Michael J. Michell MBBCH, Presenter: Nothing to Disclose

Asif Iqbal MBBS, Abstract Co-Author: Nothing to Disclose

Rema Wasan MBBCH, Abstract Co-Author: Nothing to Disclose

Abdel Douiri PhD, Abstract Co-Author: Nothing to Disclose

David Evans MD, Abstract Co-Author: Nothing to Disclose

Clare Peacock MBBS, Abstract Co-Author: Nothing to Disclose

Juliet Claire Morel, Abstract Co-Author: Nothing to Disclose

To compare reader performance between single view (MLO and CC alone) full-field digital mammography (FFDM) and single view (MLO and CC alone) digital breast tomosynthesis (DBT).

In this blinded study, 154 cases comprising normal, benign and malignant lesions were randomly selected from a database of images of recalled women who underwent bilateral FFDM and DBT in the MLO and CC and projections using the Hologic Selenia Dimensions unit. 13 participating readers were randomly recruited from a pool of 80 specialist breast screening radiologists. Each reader read the series of 154 cases twice, one month apart, in this sequence: read one, cases 1-77 FFDM and 78-154 DBT; read two, cases 154-78 FFDM and 77-1 DBT. Readers were instructed to separately score lesions first on the MLO and then the CC views. Readers were not allowed to revert to previous view. The ground truth of the cases was determined by the outcome of assessment with histology where relevant.   Using the Royal College of Radiologists Breast Group classification of five point mammographic rating scale, a multi-reader multi-case (MRMC) receiver-operating characteristic (ROC) analysis was performed by using the Wilcoxon nonparametric curve fitting methodology.  

Reader-by-reader and overall area under the curves (AUCs) were evaluated. For CC view, the averaged AUCs of the readers were 0.78, 95% Confidence Interval (CI) [0.71-0.849] for FFDM and 0.82, 95% CI [0.766-0.875] for DBT [difference 0.0407] with p-value 0.051. The overall AUCs for MLO view were 0.83, 95% CI [0.776-0.891] and 0.87, 95% CI [0.827-0.912] for FFDM and DBT respectively [difference 0.036] with p-value 0.129.

Our study demonstrated that there is borderline statistical significant difference between diagnostic accuracy of DBT and FFDM on single view CC assessment. However no statistical significant difference was observed for single view MLO reading. Further reader studies are required to assess the accuracy of combined FFDM and DBT reading.

The addition of breast tomosynthesis to digital mammography may improve the diagnostic accuracy of breast screening and further workup of recalled clients and symptomatic cases.

Michell, M, Iqbal, A, Wasan, R, Douiri, A, Evans, D, Peacock, C, Morel, J, Phase II Trial: A Multireader Multicase Study Comparing the Diagnostic Performance of Full-Field Digital Mammography with Digital Breast Tomosynthesis.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11015113.html