How does accuracy compare in automated breast volume ultrasound and hand-held ultrasound?
Automated volume breast ultrasound (AVBS) is equal in diagnostic accuracy to hand-held (HH) ultrasound, whether performed by a sonographer or mammography technologist, according to a study presented at RSNA 2016.
Researchers from the United States and Italy sought to compare the diagnostic accuracy and inter-observer variability of HH US and a single volume using AVBS centered over the clinical abnormality and to compare if there is a significant difference if the AVBS is performed by a sonographer (UT) or mammography technologist (MT).
A total of 90 patients, median age 53.1, participated in the trial, undergoing diagnostic ultrasounds; 60 for a palpable mass, 25 for mammogram abnormality, one for follow-up study, and four for breast discharge. The patients were randomized to have either a HH or AVBS first. HH was performed using a 14MHz transducer. The AVBS was performed using a L15-9 transducer. The technologist performing the second study was blinded to results of the first exam. The AVBS was randomized between a UT and a MT. The studies were blinded, randomized, and read by two radiologists each with greater than 10 years’ experience in breast ultrasound. The lesion with the highest BI-RADS score was used in the analysis. The HH studies were read six months before the AVBS studies. Final diagnoses where made by core biopsy for follow-up for two years. Lesions included nine malignant lesions and 81 benign lesions.
The results showed that the K for benign/malignant was 0.831 (while the global agreement using a 7-point BI-RADS score was 0.488). The K agreement between AVBS and HH in detecting breast pathology was 0.831.
The first rater had a K of 0.910 (0.787-1.000) while the second 0.760 (0.578-0.943). The agreement between AVBS and HH was nearly the same when AVBS was performed by a MT (K=0.858 (0.723-0.963) or UT (k=0.803(0.596-1.000), p=0.47. The AUC for lesion characterization was AVBS reader 1 0.91 (0.84-0.96), AVBS reader 2 0.91 (0.83-0.96), HH reader 1 0.91 (0.84-0.96), and HH reader 2 0.83 (0.74-0.90) with no statistical difference. The inter-observer agreement based on BI-RADS was 0.568 (0.468-0.647), with the HH K of 0.631(0.584-0.665) and for AVBS 0.492 (0.457-0.564). The agreement based on pathology was K=0.831 (0.718-0.944) with HH K=0.795 (0.623-0.967) and AVBS 0.869 (0.725-1.000).
The researchers concluded that one view diagnostic AVBS was equivalent to a HH in diagnostic US work-up, and there was no difference if the AVBS is performed by a trained UT or MT.
AI-Initiated Recalls After Screening Mammography Demonstrate Higher PPV for Breast Cancer
March 18th 2025While recalls initiated by one of two reviewing radiologists after screening mammography were nearly 10 percent higher than recalls initiated by an AI software, the AI-initiated recalls had an 85 percent higher positive predictive value for breast cancer, according to a new study.