Guangdong's national pilot base for medical AI applications has deployed 84 vertical medical and health models in the year since its launch, officials said at a press conference on the base's progress in Guangzhou on October 21.
The 84 vertical models form a matrix covering public services, clinical care and public health, accelerating the broader application of artificial intelligence in healthcare.
The base, launched in Guangzhou in September 2025, has built a digital foundation of one portal and five platforms and deployed 180 petaflops of domestically developed intelligent computing capacity. It has also set up a closed loop that takes applications from research to pilot testing and trial use in hospitals.
At the Third Affiliated Hospital of Sun Yat-sen University, a series of medical AI applications has already produced results, including a cloud-based pathology platform for AI-assisted diagnosis of sinusitis, AI-assisted diagnosis and treatment for autism spectrum disorder, and a hospital-developed AI doctor.
Yang Qintai, president of the hospital, said the hospital has trained its AI doctor on decades of clinical experience and 1.4 million real cases, building more than 470 disease-specific models. Patients can scan a QR code to receive AI pre-consultation. The AI doctor currently achieves diagnostic accuracy of more than 85 percent for neurological, liver and heart diseases, he said.
"The pilot base has provided computing power, helped optimize models and ensured stable operation, enabling the hospital to put AI into clinical use," said Yang.
To improve the management of peritoneal dialysis patients, the First Affiliated Hospital of Sun Yat-sen University released what it calls the world's first peritoneal dialysis large model and a supporting AI assistant in March 2025.
Chen Wei, vice president of the hospital and a leading expert in nephrology, said the model and assistant draw on more than 60 years of the hospital's kidney disease research knowledge base and clinical big data.
"They can assist primary-level medical staff, support online training, provide warnings of disease risks, and use a smartphone-based visual recognition function for home screening of peritonitis, helping patients conduct preliminary checks without leaving home," Chen said.
With deep cooperation and computing support from the pilot base, the rollout of the peritoneal dialysis model is accelerating. In September, the hospital established a national intelligent management network for peritoneal dialysis, with 93 medical centers included in the first batch. Related systems are being deployed gradually.
Chen said that with computing support from the base, the team plans to expand the model to more regions by 2027 to benefit more peritoneal dialysis patients.
Reporter & Photos | Chen Jinxia
