Automated identification of a radiological beads in breast tumors for purpose of calibrating quantitative ultrasound

Abstract

Quantitative Ultrasound (QUS) is an imaging method that has demonstrated the ability to detect the response of breast cancer to neoadjuvant chemotherapy. QUS is sensitive to tumor cell death, which is a hallmark of response. Traditionally, QUS requires an external reference material, e.g., reference phantom, to calibrate the system for backscatter coefficient based parameters. However, this approach does not account for losses in vivo and requires additional scans. To alleviate these issues, we used an in situ calibration target, i.e., a 2-mm diameter titanium bead, to obtain a reference signal. To automate the procedure to locate the bead in a B-mode image, we constructed a data mining method to distinguish the bead signal from other similar signals recorded from tumors having an embedded bead. The method was able to successfully locate the bead signal in a tumor image with good accuracy.