Vispro

Vispro is a command line tool for Spatial Transcriptomics image processing, including modules of fiducial marker detection, image restoration, tissue detection, and disconnected tissue segregation. It is dominantly develped and tested on high-resolution images, with resoluton of ~2000*2000. It can be applied to original microscopy image, with resolution of ~20,000 *~20,000 by dividing them into smaller patches and stitch the result togetehr. The project is built with PyTorch and Python and can run efficiently on a GPU.

Requirements

• Python 3.8 or later
• CUDA (for GPU support, optional but recommended)
• Python Libraries:
  • numpy>=1.18.0
  • torch>=1.7.0
  • scikit-image>=0.16.2
  • Pillow>=7.0.0
  • torchvision>=0.8.0
  • scipy>=1.4.1
  • matplotlib>=3.1.3
  • pymatting>=1.1.5
  • kornia>=0.5.0

Installation

Vispro is a pure-python based tool. We recommend use a miniconda environment for library management and running the code.

1. Clone the repository

git clone https://github.com/HuifangZJU/Vispro.git
cd Vispro

2. Set up environment

• 2.1 Set up a conda environment with all dependency libraries (recommended)

conda env create -f environment.yml
conda activate Vispro

For the installation of miniconda, please refer to the link https://docs.anaconda.com/miniconda/.

• 2.2 Install dependency directly with pip

pip install -r requirements.txt

Usage

The full list of pretrained models can be found at Direct Download. Please place the pretrained models in the folder Vispro/pretrained_models.

Process the high-resolution image only.

python process_high_res.py --image_path /test_data/tissue_hires_image.png

Process the original large image.

python process_large_image.py --high_res_image_path /path/to/your/tissue_hires_image.png --original_image_path /path/to/original/microscopy/image.tif

Licensed under MIT License