Longitudinal Imaging Performance Assessment of Grayscale Breast Ultrasound Using the Random Hypoechoic Sphere Phantom
Abstract
Purpose
Grayscale ultrasound is widely used in breast screening, diagnosis, and ultrasound-guided biopsy, particularly for detecting breast lesions. However, current Quality Control (QC) methods provide limited and often subjective assessment of lesion detectability in breast ultrasound systems. To address this limitation, we developed an automated analysis method that uses freehand scanning of a random hypoechoic sphere phantom (RHSP) to quantify lesion signal-to-noise ratio (LSNR) in breast ultrasound systems. Using this approach, we monitored longitudinal drifts in imaging performance for two breast ultrasound scanners over an eight-week period.
Methods
A RHSP with 2 mm spheres was scanned on two Philips EPIQ ELITE ultrasound scanners—one biopsy-guidance unit (Scanner 1) and one diagnostic unit (Scanner 2)—each operated under two protocols, yielding four acquisition configurations. For each configuration, data were collected weekly across nine time points over eight weeks. At each time point, a single operator performed a uniform 3‑second freehand cine sweep, repeated six times. The analysis algorithm generated LSNR‑versus‑depth curves averaged across repeats. Longitudinal drifts were assessed using linear regression, and variability was analyzed for both peak LSNR at 1cm depth and LSNR averaged over all depths.
Results
Peak LSNR regression slopes ranged from -0.114 to 0.045 per week, with r2 values between 0.03 and 0.42, indicating no strong longitudinal change. Overall LSNR slopes showed a similar trend. Peak LSNR variability (1.6% to 5.0%) and overall LSNR variability (2.9% to 3.3%) demonstrated stable scanner performance and minimal phantom aging. Scanner 1 showed slightly worse LSNR than scanner 2 in Adv. Breast protocol (-6.75 versus -7.08), consistent with clinical impressions that the biopsy-guidance unit has lower image quality than the diagnostic one.
Conclusion
No drift on LSNR was observed during the eight-week study. RHSP and the automated analysis algorithm can be routinely used to maintain image quality of breast ultrasound scanners.