US Pharma MarketingIt establishes a preclinical gateway for testing LLM safety without risking actual patients. Researchers from Seoul National University Hospital and Harvard Medical School have created what could be the world's first virtual hospital framework for validating large language model-based medical AI. The Clinical Environment Simulator "dynamically evaluates" medical AI using two synchronised core engines. The Patient Engine prompts the LLM to generate various virtual symptom paths and treatment responses based on disease trajectory templates defined by specialists and initial patient data from EMR. The Hospital Engine replicates the actual step-by-step workflow based on real hospital time data, tracking bed status,…
Researchers from Seoul National University Hospital and Harvard Medical School have created what could be the world's first virtual hospital framework for validating large language model-based medical AI.
HOW IT WORKS
The Clinical Environment Simulator "dynamically evaluates" medical AI using two synchronised core engines.
Source: MobiHealthNews