Automated breast ultrasound as supplemental imaging for patients with dense breasts

For many patients, the path to early detection can be fraught with uncertainty and quiet fear, more complex than it might first appear. Breast cancer has been chronicled across the ages, from the papyri of ancient Egypt to the genomic lexicon of modern medicine, and this long arc of understanding naturally extends into the evolution of imaging. The discovery of X-rays set the stage for mammography, while the rise of ultrasound offered a complementary lens, one that probes breast tissue without ionizing radiation.

Today, mammograms remain the cornerstone of breast cancer screening, but their effectiveness can be limited in women with dense breast tissue¹, where abnormalities are harder to distinguish. Both dense tissue and tumors show up as white on the scan, which means cancers can sometimes be hidden. In extremely dense breasts, up to half of cancers might be missed.² Dense tissue is also common, affecting about 40 percent of women³ and up to 70 percent of Asian women⁴, and carries a higher risk of developing breast cancer.⁵

Addressing this gap calls for scalable screening that can better detect differences in tissue, deliver consistent results and fit the realities of a strained radiology workforce.

The promise of ultrasound

Ultrasound addresses the challenge of dense breast tissue using a different set of physical principles. While mammography relies on X-ray attenuation, ultrasound uses sound waves rather than ionizing radiation and may be used as a supplemental imaging approach when consistent with labeling, clinical guidelines, and local practice. Some lesions that are less conspicuous on mammography in dense tissue may be more conspicuous on ultrasound, depending on lesion characteristics, image quality, and clinical context. Traditional ultrasound can increase sensitivity in dense breast tissue, and pairing it with mammography can reduce interval cancers, those diagnosed between routine screenings.⁶

In clinical practice, supplemental imaging is often performed as handheld breast ultrasound, in which a physician or sonographer manually sweeps a transducer across the breast to acquire images in real time. While this approach has long held clinical value, its reliance on operator technique introduces variability. Image quality, coverage, lesion documentation and reproducibility can differ depending on the physician or sonographer’s approach.

Automated Breast Ultrasound, or ABUS, was developed to reduce this variability through a standardized scanning process. A mechanized scan head guides the transducer across the breast in a predefined path, capturing more than 300 two-dimensional images during transverse sweeps from skin to chest wall. These images are reconstructed into a three-dimensional breast volume, a stack of slices that can be examined in multiple planes.

ABUS also separates image acquisition from interpretation. Technologists can perform the scans while radiologists review the completed volumes later, using preset viewing layouts designed for three-dimensional image sets. This separation of acquisition and interpretation may support workflow flexibility, depending on site staffing, protocol, and local requirements.

In the interim BRAID randomized controlled trial, ABUS was one of several supplemental imaging arms studied in women aged 50–70 with dense breasts and a negative mammogram; the ABUS arm reported 4.2 cancers per 1,000 examinations and a 4.0% recall rate.⁷ Women enrolled in the trial generally found supplemental screening acceptable and well tolerated —  especially with ABUS compared to contrast enhanced techniques, such as CEM and MRI — an important consideration for programs that depend on patients returning over time.⁸

The coronal plane

An important feature of automated breast ultrasound is the coronal plane, a front-to-back view reconstructed from the three-dimensional ultrasound volume. In this orientation, radiologists can observe architectural patterns across a wide field. Spiculations, the radiating lines that may signal malignancy, can appear as a star-like formation.

The stored volume also allows for virtual rescanning in multiple planes, precise lesion localization using clock-face references and distance from the nipple, and comparison with prior studies. A radiologist can return to the dataset after the examination and review the same area from different angles without asking the technologist to reproduce a manual sweep.

The Invenia^TM ABUS Premium system builds on these capabilities with refinements to image quality, acquisition and interpretation. A redesigned probe with a patient-friendly reverse curve design enhances resolution and reduces artifacts⁹, while updated algorithms address nipple shadowing, scan quality and anatomical documentation.

Probe design and image quality

A central change is the Room Temperature Vulcanizing Silicone lens within the imaging probe.

The silicone lens allows more precise control over the thickness of the ultrasound beam. A thinner beam reduces overlapping tissue information within each reconstructed slice, improving spatial and contrast resolution. A wider acceptance angle improves visualization around curved anatomy and beneath the nipple.¹⁰

White-stripe and cat-scratch artifacts are bright bands or fine, scratch-like lines created by the imaging process rather than the breast itself. Because artifacts may affect image review, artifact-reduction features are intended to support image interpretation. The redesigned probe is intended to reduce nipple-region shadowing; include approved technical or clinical support for the extent and conditions of this effect.

