RSNA: Automated breast ultrasound system provides reproducible images
“Ultrasound is a widely adopted tool for the characterization of breast lesions in conjunction to mammography,” noted Jung Min Chang MD, of the Breast Cancer Center at the Seoul National University Hospital in Seoul, Korea, who presented the study. However, she noted that "increased cancer detection is possible by addition of screening ultrasound in high risk patients.”
However, Chang noted that problems still exist in regard to operator tendency, which often has a high degree of error. The use of ABUS was designed to curb these problems and can also facilitate reproducible images.
The purpose of the study was to “retrospectively evaluate the reproducibility of data acquisition of an automated breast ultrasound system for mass detection, localization and characterization,” Chang stated.
The researchers identified 24 patients with 33 breast lesions between October 2007 and March 2008. Two trained technicians performed the image acquisition and lateral and medial scans for each breast. The technicians then reviewed 3D coronal volumetric data and assessed image quality on a five-grade scale.
The technicians reviewed the image quality and lesion visibility of the 3D images using a side-by-side display reported a total of 24 breast cancers and nine benign pathologies. In addition, they used a five point scale and reported the reproducibility of the documented location—clock face, distance from nipple and lesion depth—and the size of lesions and lesion similarity classified as identical, similar or different.
Results showed that the mean image quality score was 4.6 and there were three reported cases of poor image quality.
“Image qualities were good to excellent in all cases except three with focal defects that might affect lesion detection,” Chang said.
Of the 24 patients, two had mass lesions in the periphery that were not included in one of the paired 3D volume data. As for lesion similarity, the technicians reported 21 (16 cancers and five benign) that were assessed as identical, eight (six cancers and two benign) that were identified as similar and two benign that were classified as different.
A ‘similar’ classification referred to lesions that were not identical but had the same BI-RADS US lexicon and defined a classification of ‘different’ to be lesions that had at least one different BI-RADS features that led to a difference in the final assessment.
“ABUS provided reproducible images for mass localization and size measurements,” concluded Chang. “For better reproducibility current placement of the receptor is important and adjustment of scanning pressure is still essential.”