In a groundbreaking development, a team from the Chinese mainland has unveiled PathOrchestra, the country’s first versatile artificial intelligence (AI) model designed to analyze a wide range of pathological images. This state-of-the-art model can examine over 20 human organs, including the lungs, breast, and liver, marking a significant advancement in AI-assisted disease diagnosis.
PathOrchestra represents a pivotal shift from specialized AI models focused on specific cancers to a more adaptable system capable of addressing multiple cancer types. Developed by researchers from Air Force Medical University (AFMU), Tsinghua University, and SenseTime, the model leverages China’s largest domestic dataset, consisting of nearly 300,000 whole-slide digital pathology images amounting to an impressive 300 terabytes of data.
Utilizing self-supervised learning, PathOrchestra has been trained to perform a variety of clinical tasks. These include pan-cancer classification, lesion identification, multi-cancer subtype differentiation, and biomarker assessment. The complexity of analyzing diverse pathological images has earned this field the nickname “jewel in the crown” of image processing, according to Wang Zhe, a professor at the Basic Medical Science Academy under AFMU.
The model has achieved an accuracy rate exceeding 95 percent in nearly 50 clinical tasks, such as lymphoma subtype diagnosis and bladder cancer screening, as reported in an AFMU news release. This high level of accuracy not only enhances diagnostic precision but also significantly reduces the workload for pathologists, thereby increasing efficiency in medical image reviews.
PathOrchestra stands as a testament to the rapid and vibrant growth of AI in the Chinese mainland. With 36 percent of the world’s more than 1,300 AI large language models (LLMs) originating from China, the nation holds the second-largest share after the United States, according to a recent white paper by the China Academy of Information and Communications Technology presented at the Global Digital Economy Conference 2024.
Reference(s):
Chinese team develops AI model capable of reading cancer images
cgtn.com
