Automated breast ultrasound system (ABUS): reproducibility of mass localization, size measurement, and characterization on serial examinations.

Abstract

Automated breast ultrasound (ABUS) is gaining popularity for breast cancer detection and diagnosis with its high reproducibility. Consistent recognition of breast lesions on serial exams is critical, and high reproducibility for lesion characterization is expected with ABUS, but not well reported. To retrospectively evaluate the reproducibility of ABUS for mass localization, size measurement, and characterization. Bilateral whole breast US images were obtained twice using a commercially available ABUS within a mean interval of 1.3 days. In total, 24 patients were imaged before biopsy or surgery. There were 24 breast cancers and nine benign diagnoses. Two breast radiologists reviewed every paired three-dimensional dataset with regard to lesion visibility, reproducibility of documented location (clockface location, distance from nipple, and lesion depth), size of the lesions, and similarity for lesion characteristics. Lesion similarity was classified as being identical, similar, or different by consensus reading using the SomoVu workstation. Intraclass correlation coefficients (ICCs) and the Bland-Altman method were used to determine the amount of agreement between assessments of lesion location and size. Among 33 breast lesions, 31 lesions were depicted in both serial examinations. ICCs for the displayed lesion location (clock face location, distance from nipple), and the individual size of detected lesions were 0.994, 0.926, and 0.980, indicating excellent reliability. However, the ICC for lesion depth from the skin was 0.342 showing low reliability. For lesion similarity, 16 cancers and five benign lesions were classified as being identical, and six cancers and two benign lesions were classified as being similar. Two benign lesions were assessed to have different lesion characteristics and final assessment categories. The ABUS provided reproducible images for mass localization, size measurement, and characterization, which may be useful for follow-up studies.