A new AI tool could help detect breast cancers that radiologists might miss on MRI scans. According to a study published in Academic Radiology, this AI tool could spot up to 30% more cancers by providing a “second look” at MRI images. The authors believe that the AI’s re-evaluation may reveal early cancer signs that radiologists sometimes overlook, potentially leading to earlier treatment.
Studies show that many breast cancers (up to 75% in some cases) are visible on MRI scans before they are officially diagnosed. While radiologists are skilled at spotting these cancers, they sometimes miss subtle signs. This new AI model, developed by researchers at City College of New York, was trained on over 3,000 breast MRI scans, including 115 cases where cancer was diagnosed within a year of a negative MRI.
The goal of the AI tool is to identify cancers that radiologists might miss up to a year sooner. In tests, the tool showed promising results, detecting high-risk cancers with an accuracy score (AUC) of 0.72. It also correctly identified the area where cancer would later develop in over half of the cases. If validated, this tool could help detect more cancers without adding extra work for radiologists, potentially leading to earlier treatments for high-risk patients.
The researchers highlight that early detection is crucial for better treatment outcomes. Although re-evaluating MRIs takes effort, the added benefit of earlier detection may outweigh this extra work. The team believes this AI approach will become even more useful as more data becomes available.
Source: healthimaging
