Michael Galperin PhD, Presenter: Employee, Almen Laboratories, Inc

Linda Kay Olson MD, Abstract Co-Author: Research Consultant, Almen Laboratories, Inc Research Consultant, TechniScan Medical Systems

Annina Nicholas Wilkes MD, Abstract Co-Author: Nothing to Disclose

Michael Paul Andre PhD, Abstract Co-Author: Scientific Advisory Board, TechniScan Medical Systems Advisory Board, Almen Laboratories, Inc

Flemming Forsberg PhD, Abstract Co-Author: Equipment support, Toshiba Corporation Equipment support, Siemens AG Research collaboration, General Electric Company Research collaboration, Ultrasonix Medical Corporation Research collaboration, Toshiba Corporation Advisory Board, Siemens AG Advisory Board, Toshiba Corporation

Barry B. Goldberg MD, Abstract Co-Author: Nothing to Disclose

The developed Breast Ultrasound diagnostic tool performed exceedingly well in two independent “stress test” evaluations. BC application promises to reduce well-known variability of radiologists, increase overall accuracy of reading performance by using a more structured protocol and flow, and therefore create an environment that would lead to substantial reduction in the number of biopsies for benign masses (reduction of False Positives)

Breast Companion® (BC) is diagnostic tool that assists radiologists in their interpretation during BI-RADS assessment. BC has been extensively validated for diagnostic breast sonography. As part of clinical utility valuation two luminary institutions (TJU & UCSD) completed an independent "stress test" evaluation of the tool performance and its improvement of the existing reading practice

BC utility was valuated using "stress test" cohorts retrieved from the database of confirmed cases. The 1st cohort: 82 cases with 15% cancers. The 2nd cohort: 80 cases with 43% cancers. Both cohorts represent statistically acceptable "stress test" deviations (average cancer prevalence for ultrasonography is 20-25%%). The two-step testing protocol was used: 1. Images were analyzed with BC in three separate sessions. The radiologist segmented the margins of the mass using the manual tool or if preferred, the automatic segmentation tool 2. After completing the BI-RADS Assessment report the radiologist was asked to allow or not to allow (if they disagree) BC’s computerized lesion assessment (CLA) score to be displayed and included into the final report for the segmented mass produced earlier. The CLA scores from the reports were analyzed by applying ROC analysis and compared

Accuracy ROC curve for the scores produced by the tool and accepted by the radiologists was made. Numeric outcomes show that the computed scoring accuracy reached its maximum AUC ROC 95% and 87% on “stress test” cohort with 15% and 43% cancers respectively (SE 4%, p<0.0001). Those results were compared to the radiologists’ performance on the same set of cases in clinical environment without use of the tool. They achieved AUC ROC of 89% and 71% for the correspondent “stress test” cohorts (SE 6%, p<0.0006)

Galperin, M, Olson, L, Wilkes, A, Andre, M, Forsberg, F, Goldberg, B, Practical Improvement of BI-RADS Interpretation and Assessment Reading While Using Validated Breast Ultrasound Diagnostic Scoring Tool as Decision-making Support System.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11003249.html