Nanjing releases medical world model appliance and five chronic-disease models

On the afternoon of October 19, the medical and health world model session of the 2026 Physical AI and World Models Conference took place at the Nanjing International Expo Center. Under the theme “Foresee disease, simulate life,” it brought together academicians, clinicians, university researchers and representatives from the AI, pharmaceutical and data industries.

Unlike medical AI built around single-point detection and static prediction, the session examined how a patient's state changes over time and what different interventions might produce. Eight keynotes ran from clinical need to modeling methods, research validation and application, covering chronic disease management, continuous-domain generalization, rehabilitation forecasting, ontology-based modeling, wet-lab validation, gene-data target discovery, embodied models for laparoscopic surgery and synthetic treatment trajectories.

The concrete products arrived in the results segment. Jiangsu Province Hospital and Nanjing Credible Data Services, part of Nanjing Big Data Group, released a medical and health world model appliance together with H3C, UnionTech Software and YashanDB, packaging world model technology with domestic computing hardware, operating systems and databases into a platform hospitals can run locally.

Jiangsu Province Hospital and Nanjing Credible Data Services also released five chronic-disease world models covering insulin resistance, diabetes medication, continuous hypertension management, lipids and atherosclerosis, and chronic kidney disease risk. Wuzhou Workers' Hospital and the Guangxi Institute of Precision Medicine released a thalassemia screening world model designed to keep raw data inside the hospital, with the screening capability later extended to medical cooperation with ASEAN partners.

Data infrastructure featured prominently. The Nanjing Health Commission information center, with Jiangsu Province Hospital, Nanjing Drum Tower Hospital and several other municipal hospitals, launched a joint plan to build high-quality medical world model datasets under compliance and privacy safeguards. Southeast University Zhongda Hospital and Yidu Tech released a high-quality data foundation focused on quality governance, unified medical semantics and temporal information.

Applications followed: Neusoft released a city-level digital health solution built on its Tianyi medical model, and Nanjing Hanwei Public Health Research Institute released a world-model-based medical research product. Together they trace the practical gates for medical world models: data, compute, models and scenarios.

The organizers were notably measured. The released results still require external validation, clinical evaluation and ongoing regulatory review to define their scope, and model simulation cannot replace physicians' clinical decisions. For technology that reaches into diagnosis and treatment, that caveat is a prerequisite rather than a formality.

The bigger signal is where the deployment happens: inside hospital data centers, on domestic hardware and software. What to watch is whether these models clear external validation on real clinical data, whether the appliance can be replicated across more hospitals, and whether the cross-border ASEAN arrangement proves workable.