A research team led by the Hong Kong University of Science and Technology (HKUST) has developed a pioneering artificial intelligence (AI) pathology analysis system that can accurately identify many types of cancer using only a minimal number of samples—without the need for additional training. This breakthrough greatly enhances the flexibility and efficiency of AI-assisted medical care, marking an important step forward towards the widespread adoption of intelligent pathology.
Each year, nearly 20 million new cases of cancer are diagnosed worldwide, with pathological examination playing a key role in clinical diagnosis and treatment decision-making. However, amid a severe global shortage of pathologists, the medical community increasingly needs innovative solutions to improve the efficiency of pathological analysis.
While artificial intelligence has great potential for automating pathological diagnostics, its practical deployment remains limited by multiple challenges. Conventional AI models typically require the collection and training of tens of thousands of pathology images and training datasets for each specific cancer type or diagnostic task, resulting in long development cycles and significant computational and manpower costs. In addition, existing fundamental pathology models often lack sufficient generalization, which requires extensive detail when applied to different tumor types in real clinical settings, thus limiting their scalability and adoption, particularly in resource-constrained areas.
To address these challenges, a research team led by Professor LI Xiaomeng, Assistant Professor of the Department of Electronic and Computer Engineering, and Associate Director of the Center for Medical Imaging and Analysis at HKUST, in collaboration with Guangdong Provincial People’s Hospital and Harvard Medical School, developed a new Receranco PRETcan analysis system (Example Training).
The system is the first to introduce the concept of “learning in context” from natural language processing to pathological image analysis. It allows the model to be instantly adapted to new cancer types and to perform diagnostic tasks, such as cancer screening, tumor subtyping, and tumor segmentation, during the inference stage by reporting only one to eight annotated tumor slides. Acting as an intelligent plug-and-play diagnostic tool, PRET essentially overcomes the need for task-specific fine-tuning in traditional AI models.
The research team performed comprehensive validation of the PRET system using 23 international reference datasets from medical institutions in mainland China, the United States, and the Netherlands, covering 18 types of cancer and various diagnostic tasks. The results showed that the system outperformed existing methods in 20 tasks, with the area under the curve (AUC)—a measure of diagnostic accuracy—exceeding 97% in 15 of those tasks.
Specifically, PRET achieved an AUC of 100% in colon cancer screening and an AUC of 99.54% in esophageal squamous cell carcinoma tumor segmentation. In the extremely difficult task of detecting lymph node metastases, PRET achieved an AUC of approximately 98.71% using only eight slide samples, surpassing the average performance of 11 pathologists, whose AUC averaged approximately 81%. In addition, PRET has shown consistent and robust generalizability to different populations and regions with different levels of medical resources.
Professor Li Xiaomeng said, “The core value of the PRET system lies in breaking down the traditional barriers of ‘bulk data and iterative training’, allowing AI-powered pathology systems to be applied to real clinical settings with lower cost and greater flexibility.”
This not only helps alleviate the workload faced by pathologists, but also has the potential to improve access to cancer diagnosis in underserved areas. Through this plug-and-play system, we hope that advanced and accurate AI-powered diagnostic services can transcend geographical and resource limitations, thereby promoting global healthcare equity.”
Li Xiaomeng, Hong Kong University of Science and Technology
Looking to the future, the research team plans to further improve the system’s diagnostic performance and expand its applications to additional clinical tasks, such as predicting genetic mutations and assessing patient prognosis, opening up new directions for the future of AI-based pathology diagnosis.
The research findings have been published in the leading international journal Nature Cancer.
Source:
Journal Reference:
Lee, Y., et al. (2026). PRET is a partial download system for pancancer recognition without a training example. Nature Cancer. DOI: 10.1038/s43018-026-01141-2. https://doi.org/10.1038/s43018-026-01141-2.
