🧬 MorphoFM

MorphoFM is a unified framework for extracting multi-level morphological features (including single-cell level) from spatial single-cell imaging data (e.g., H&E images).

It provides three complementary feature modules:

• Morpho1 – classical single-cell morphology (shape, size, intensity, texture)
• Morpho2 – deep visual embeddings from foundation models (e.g. UNI2-h, Virchow2)
• Morpho3 – spatial microenvironment & niche features (density, neighborhood geometry, contextual morphology)

More details are available in the demo notebook.

---

Installation

git clone https://github.com/Teichlab/MorphoFM.git
cd MorphoFM

conda env create -f morpho_env.yml
conda activate morpho

pip install -e .

Quick Start

morphofm \
  --image your_path/he.tiff \
  --labels your_path/he.npz \
  --feattype all \
  --sample_id Image1 \
  --mpp 0.25 \
  --crop_size_um 32 \
  --geometry circle \
  --radius_um 32 \
  --cuda_visible_devices 1 \
  --hf_home your_path_hf_cache