatial operations such as polygon creation, scaling, and spatial relationship computations.
  • RAPIDS: Provides GPU-accelerated computation for tasks like k-nearest neighbors (KNN) graph construction.
  • AnnData & Scanpy: Efficient processing for single-cell datasets.
  • SciPy: Facilitates spatial graph construction, including distance metrics and convex hull calculations for transcript clustering.

Contributions

segger is open-source and welcomes contributions. Join us in advancing spatial omics segmentation!