Abstract
BACKGROUND: The application of artificial intelligence (AI) in veterinary oncology is rapidly expanding, mirroring its advancements in human medicine. This field is uniquely positioned to offer bi-directional insights due to the spontaneous development of cancers in companion animals that are similar to those in humans. However, a comprehensive understanding of the current research landscape is lacking. This scoping review was conducted to systematically map the literature on AI in veterinary oncology, identifying the clinical applications, techniques, and data sources being utilized, as well as the major challenges hindering clinical translation. RESULTS: The review included 69 studies, revealing a field with a strong focus on diagnostic applications in canine patients, particularly for common tumor types such as lymphomas, (sub-)cutaneous and mammary tumors. The most mature applications involve image-based diagnostics, including digital pathology and radiomics, where deep learning models have demonstrated high performance in tasks like tumor grading and non-invasive characterization. While emerging applications in treatment planning and multimodal data fusion show great promise, the overall field is limited by a pervasive reliance on small, single-source datasets and a lack of external and prospective validation. CONCLUSIONS: The application of AI in veterinary oncology has produced powerful proof-of-concept models, particularly in diagnostics, with a clear potential to augment clinical practice. However, the path from research to clinical implementation is hindered by fundamental challenges, including the data bottleneck and validation gap. To fulfill its transformative potential, the field must prioritize a shift from isolated studies to collaborative, large-scale research efforts that generate standardized, public datasets and emphasize rigorous external validation. By doing so, the community can ensure the development of generalizable AI models that will truly improve cancer care for veterinary patients.