You can deploy this model and start earning money today!

cbeeche/
percival

Name: percival
Rating: 2.5 (8 reviews)

three-dimensional vision-language model

apache-2.0

Views

No views yet

Percival: A Generalizable Vision-Language Model for Computed Tomography

This repository provides the pretrained weights for Percival, a vision-language foundation model trained on over 400,000 CT volumes and paired radiology reports from more than 50,000 participants in the Penn Medicine BioBank (PMBB). Percival leverages multi-view, multi-contrast data to learn generalizable 3D representations for clinical and biological tasks.

For details on model architecture, training data, and evaluation, see the preprint and the GitHub repository.

Citation

If you use Percival in your work, please cite:

@article {Beeche2025.07.03.25330654,
	author = {Beeche, Cameron A and Kim, Joonghyun and Tavolinejad, Hamed and Zhao, Bingxin and Sharma, Rakesh and Duda, Jeffrey and Gee, James and Dako, Farouk and Verma, Anurag and Morse, Colleen and Hou, Bojian and Shen, Li and Sagreiya, Hersh and Davatzikos, Christos and Damrauer, Scott M. and Shad, Rohan and Ritchie, Marylyn D and Rader, Daniel J and Long, Qi and Eaton, Eric and Chen, Tianlong and Kahn, Charles E and Chirinos, Julio A and Witschey, Walter R and Penn Medicine Biobank},
	title = {Generalizable CT Vision-Language Modeling for Population Health and Disease Risk},
	elocation-id = {2025.07.03.25330654},
	year = {2026},
	doi = {10.1101/2025.07.03.25330654},
	publisher = {Cold Spring Harbor Laboratory Press},
	URL = {https://www.medrxiv.org/content/early/2026/01/26/2025.07.03.25330654},
	eprint = {https://www.medrxiv.org/content/early/2026/01/26/2025.07.03.25330654.full.pdf},
	journal = {medRxiv}
}

cbeeche/percival

Percival: A Generalizable Vision-Language Model for Computed Tomography

Citation

cbeeche/
percival