An exemplification of the deployed AI large model for nasopharyngeal–otolaryngology at the launch event. (Photo/CBG)

Chongqing - Chongqing has newly unveiled China's first AI large model for ear, nose, throat (ENT) care, developed by Chongqing University of Posts and Telecommunications (CQUPT). With up to 90% diagnostic accuracy, it streamlines clinical workflows by generating detailed electronic medical records.

AI-powered tools are already in use across Chinese hospitals for interpreting lung CT scans, assisting in breast cancer imaging, and retinal screenings. This development aligns with priorities outlined in the Healthy China 2030 initiative and China’s national AI strategy, both of which include provisions for advancing intelligent diagnostic tools in healthcare. Regulatory progress has also been evident, with China's National Medical Products Administration approving multiple AI-enabled devices in recent years, primarily in imaging and pathology.

Despite these gains, applications in ENT, particularly nasopharyngeal diagnosis, remain relatively limited. Nasopharyngeal-focused tools are still underrepresented compared to those targeting hearing loss, sinus conditions, or general head and neck imaging. CQUPT’s new model is among the first to address this clinical subfield in China directly.

Traditional ENT diagnostics still rely heavily on visual inspection and physician experience. This can result in missed early signs, inconsistent conclusions between practitioners, and a high reliance on senior-level clinical expertise. With rising patient volumes and increased attention to early-stage intervention, hospitals are beginning to explore AI-based tools as potential aids in enhancing diagnostic consistency and efficiency in ENT care.

According to Wang Wei, head of the research team and deputy dean of the School of Life and Health Information Science and Engineering, the model uses advanced algorithms and big data to learn from large volumes of clinical cases. It enables precise oversight of diagnosis and treatment, improving both consistency and care quality, particularly in areas where expert ENT clinicians may be scarce.

Early detection is essential for improving survival rates for nasopharyngeal tumours. "Our model analyzes and interprets imaging data in less than two seconds with up to 90 percent accuracy," Wang noted. It can detect subtle early-stage tumour features, supporting earlier diagnosis and intervention."

The system also simplifies documentation. Physicians input key clinical information, and the model auto-generates structured records. This significantly reduces the time required to produce standardized EMRs and contributes to greater efficiency in clinical workflows.

Notably, the model adopts an on-device AI architecture, enabling it to process massive datasets locally, even when working with limited data inputs. Unlike traditional AI models that depend on cloud infrastructure and high computational costs, this design allows for fast, secure, and real-time analysis on local devices. Because it is designed to operate locally, the system may also be suitable for hospitals with limited access to high-bandwidth infrastructure or large-scale cloud processing capacity.

"As clinical adoption expands, more real-world data will help us continue refining the model," Wang said. "Our goal is to ensure consistent diagnostic quality across hospitals and regions, so that wherever patients seek care, they receive reliable service."

The AI model has already been deployed in top-tier hospitals in Chongqing and Beijing. Developed over a two-year period by CQUPT in partnership with Beijing Jimai Health Technology, the project is now based in the university's science and technology park. Jimai also signed on as the 100th enterprise to join the park. As part of its commercialization path, the team plans to scale the model for additional hospital settings and explore integration with broader digital health systems.

Its release follows ongoing growth in China's AI healthcare sector, where companies such as InferVision, Yitu Healthcare, and iFLYTEK's Xiaoyi have introduced tools in radiology and general diagnostics. CQUPT's model contributes to this landscape by focusing on a specialized clinical area and employing an edge-computing approach aimed at flexibility and local deployment.