FRIDAY, May 23, 2025 (NestheoNews) — A study published online on April 18 in the journal indicates that artificial intelligence (AI) frequently identifies and precisely pinpoints interval breast cancers (IBCs), which can be seen during screenings, with greater accuracy compared to true interval or hidden cancers.
National Cancer Institute Journal
.
Dr. Tiffany T. Yu from UCLA collaborated with others to identify inflammatory breast cancers (IBCs) detected less than twelve months following a clear mammogram. These cases came from both digital mammographies and digital breast tomosyntheses performed between 2010 and 2019 at an American tertiary-care teaching hospital. The researchers categorized these IBCs retroactively into categories such as reading errors, minor signs requiring action, minor non-actionable signs, genuine intervals, hidden occurrences, or technical mishaps. A deep learning artificial intelligence system rated each initial negative screening mammogram with risk scores ranging from one to ten; those scoring eight or above were marked as flagged.
Out of 184,935 screening mammograms, 148 cases of invasive breast cancer (IBCs) were detected in 148 different women. These cancers fell into several categories: minimal signs requiring action accounted for 26%, occult cancers made up 24%, minimal signs not needing immediate action comprised 22%, reading mistakes led to 17% being missed, true intervals contributed 6%, and technical errors caused another 5%. In an analysis using artificial intelligence (AI), out of 131 reviewed mammogram images (excluding 17 deemed erroneous), instances involving missed readings due to reader error, actionable minor signs, and non-actionable minor signs were highlighted most often by the AI system—occurring at rates of 90%, 89%, and 72% respectively. Furthermore, when compared, visually apparent tumours had better localization accuracy via AI as opposed to those that weren’t visibly detectable; precise locations ranged between 35% to 68% success rate versus only 0% to 50% for less obvious ones.
The authors explain that AI could help identify mammographically detectable inflammatory breast cancer cases (including missed readings and subtle signs) during screenings, potentially decreasing the occurrence of these cancers so that they mainly consist of true interval cancers.
The research received partial funding through a grant from EarlyDiagnostics, which was awarded to one of the authors.
Summary/Full Content (access might require subscription or payment)