GEM-pRF

Welcome to GEM-pRF, a standalone, GPU-accelerated tool for population receptive field (pRF) mapping, built for large-scale fMRI analysis.

For theory and full method details, see our paper:

👉 Mittal et al. (2025): GEM-pRF: GPU-Empowered Mapping of Population Receptive Fields for Large-Scale fMRI Analysis https://doi.org/10.1016/j.media.2025.103891

Documentation

• Documentation and examples are available at: https://gemprf.github.io/

• For a deeper look into the mathematical and computational foundations, the paper above is the best reference.

Installation

GEM-pRF relies on an NVIDIA GPU and CUDA. Make sure your system has:

• A compatible NVIDIA GPU
• A matching CUDA toolkit
• A matching NVCC compiler

1. Install GEM-pRF

pip install gemprf

Latest versions: https://pypi.org/project/gemprf/

2. Install CuPy (required)

GEM-pRF depends on CuPy, but CuPy must match your CUDA version — so it is not installed automatically.

Install the correct CuPy wheel for your system:

• https://docs.cupy.dev/en/stable/install.html#installing-cupy
• or via pip, for example:

pip install cupy-cuda12x

Caution

Install the CuPy variant that matches your CUDA version.

You must install CuPy before running GEM-pRF.

Running GEM-pRF

After installing gemprf and a compatible CuPy build, you can run GEM-pRF directly from Python.

Example

import gemprf as gp

gp.run("path/to/your_config.xml")

Configuration files

GEM-pRF uses XML configuration files to define analysis settings. See a sample config here:

https://github.com/siddmittal/GEMpRF_Demo/blob/main/sample_configs/sample_config.xml

---

Quick workflow

1. Install GEM-pRF → pip install gemprf

2. Install the correct CuPy for your CUDA environment

3. Prepare your XML config file

4. Run:

   import gemprf as gp
   gp.run("config.xml")