From Data to Animal Health: Building Benchmarks for AI-Driven Veterinary Innovation
Principal Investigator: Renata Ivanek
Co-PI: Casey Cazer; Parminder Basran; Jennifer Sun
DESCRIPTION (provided by applicant):
Benchmark datasets drive the advancement of artificial intelligence (AI), as demonstrated by their transformative impact in vision, language, science, and human healthcare. Yet, veterinary medicine lacks standardized, high-quality datasets, limiting the development and validation of AI tools tailored to veterinary needs. To fully realize AI’s potential in animal health, benchmark datasets are urgently needed across all veterinary domains—from companion animals to food-producing species. These resources will drive innovation that benefits animals, veterinary professionals, animal owners, and public health.
Creating datasets and benchmarks for AI in veterinary medicine presents unique challenges, including species diversity, lack of central regulations, varied data modalities, and ethical considerations of animals. While FAIR (Findable, Accessible, Interoperable, Reusable) data principles remain a hurdle in human healthcare, veterinary data faces dramatically different socio-economic, legislative, regulatory, and protection barriers. Tackling these challenges requires transdisciplinary collaboration across veterinary medicine, computer science, ethics, law, and related fields. Such partnerships are essential for building representative datasets, demonstrating benchmark utility and trust, and designing interfaces that make datasets and benchmarks updatable and accessible to both veterinary and AI communities.
To address this need, we propose a “Thought Summit” at Cornell University to catalyze the development of VETNET—an ecosystem of live benchmarks for AI-driven veterinary innovation guided by the principle: any animal, any disease, any task. The summit will convene experts to define key problems, build a stakeholder community, and develop and publish a whitepaper that will serve as the foundation for a large-scale, multi-institutional academia-industry partnership and grant proposal to develop public benchmark datasets across veterinary medicine. By merging computational innovations with applications in real-world veterinary medicine, this initiative will lay the groundwork for transformative progress in animal health and establish a foundation for securing large-scale, multi-institutional funding.
