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 recognize multiple types of cancer without requiring additional training and using minimal samples. This breakthrough greatly increases the flexibility and efficiency of AI-assisted medicine and represents a major step toward widespread adoption of intelligent pathology.
Approximately 20 million new cancer cases are diagnosed worldwide each year, and pathological examinations play a vital role in clinical diagnosis and treatment decision-making. However, with a severe global shortage of pathologists, the medical community increasingly needs innovative solutions to improve the efficiency of pathology analysis.
Although AI has great potential to automate pathological diagnosis, multiple challenges still constrain its practical implementation. Traditional AI models typically require training by collecting tens of thousands of pathology images and datasets for each specific cancer type or diagnostic task, resulting in long development cycles and significant computational and human costs. Furthermore, existing basic pathology models often lack sufficient generalizability and require extensive fine-tuning when applied to different tumor types in real-world clinical settings, limiting their scalability and adoption, especially in resource-constrained regions.
To address these challenges, a research team led by Professor LI Xiaomeng, assistant professor in the Department of Electrical and Computer Engineering and deputy director of the Hong Kong Academy Medical Image Analysis Center, collaborated with Guangdong Provincial People’s Hospital and Harvard Medical School to develop a new pathology analysis system called PRET (Pan Cancer Recognition without Sample Training).
This system is the first to introduce the concept of “in-context learning” from natural language processing to pathological image analysis. This allows the model to instantly adapt to new cancer types and perform diagnostic tasks such as cancer screening, tumor subtyping, and tumor segmentation by simply referencing one to eight annotated tumor slides during the inference stage. Acting as a “plug-and-play” intelligent diagnostic tool, PRET fundamentally overcomes the need for task-specific fine-tuning in traditional AI models.
The research team conducted a comprehensive validation of the PRET system covering 18 cancer types and various diagnostic tasks using 23 international benchmark datasets from medical institutions in mainland China, the United States, and the Netherlands. As a result, the system outperformed existing methods on 20 tasks, and the area under the curve (AUC), a measure of diagnostic accuracy, exceeded 97% on 15 of those tasks.
In particular, PRET achieved 100% AUC for colorectal cancer screening and 99.54% AUC for 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 8 slide samples, outperforming the average performance of 11 pathologists (average AUC of approximately 81%). Furthermore, PRET demonstrated stable and robust generalizability across different populations and regions with varying levels of health care resources.
Professor Li Xiaomeng said, “The core value of the PRET system lies in breaking the traditional barrier of ‘large amounts of data and repetitive training’, and making AI-powered pathology systems applicable to real clinical settings at lower cost and with greater flexibility.”
This not only reduces the workload pressures 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 will transcend geographic and resource constraints, thereby promoting global health equity. ”
Li Xiaomeng, Hong Kong University of Science and Technology
Looking to the future, the research team plans to further enhance the system’s diagnostic performance and expand its application to additional clinical tasks such as gene mutation prediction and patient prognosis assessment, opening new directions for the future of AI-driven pathology diagnosis.
Research results have been published in leading international journals natural cancer.
sauce:
Hong Kong University of Science and Technology
Reference magazines:
Lee, Y. others. (2026). PRET is a few-shot system for pan-cancer recognition without sample training. natural cancer. DOI: 10.1038/s43018-026-01141-2. https://doi.org/10.1038/s43018-026-01141-2.
