Related Experiment Video
Updated: Jul 8, 2026
05:47
Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
MultiGraph-vet: A multimodal knowledge-augmented decision-support framework for safe dairy cow disease assessment on
Cunjin Zhang1, Maoxu Wang1, Xu You1
1College of Computer and Artificial Intelligence, Northeast Forestry University, Harbin 150040,China.
Research in Veterinary Science
|July 6, 2026
Summary
MultiGraph-Vet enhances dairy cow disease assessment by integrating knowledge graphs and visual data, improving diagnostic accuracy and safety. This AI framework reduces errors in veterinary decision support for healthier herds.
Area of Science:
- Veterinary Medicine
- Artificial Intelligence
- Data Science
Background:
- Accurate dairy cow disease assessment is crucial for herd health management.
- Current clinical data is often ambiguous, leading to potential errors in AI decision-support systems.
- Large language models (LLMs) face challenges with factual accuracy and safety in veterinary contexts.
Purpose of the Study:
- To introduce MultiGraph-Vet, a multimodal framework for safer dairy cow disease assessment.
- To enhance the reliability and safety of AI-driven veterinary decision support.
- To address limitations of generic LLMs in processing complex clinical data.
Main Methods:
- Integration of a bovine veterinary knowledge graph, visual lesion parsing, and environmental context.
- Implementation of a multimodal safety gate for cross-modal consistency checking.
- Utilizing a dual-path knowledge graph strategy and rule-constrained reasoning for evidence-grounded diagnostics.
Main Results:
- Achieved 80.36% diagnostic accuracy and 87.27% candidate recall@K on a 338-case benchmark.
- Demonstrated a 96.83% safe interception rate and a 3.35% hallucination-related commitment rate.
- Outperformed generic multimodal LLM baselines in accuracy and safety.
Conclusions:
- Explicit veterinary knowledge grounding, visual standardization, and conflict checking significantly improve AI reliability and safety.
- MultiGraph-Vet offers a robust approach to AI-assisted dairy cow disease assessment.
- The framework supports preliminary assessment and risk-aware decision-making, not replacing veterinary diagnosis.