Gisella Gennaro PhD, Presenter: Nothing to Disclose

R. Edward Hendrick PhD, Abstract Co-Author: Consultant, General Electric Company Advisory Board, Bracco Group Medical Advisory Board, Koning Corporation

Cosimo Di Maggio MD, Abstract Co-Author: Nothing to Disclose

Manuela La Grassa MD, Abstract Co-Author: Nothing to Disclose

Luigi Pescarini MD, Abstract Co-Author: Nothing to Disclose

Ilaria Polico MD, Abstract Co-Author: Nothing to Disclose

Alessandro Proietti MD, Abstract Co-Author: Nothing to Disclose

To compare clinical performance with digital breast tomosynthesis (DBT) obtained by individual radiologists in two subsequent reading studies.

Two multi-reader studies were conducted, separated in time by 1.5 years, to compare two-view mammography to one-view DBT. The first and second study included 371 and 469 breasts respectively, from 250 patients with and without lesions. In each study, six breast radiologists were involved, five radiologists participating in both. Readers, blinded to any clinical information, evaluated images from the two modalities independently and classified eventual findings according to the BI-RADS lexicon. For the five readers involved in both studies, clinical performance was evaluated by receiver operating characteristics (ROC) and individual “operating points” (sensitivity/specificity pairs) to capture possible changes in radiologists’ approaches with tomosynthesis between the two studies. The primary additional DBT interpretation experience gained by each reader in the second study was their participation in the first study.

With mammography, individual operating points were virtually unchanged between the two studies: mean sensitivity-specificity pairs across the five readers were (Se 72.2%, Sp 85.1%) in the first study and (Se 74.4%, Sp 84.5%) in the second. With tomosynthesis, individual operating points changed from the first to the second study. Only one reader showed clinical performance slightly better in the first study, while four out of five readers improved their sensitivity at comparable or higher specificity in the second study. On average, DBT operating points were (Se 67.6%, Sp 86.1%) in the first study and (Se 74.8%, Sp 86.8%) in the second.

The shift of individual operating points over time demonstrates a learning process in DBT interpretation performance.

Radiologist's interpretation of images from DBT improved with clinical experience. Experience is necessary to allow readers to establish appropriate operating points with a new technology.

Gennaro, G, Hendrick, R, Di Maggio, C, La Grassa, M, Pescarini, L, Polico, I, Proietti, A, Digital Breast Tomosynthesis: Radiologists’ Learning Process.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12032585.html