AI Model DeepGEM 2.0 Predicts Lung Cancer Gene Mutations in One Minute
Chinese researchers have launched an AI model that can predict lung cancer gene mutations from routine digital pathology slides in as little as one minute, potentially reducing genetic testing costs from thousands of yuan to a few hundred yuan.
The multimodal large model, DeepGEM 2.0, was jointly developed by the National Center for Respiratory Medicine and KingMed Diagnostics. It can pre-screen for genetic mutations associated with all targeted therapies approved for lung cancer in China, helping identify patients who may benefit from more focused genetic testing, according to Dr. Liang Wenhua, deputy director of the NCRM.
The model analyzes routine digital pathology slides without requiring additional tissue samples, with prediction accuracy reaching as high as 90 percent. Patients identified as having a high probability of specific genetic mutations can undergo targeted genetic testing, cutting costs from several thousand yuan, or up to CNY20,000, to a few hundred yuan while shortening waiting times.
Liang noted that targeted therapies can significantly extend survival of lung cancer patients, but genetic testing is a prerequisite. Conventional genetic testing is often expensive, time-consuming, may require multiple tissue sampling, and has limited coverage. By identifying likely mutations in advance, DeepGEM 2.0 allows doctors to prioritize smaller genetic testing panels covering a few or dozens of genes, reducing unnecessary expenses and helping patients obtain confirmation sooner.
