Using Single-View Wide-Angle DBT with AI for Breast Cancer Screening
Menée à partir de données portant sur 190 femmes (âge médian : 54 ans), cette étude évalue la performance d'un algorithme d'apprentissage profond pour aider les radiologues à interpréter des images de tomosynthèse mammaire numérique avec acquisition grand angle et procédure à incidence unique
Digital breast tomosynthesis (DBT) is increasingly used in both screening and diagnostic work-up for breast cancer. DBT from different vendors has different screening protocols approved by the U.S. Food and Drug Administration, ranging from two-view DBT combined with two-view two-dimensional digital mammography (DM), craniocaudal-view DM combined with mediolateral oblique (MLO)– view DBT, to two-view DBT without DM. Although DBT has been shown to increase detectability of masses and architectural distortions, radiologists still rely on two-dimensional DM for visualization of asymmetry, new density, and grouped microcalcifications in a single plane. With the advance in technology that can generate a two-dimensional DM-like synthetic mammogram (SM) from DBT, some studies have shown that SM can provide similar functions as DM, suggesting that SM may replace DM and reduce the associated radiation dose. SM is a no-cost companion to all DBT examinations, but some radiologists still prefer DM for detection and characterization of microcalcifications. Regardless of whether DM views are kept or replaced with SM in DBT, a screening examination that includes DBT may increase the reading time for radiologists, or the advantages of DBT may not be fully utilized if radiologists speed up their reading. The need to increase efficiency and diagnostic accuracy in the clinic has prompted studies of alternative imaging or reading protocols.
Radiology , éditorial, 2020