Nipple detection and shadow compensation

Nipple documentation provides a consistent anatomical reference. Breast findings are commonly described by their clock-face position and distance from the nipple, allowing radiologists to locate lesions and compare them across examinations and imaging modalities.

Enabled by Verisound^TM AI, Auto Nipple Detection uses image segmentation to identify that landmark. A U-Net neural network examines multiple coronal projections near the skin, combines the results and calculates the nipple’s location. The proposed marker position can be reviewed and adjusted by the user, consistent with the system workflow and labeling. The final coordinate is stored in the imaging file for later interpretation and comparison.

Nipple Shadow Compensation addresses the separate problem of tissue obscured beneath the nipple. Differences in acoustic properties between the nipple and surrounding tissue can produce strong reflections and darken the anatomy below.

The compensation algorithm uses a wavelet transform to separate the image into layers of detail, allowing localized adjustments rather than uniform brightening. Within a defined region beneath the nipple, the system evaluates measures such as mean intensity and variance, then applies a controlled enhancement that diminishes outward from the center. The aim is to recover signal while limiting noise amplification.

Speed and performance

Invenia^TM ABUS’ Fast Scan feature increases acquisition speed by 40 percent (compared to the prior Invenia^TM ABUS 2.0) with minimal difference in image quality.¹¹ Auto Nipple Detection was tested on 439 breast volumes from institutions separate from those used to develop the system. It placed the nipple marker more consistently than the variation typically seen between human readers.¹²

Scan Quality Assessment was evaluated on another 427 volumes. Across probe positioning, imaging depth and tissue coverage, sensitivity ranged from 96.0 to 98.7 percent and specificity from 97.4 to 100 percent.¹³

The open architecture of  Invenia^TM ABUS enabled integration with a variety of tools such as QVCAD^TM14 used during interpretation. These  have reported lesion-detection sensitivity of up to 93 percent and reading-time reductions of about one-third.¹⁵ Some helped less experienced readers approach the performance of senior specialists. In one triage study, AI (MONCAD ABS DX) identified 84 percent of negative screening cases that did not require full radiologist review.¹⁶

For patients with dense breast tissue, confidence in screening depends on the completeness of the examination and the clarity of the images. A missed area may mean another appointment. Tissue obscured beneath the nipple may leave an important question unresolved. An inconsistently documented finding can make comparison more difficult the next time a patient returns.

ABUS is designed to reduce those sources of uncertainty. A technologist can rescan an incomplete area before the patient leaves. A radiologist can revisit a stored three-dimensional volume and examine a finding from several angles. Automated landmarks and quality checks can make one examination easier to compare with the next.

The fear surrounding breast screening cannot be engineered away. Careful acquisition and clearer imaging can keep that fear from being compounded by an incomplete view.

---

1. Kolb et al, Radiology, Oct 2002;225(1):165-75 ↩︎
2. Kolb et al, Radiology, Oct 2002;225(1):165-75. ↩︎
3. Pisano et al. NEJM 2005; 353: 1773. ↩︎
4. Ref. Ellison-Loschman, et al, PLOS ONE July 2013. ↩︎
5. Kolb et al, Radiology, Oct 2002;225(1):165-75. ↩︎
6. https://www.itnonline.com/article/dense-breast-tissue-supplemental-imaging ↩︎
7. Gilbert et.al., Comparison of supplemental breast cancer imaging techniques—interim results from the BRAID randomised controlled trial; The Lancet; May 2025; https://doi.org/10.1016/ S0140-6736(25)00582-3 ↩︎
8. Allajbeu et.al., Acceptance, experience, and feedback for supplemental screening in dense breasts among women participating in the BRAID trial. Insights into Imaging, (2026) 17:14 https://doi.org/10.1186/s13244-025-02170-8 ↩︎
9. GE HealthCare ABUS Technical Whitepaper 2026 (JB38391XX) ↩︎
10. GE HealthCare ABUS Technical Whitepaper 2026 (JB38391XX) ↩︎
11. GE HealthCare ABUS Technical Whitepaper 2026 (JB38391XX) ↩︎
12. Appropriate clinical data collection and/or research agreements have been duly established.. Results are based on internal validation study and customer feedback from clinical evaluations. ↩︎
13. Appropriate clinical data collection and/or research agreements have been duly established.. Results are based on internal validation study and customer feedback from clinical evaluations.
    ↩︎
14. QVCAD is a trademark of QView Medical, Inc. ↩︎
15. Performance and Reading Time of Automated Breast US with or without Computer-aided Detection. Read More: https://pubs.rsna.org/doi/10.1148/radiol.2019181816 ↩︎
16. MONCAD ABS DX is a product from Monitor Corporation ↩︎