BEST IN PHYSICS (IMAGING): Development of Supine MRI-Localized, 3D Ultrasound-Guided Breast Biopsy Using a Wearable 3D Abus Device
Abstract
Purpose
Ultrasound (US)-guided breast biopsy is widely used due to its accessibility and real-time visualization, but operator dependence can contribute to high false-negative sampling. Magnetic resonance imaging (MRI)-guided biopsy offers high sensitivity but is complex, costly, and limited by the need for dedicated prone biopsy tables. We developed and validated a wearable 3D Automated Breast Ultrasound (ABUS) biopsy device that enables supine MRI-localized, 3D US-guided targeting to improve procedure workflow and accessibility.
Methods
A cost-effective 3D ABUS device was redesigned to accommodate both core-needle and vacuum-assisted needle biopsy procedures in supine position. The device consists of a wearable base that conforms to the chest wall, an adjustable compression mechanism in mediolateral and anteroposterior directions, and a motorized linear scanner for volumetric imaging. Two clinical workflows were implemented: 1) 3D ABUS localization with real-time US-guided biopsy for US-visible lesions, and 2) MRI-to-3D ABUS localization via automated rigid registration between MRI and ABUS volumes for MRI-only visible lesions. Registration accuracy was evaluated using breast-mimicking phantoms with embedded inclusions, quantified by fiducial localization error (FLE) and target registration error (TRE). Needle guidance accuracy was assessed through mock biopsies using needle targeting error and trajectory error.
Results
Automatic MRI-3D ABUS registration achieved a mean TRE of 2.57 mm with a mean FLE of 0.51 mm. Mock biopsies demonstrated a mean needle targeting error of 1.43 ± 0.56 mm and angular trajectory error of 1.20 ± 0.72º, corresponding to a minimum reliably sampled lesion diameter of 2.06 mm at 95% confidence.
Conclusion
The proposed 3D ABUS biopsy device enables accurate supine-based MRI-localized and US-guided breast biopsy with high targeting accuracy. This approach has potential to decrease procedural complexity and accessibility to MRI-informed sampling, particularly for lesions that are challenging to localize with conventional US.