A Multi-Center Study of ABUS for the Diagnosis of Breast Cancer in China

Abstract

Breast cancer is the most common cancer in women in China and worldwide. Mammography (MAM) is the first choice of breast screening. However, its practicability is limited in Asian due to dense breast mostly among Chinese women and the younger age of onset. Meanwhile, hand held ultrasound (HHUS) has obvious operator dependent and poor repeatability. Automated Breast Ultrasound System (ABUS) is a potential method to alleviate current challenges in accessible breast cancer screening. This study aims to evaluate the initial effectiveness of ABUS by comparing it with Hand Held Ultrasound (HHUS) and Mammography (MAM) in a hospital-based multi-center study. Women between the ages of 30 to 69 who visited breast surgeons for the first time without visible, suspicious signs of breast cancer were eligible for our study. All participants underwent HHUS and ABUS, and women in the older group (40 to 69 years old) also received MAM. Using the Breast Imaging Reporting and Data System (BI-RADS) lexicon, the images of the lesions were interpreted independently. Image interpretations were done without knowledge of clinical or other imaging results. Lesions classified as BI-RADS 4 or 5 were considered to be “suspicious”, and BI-RADS 1 to 3 lesions were assessed as “benign”. So far, 1010 eligible women have been enrolled and 656 in the older group. Taking breast as the unit of analysis, we have acquired 2020 HHUS results, 2020 ABUS results and 1312MAM results. The consistency rates and Kappa statistics were calculated to assess the reliability of ABUS compared with HHUS or MAM. The average subject age was 45.32 (SD 9.756) in the whole group and 50.82 (SD 7.489) in the older group. Of all the 2020 breasts, HHUS detected 395 suspicious lesions and ABUS detected 356 suspicious lesions. Among the 395 suspicious lesions detected by HHUS, ABUS detected 327; In the older group, ABUS detected 275 suspicious lesions and MAM detected 259 suspicious lesions. Among the 259 suspicious lesions detected by MAM, only 41 lesions were undetected by ABUS. Among the 275 lesions detected by ABUS, only 57 were not detected by MAM. The reliability of ABUS and the other two modalities were quite good. Specifically, the consistency rate between HHUS and ABUS was 95.2%, and that between ABUS and MAM was 92.53%. The Kappa value between ABUS and HHUS was 0.8414 and that of ABUS and MAM in the older group was 0.7696. Fairly good reliability was observed in comparisons between ABUS and HHUS or MAM in our initial analysis. As ABUS is operator independent and it is feasible for all the radiologists with a short term of training to have similar lesion interpretation as HHUS or MAM by specialists. As a result, ABUS is a promising modality in breast imaging.