Automation of clip localization in Digital Tomosynthesis for setup of breast cancer patients

Abstract

The objective of this study is to develop an automatic clip localization procedure for breast cancer patient setup based on Digital Tomosynthesis (DTS) and to characterize its performance with respect to the overall registration accuracy and robustness. The study was performed under an IRB-approved protocol for 12 breast cancer patients with surgical clips implanted around the tumor cavity. The registration of DTS images to planning CTs was performed using an automatic algorithm developed to overcome specific challenges of localization and registration of clips in the breast setup images. The automatic method consisted of auto-segmentation (intensity-based thresholding with a priori knowledge about clip size and location to distinguish clips from bony features) and auto-registration of the segmented clip clusters. To determine the inherent accuracy and robustness of the registration algorithm, additional simulated DTS data was analyzed. The developed algorithm is efficient in removing false positives and negatives and provides an accuracy of better than 2.3mm for 60° and 3.3mm for 40° DTS. When incorporated in clinical software, this algorithm helps to facilitate fast and accurate setup evaluation with minimal dose delivered to patients.