Piscis

piscis is an automatic deep learning algorithm for spot detection, written in Python using the PyTorch framework. It is named after the Latin word for fish, as it was designed specifically for microscopy images generated by RNA fluorescence in situ hybridization (FISH). However, we have found it to be useful for other imaging methods, such as immunofluorescence (IF) and FISH-based spatial transcriptomics. To learn more about piscis, please read our Cell Systems paper or bioRxiv preprint.

This Python package allows users to apply pre-trained models from Hugging Face to both single plane images and z-stacks or to train new models using custom datasets. It provides a simple API for both training and inference that can be used in traditional Python scripts or Jupyter notebook environments such as on Google Colab. It also provides a command line interface for those who prefer the terminal. For a user-friendly graphical user interface, we have implemented piscis as a Docker image for NimbusImage, a cloud platform for biological image analysis enabling researchers to interactively visualize their data while leveraging state-of-the-art machine learning algorithms.

For more information, please refer to the comprehensive documentation available at https://piscis.netlify.app.

Examples

Installation

Install piscis from PyPI with pip.

pip install piscis

Usage

If you want to use piscis with its Python API, check out the inference example and training example notebooks.

If you want to use piscis with its command line interface, run the following commands.

# Run Piscis.
piscis predict INPUT_PATH OUTPUT_PATH [OPTIONS]

# Train Piscis.
piscis train MODEL_NAME DATASET_PATH [OPTIONS]

To see the full list of options, run piscis predict --help or piscis train --help.

Citation

If you use piscis in your research, please cite our paper.

Niu, Z., O’Farrell, A., Li, J., Reffsin, S., Jain, N., Dardani, I., Goyal, Y., & Raj, A. (2025). Piscis: A loss estimator of the F1 score enables accurate spot detection in fluorescence microscopy images via deep learning. Cell Systems. https://doi.org/10.1016/j.cels.2025.101448

@article{Niu2025-Piscis,
    title={Piscis: A loss estimator of the F1 score enables accurate spot detection in fluorescence microscopy images via deep learning},
    author={Niu, Zijian and O’Farrell, Aoife and Li, Jingxin and Reffsin, Sam and Jain, Naveen and Dardani, Ian and Goyal, Yogesh and Raj, Arjun},
    year=2025,
    journal="Cell Systems",
    DOI={10.1016/j.cels.2025.101448}
}

License

piscis is licensed under the MIT License. The copyright and permission notices found in the LICENSE file shall be included in all copies or substantial portions of the Software.