Errors from radiology trainees increase when interpreting digital breast tomosynthesis scans deemed to be difficult by expert imaging readers, according to results of a new study published online in the Journal of the American College of Radiology.
Tomosynthesis continues to see increased adoption as a breast imaging modality due to its advantages over traditional mammography. In turn, the ability to confidently interpret digital breast tomosynthesis is now a critical part of the women’s imaging skill set for radiologists.
But training future radiologists how to interpret an emerging imaging tool is sometimes tricky, according to lead author Lars Grimm, MD, and his colleagues from Duke University Medical Center in Durham, N.C.
“[Digital breast tomostynthesis] is similar to mammography in principle and relies on the same BI-RADS lexicon, but many of the rules for mammography do not apply to [digital breast tomosynthesis],” the authors wrote. “Incorporating [digital breast tomosynthesis] into existing educational training is a unique challenge, in part because this is also a new imaging modality for radiology faculty members.”
Grimm and his team wanted to better understand radiology trainee performance by investigating the relationship between error making and difficulty for digital breast tomosynthesis. To do so, they asked 27 radiology residents with no training on tomosynthesis and three expert breast imagers to review 60 digital breast tomosynthesis studies. Each reader provided assessments of difficulty and BI-RADS scores for each study, with expert consensus interpretations used to determine the ground truth. Calculations were then made for trainee sensitivity, specificity and area under the receiver operating characteristic curve (AUC).
They found that the radiology trainees increase their sensitivity to compensate for decreased specificity when presented with tomosynthesis cases they deemed to be more difficult, while overall trainee performance decreased most in cases that expert breast radiologists deemed to be more difficult.
“These relationships were irrespective of years of training, indicating that they are independent of prior breast imaging and general radiology experience,” the researchers added.
Grimm and his colleagues believe the results should be useful to those designing radiology training curriculum and will help improve radiologist performance as the use of tomosynthesis continues to grow.
“Educators should include knowledge of these findings when developing educational materials to improve trainee performance,” the authors concluded. “Notably, they should focus teaching approaches to [digital breast tomosynthesis] interpretations and select cases they find to be particularly difficult to maximize their teaching yields.”