Clinically and Mammographically Occult Breast Cancers Initially Detected by Hand-Held Ultrasound (HHUS): Detection Performance of Radiologists Using 3D Breast US Data Obtained by a Commercially Available Automated Breast US System (ABUS)

Abstract

PURPOSE To retrospectively assess the performance of radiologists for the detection of breast cancers initially detected by HHUS using 3D breast volume data obtained by ABUS.

METHOD AND MATERIALS From 2007.11-2008.02, bilateral whole breast US was obtained using ABUS (U-systems) in 61 consecutive women who were scheduled to undergo US guided needle biopsy due to suspicious breast masses detected at supplemental screening HHUS. Finally 14 cancers (10 invasive and 4 intraductal cancers) in 13 women (aged 35-50 years) were found and mean lesion size of IDC was 0.56cm (range, 0.2-1.5cm). For the control group, 48 normal breasts were selected from 48 women with benign breast diseases in the contralateral breasts. Two breast radiologists assessed lesion size and conspicuity of 14 cancers at 3D breast volume data by five grade scales in consensus with the information of HHUS. Another three breast radiologists (experience range 5-11 years), who did not perform the examinations and blinded to histology, independently reviewed 3D breast volume data of 61 (13 diseased and 48 normal) breasts to detect any suspicious lesions and documented the lesion features and BI-RADS final assessment. Sensitivities, and false positive rates were calculated, and results of agreement on lesion features and final assessments were evaluated by using κ statistics.

RESULTS Sensitivities of each reader for cancer detection were 78.6% (11/14), 78.6% and 57.1% (8/14), respectively, with false positive rates of 20.8% (10/48), 12.5% (6/48) and 8.3% (4/48). Seven cancers (mean conspicuity score 4; mean US size, 1.7cm) were detected by all three readers, four (3.5; 1.7 cm) by two readers, and one (3; 2cm) by only one reader. Two invasive cancers (1; 0.6cm) were not detected by all three reviewers. Regarding the descriptions of US findings, all features including lesion shape and margin were discordant between the radiologists, and for final assessment, ĸ values were <0.2.

CONCLUSION Only 57.1-78.6% of HHUS detected cancers were detected by the radiologists when they read the 3D breast volume data obtained by a commercially available ABUS. Also, slight agreement was found for description of lesion features, and final assessment.

CLINICAL RELEVANCE/APPLICATION For interpretation of the 3D breast volume data obtained by current ABUS, substantial degree of radiologists` experience and training seems to be necessary for detection of small and subtle cancers.