An Auto-Accessing Method for Reducing the Reading Time of Digital Breast Tomosynthesis with a Synthetic Mammogram

Abstract

Purpose: While combinatorial acquisition of digital mammogram (DM) and digital breast tomosynthesis (DBT) increases breast cancer diagnosis performance, this combo mode acquisition raises concerns on increased data acquisition time, interpretation time, and radiation dose. One possible solution to reducing radiation dose is compositing a synthetic mammogram (SM) from the DBT slices. Accessing corresponding DBT slice-in-focus of a selected region-of-interest (ROI) in the SM image is important in such a diagnostic approach. In this study, we propose an efficient method for automatically accessing DBT slice-in-focus of the selected areas in the SM image.

Methods: In synthesizing a mammogram from DBT slices, we detected linear structures in DBT slices by use of the oriented bins operator and generated a linear structure weighting map accordingly. We computed a weighted forward projection by use of the linear structure weighting map. For an auto-accessing method, we additionally generated a linear structure weighted maximum intensity projection (LS-MIP) and a linear structure weighted maximum intensity depth map (LS-MID). For a selected ROI in the SM image, we estimated the depth of the object using the LS-MIP and the LS-MID. Since ROI would also contain background regions, we only took pixels having the top 10% high values inside the ROI in the LS-MIP. We collected the values of the chosen pixels in the LS-MID and calculated the counts of each depth. After that, we displayed the reconstructed DBT slices at the calculated top three counted depths. For evaluation, we used the CIRS breast phantom and patient data.

Results: The proposed auto-accessing method accurately figured the depth of selected ROIs from the SM image in the DBT slices.

Conclusion: We have developed an auto-accessing method to seek reconstructed DBT slices contributing mainly to the selected ROI in the SM, which is supposed to reduce the reading time of DBT.