Digital breast tomosynthesis (DBT) was approved by the FDA in 2011 as a screening modality for the early detection of breast cancer. There are now more than 11,000 FDA-accredited DBT units in the USA, with over 80% of breast screening clinics offering this modality. Despite its proven advantages over full-field digital mammography (FFDM), DBT does suffer from the challenge of many acquired images and consequently longer interpretation times. Artificial intelligence (AI) algorithms have come to the fore in the past five years and now have the potential to boost the significant benefits of DBT.
A group of radiologists from New York City led by Dr Julia Goldberg has recently published an article in RadioGraphics in which they reviewed the ways that AI could enhance screening for breast cancer with DBT. As Dr Manisha Bahl pointed out in an invited commentary, there are currently more than 10 FDA-cleared applications of AI for DBT. Kate Madden Yee has highlighted in an AuntMinnie.com article that Goldberg and colleagues identified at least seven ways that AI shows significant promise when applied to DBT images.
First, the combination of DBT and AI will improve the sensitivity – and therefore the accuracy – in detecting breast cancer. They wrote that “the use of AI algorithms results in noninferior or improved sensitivity for cancer detection compared with sensitivity of traditional methods of reading screening DBT studies in the clinical setting.” Second, there will be a decreased workload in which a triage system labels all images and only abnormal images are evaluated by the radiologist.
Third, there should be decreased recall rates, where the authors cited a clinical simulation study based on over 13,000 DBT examinations. Fourth, AI algorithms have the potential to classify abnormal findings in which bounding boxes may be associated with different confidence levels for malignancy (see above left, © RSNA). Fifth, suspicious lesions will be better visualized when they have been identified and projected onto a synthesized 2D image.
Sixth, there is the potential to reduce the radiation dose while also improving image quality. Finally, the use of AI will enable breast cancer risk to be assessed, particularly in women who have dense breast tissue. In her commentary, Bahl (seen right) suggested that “the review article by Goldberg and colleagues is a welcome addition to the growing breast imaging literature on AI.” Ultimately, of course, the combination of DBT and AI should lead to improved health outcomes for women whose cancer is detected early enough.