New reporting guidelines improve transparency in veterinary pathology AI research
Artificial intelligence (AI) is rapidly transforming veterinary pathology, offering powerful tools for automated image analysis (AIA). These advancements promise greater reproducibility, richer spatial data, and efficiency in interpreting whole-slide images. However, this progress comes with a challenge: ensuring transparency and reproducibility in research that uses these complex, often “black box” models.
The article emphasizes a critical issue: while AI-based AIA is gaining traction, many studies lack sufficient detail for reviewers to assess quality or for other scientists to replicate findings. To address this, Veterinary Pathology introduces a 9-point reporting checklist for authors submitting AI-related manuscripts. Developed by an interdisciplinary team, the checklist focuses on detailed documentation of study design, dataset creation, model training, performance evaluation, and interaction with AI models.
This structured approach serves three key purposes:
Enable assessment and identification of bias by reviewers and readers.
Facilitate reproducibility so other researchers can validate findings.
Support practical implementation of AI tools in clinical and diagnostic workflows.
Importantly, the authors stress that reproducibility depends not just on clear reporting but also on access to supporting data, code, and trained models—elements often missing from published studies. Without these, even the most transparent methods can fall short when models behave inconsistently across labs.
By adopting these guidelines, researchers can streamline the review process, reduce back-and-forth revisions, and strengthen trust in AI applications within veterinary pathology. This initiative marks a significant step toward responsible integration of AI, ensuring that innovation translates into meaningful, reliable diagnostic tools.
Bottom line: AI in veterinary pathology is exciting—but without rigorous reporting and openness, its potential could be compromised. This new checklist sets the stage for transparent, reproducible, and clinically relevant research in the field.
