Insight

AI Assistant Supports Pathologists in Cancer Detection at Zhejiang University

chinauinfo.com

Dec 27, 2024, 13:56

AI Assistant Supports Pathologists in Cancer Detection at Zhejiang University

Zhejiang University has introduced an innovative AI-powered pathology assistant called OmniPT, designed to assist pathologists in diagnosing high-incidence cancers like gastric, colorectal, and cervical.

The technology integrates advanced vision and language models, allowing for improved human-computer interaction.

It’s currently being used at the First Affiliated Hospital of Zhejiang University School of Medicine, the first clinical institution in China to adopt this AI-powered tool.

OmniPT has made significant strides in cancer diagnostics, including classifications, grading, and identifying vascular and neural invasion.

It can also detect markers indicating the future progression of the disease.

According to the hospital, OmniPT’s predictions are 80 to 90 percent accurate across various cancer types.

The role of pathologists is crucial in understanding the nature of diseases from tissue samples, but China is facing a severe shortage of professionals in this field.

With only 30,000 registered pathologists in the country, far below the needed 150,000 to 200,000, the shortage is particularly acute in remote areas.

Young pathologists also often lack experience, adding to the problem.

OmniPT helps address this shortage by streamlining diagnostic processes and allowing pathologists to focus on critical decisions.

The AI assistant takes over repetitive tasks, such as counting mitosis in glioma slides, a task that can take an hour manually but is completed in less than 10 seconds by OmniPT.

This enhances both speed and accuracy in diagnosing complex conditions, while also alleviating the burden on pathologists.

Song Mingli’s team at Zhejiang University, in collaboration with the hospital, developed OmniPT with a focus on clinical needs.

The tool speeds up sample evaluation, ensuring that even subtle details, which could be overlooked due to fatigue or time pressure, are detected.

While OmniPT handles over 90 percent of repetitive tasks, pathologists remain in control, using their expertise for final judgments.

This collaboration is improving diagnostic efficiency, reducing costs, and helping doctors in underserved regions deliver better healthcare.