Advancements in cancer screening through blood tests and multimodal AI

Recent developments in cancer screening highlight both the potential and the challenges of new diagnostic technologies. In the United States, the PATHFINDER 2 study involving approximately 32,000 asymptomatic individuals found that the Galleri blood test, which evaluates cell-free DNA methylation patterns, achieved a positive predictive value of 60.3%.

However, large-scale trials in England, such as the NHS-Galleri trial involving 142,000 adults, failed to meet the primary endpoint of reducing late-stage diagnoses across three annual rounds. While the trial detected more early-stage malignancies, including colorectal and ovarian cancers, the absolute patient counts remained small, and the reduction in stage IV incidence lacked formal statistical significance.

In parallel, researchers at Sun Yat-sen Memorial Hospital in China have developed a multimodal AI framework to improve breast cancer screening. By integrating ultrasound (US), digital mammography (DM), and digital breast tomosynthesis (DBT), the deep learning model achieved high diagnostic accuracy. The US-DBT model configuration demonstrated the highest performance, reaching an area under the receiver operating characteristic curve (AUC) of 0.944 in internal validation and 0.934 in a pathology-confirmed cohort.