DentAgent Improves Multimodal Dental Diagnosis with Evidence-Centric AI

Oral diseases represent a global health challenge, necessitating accurate dental assessments that can synthesize information from various sources, including domain knowledge, X-rays, intraoral photos, and 3D dental data. Current AI systems in dentistry often focus on single modalities or specific tasks, and while some vision-language models can answer dental questions, their responses frequently lack explicit, traceable evidence.To address these limitations, researchers have developed DentAgent, an innovative evidence-centric multi-agent framework. This system features an Orchestrator that coordinates five specialized agents, each designed to handle different modalities. These specialists use domain-specific tools to convert raw observations into structured evidence records.A central "Evidence Blackboard" manages these records, serving as a shared state that tracks evidence coverage, identifies gaps, and resolves conflicts before generating a final response. This standardized evidence representation allows for a unified workflow that integrates previously isolated dental AI capabilities. DentAgent has demonstrated superior performance across four benchmarks, notably outperforming senior human specialists by 17.3 percentage points in multi-label diagnosis, highlighting its potential for broad application in population oral health assessment and management.