Toxicology vision models improve pathology across animals and organs
Scaling Vision Foundation Models for Toxicologic Pathology
Computer Vision and Pattern Recognition
Summary
Toxicologic pathology involves studying tissue samples from animals to detect harmful effects of chemicals or drugs. The authors developed TOX-FM, a large family of computer models trained using over a million images from diverse animal tissues collected over decades. These models can perform many different pathology tasks better than existing clinical models, including ones they were not specifically trained for. The authors also studied how training on specialized non-clinical data affects model performance on clinical tasks and vice versa.
What this means in practice
- •For veterinary pathologists: Use TOX-FM to improve accuracy and speed when analyzing animal tissue slides from toxicology studies.
- •For biopharmaceutical safety teams: Deploy TOX-FM models to assist in preclinical toxicology assessments across multiple species and organs.
Authors
Jessica Loo, Theodoros Pissas, Byunghak Kang, Vincent A. Ella, Xiaoyan Zhang, Kevin Sweeney, Walter Georgescu, Prateek Katiyar, Julie Boisclair, Holger Hoefling
Abstract
Here, we present TOX-FM, a family of non-clinical PFMs trained on large-scale diverse histopathology datasets of over one million whole slide images consisting of over 60 organs across six animal species from over 14,000 toxicology studies across six decades. We introduce a comprehensive benchmark of 18 non-clinical downstream tasks, incorporate an established benchmark of 9 clinical downstream tasks, and evaluate TOX-FM against state-of-the-art clinical PFMs across both non-clinical and clinical domains. Notably, we explore open questions surrounding domain-specific training and cross-domain generalization, offering insights into the opportunities and challenges of applying PFMs in the non-clinical domain.