RSNA 2015

Abstract Archives of the RSNA, 2015


SSA01-09

Clinical Performance of Dedicated Breast Computed Tomography in Comparison to Diagnostic Digital Mammography

Sunday, Nov. 29 12:05PM - 12:15PM Location: Arie Crown Theater



Elodia B. Cole, MS, Charleston, SC (Presenter) Consultant, FUJIFILM Holdings Corporation; Consultant, Koninklijke Philips NV; Consultant, Alan Penn & Associates, Inc
Amy S. Campbell, MD, Washington, DC (Abstract Co-Author) Nothing to Disclose
Srinivasan Vedantham, PhD, Worcester, MA (Abstract Co-Author) Research Grant, Koning Corporation
Etta D. Pisano, MD, Charleston, SC (Abstract Co-Author) Founder, NextRay, Inc CEO, NextRay, Inc Research Grant, Koning Corporation Research Grant, Koninklijke Philips NV Research Grant, Zumatek, Inc Research Grant, FUJIFILM Holdings Corporation Equipment support, Siemens AG Research Grant, Siemens AG Equipment support, Koninklijke Philips NV Research Grant, Koninklijke Philips NV
Andrew Karellas, PhD, Worcester, MA (Abstract Co-Author) Research collaboration, Koning Corporation
PURPOSE

To compare the clinical performance of a three-dimensional dedicated breast computed tomography system requiring no breast compression alone (dBCT), dBCT as adjunct to two-dimensional standard view screening mammography (SM), and two-dimensional diagnostic mammography (DxM).

METHOD AND MATERIALS

Eighteen radiologists interpreted 235 cases (52 negative, 104 benign, 79 cancer; 93/235 calcifications) that were randomly selected from 478 cases enrolled under 3 different clinical trial protocols, all in diagnostic population. Each case consisted of unilateral SM, DxM and dBCT images. Each case was randomized to 3 sessions and interpreted under 3 conditions: dBCT alone, dBCT plus SM, and DxM alone with at least a 4-week washout period. Each interpretation included an overall BIRADS score and continuous probability of malignancy (POM) score. For each case, any identified lesions assigned BIRADS category 3 or greater had its location, type, BIRADS and POM reported. Sensitivity, specificity and area under the ROC curve (AUC) were determined with either pathology or 1-year follow-up as truth.

RESULTS

All reported performance metrics were averaged across all readers. Results are reported from analysis using BIRADS score after dichotomizing at BIRADS 4.The sensitivity for dBCT alone was 81.78%, 87.93% for dBCT plus SM, and 84.07% for DxM. dBCT plus SM had significantly higher sensitivity than DxM (p=0.0081), and dBCT alone (p<0.0001). DxM and dBCT alone did not differ in sensitivity (p=0.1753).The specificity for dBCT alone was 49.67%, 39.65% for dBCT plus SM, and 44.84% for DxM. Neither dBCT alone (p=0.1148) nor dBCT plus SM (p=0.0745) statistically differed from DxM. dBCT alone had a significantly higher specificity than dBCT plus SM (p<0.0001). The AUC based on BIRADS (POM) were 0.716 (0.770) for dBCT, 0.723 (0.791) for dBCT plus SM, and 0.724 (0.792) for DxM. There were no statistically significant differences between the modalities based on POM (p=0.3311) or BIRADS (p=0.8569) score analyses.

CONCLUSION

The most effective use of dBCT for diagnostic imaging is as adjunct to standard view mammography.

CLINICAL RELEVANCE/APPLICATION

Dedicated Breast Computed Tomography has potential for use as a diagnostic breast imaging tool.