Breast Ultrasound

What it does

Breast Ultrasound reads handheld breast studies and automated breast ultrasound (ABUS) volumes: radial and antiradial stills or cine of each lesion, images of the axilla, or the two to six 3D volumes an ABUS system acquires. Body-marker and clock-face annotations on the screen are read so that each lesion gets its location.

The result is a lesion list. Every lesion is detected, outlined and measured, tracked across the images and video frames that show it, placed by clock-face position, distance from the nipple and depth, and described with the ACR BI-RADS ultrasound lexicon. A final BI-RADS assessment with its management follows, then the axillary nodes.

Intended use

Decision support for the breast imager working up a palpable lump, a mammographic finding or a dense-breast screening scan. The same lesion seen on many images is reported once, and every assessment shows the descriptors it rests on.

Who it is for

Breast radiologists, breast surgeons, gynaecologists and general radiologists who read breast ultrasound, and sonographers preparing worksheets. Developers use the run to pre-fill BI-RADS reporting fields or to pre-read ABUS volumes.

Accepted input

• Handheld studies: DICOM Ultrasound Image and Multi-frame objects, radial and antiradial planes per lesion, with or without colour Doppler.
• ABUS: Enhanced US Volume objects, two to six volumes per study.
• Whole studies: a study folder or CD; the axillary images are separated from the breast images automatically.

Requirements

Requirement	Why
Two orthogonal planes per lesion	Shape, orientation and three-axis size need both
Readable body marker or clock-face annotation	Location (side, clock-face, distance from nipple) comes from it
SequenceOfUltrasoundRegions calibration	Lesion size and nodal cortical thickness in millimetres
Axillary images for the node section	The axilla is read only when imaged

Optional context

Age, the indication (palpable lump, mammographic finding, screening of dense breasts, follow-up) and prior findings go in clinical_context. Correlation with a mammogram is a separate run of a mammography model.

The result

One JSON result in sections with findings, measurements and classifications; lesion masks are returned as DICOM SEG. Each lesion keeps a lesion_id across images, with side, clock-face position, distance from the nipple and depth.

Section	Content
Lesions	Detection, outline and size of each lesion, tracked through cine and across images
BI-RADS assessment	Shape, orientation, margin, echo pattern, posterior features, calcifications and vascularity; BI-RADS 2–5 per lesion with management
Axillary lymph nodes	Node detection, cortical thickness, fatty hilum and a metastasis probability
ABUS lesions	Lesion detection and segmentation in 3D ABUS volumes with a benign or malignant estimate
Draft report	Opt-in (options.report): an English narrative that interprets the result for the reader (findings, impression, limitations); every number is checked against the findings

Standards

• Lexicon and assessment: ACR BI-RADS v2025 ultrasound lexicon, including non-mass lesions; labels from the 2013 5th edition are mapped when legacy terms are needed.
• DICOM: SR Breast Imaging Report TID 4200; SEG for lesion masks; GSPS calipers on the source frames; key-object selection for the images that define each lesion; Encapsulated PDF.
• Terminology: RadElement common data elements for BI-RADS, SNOMED CT for findings and anatomy.

Assessment notes

Each assessment category comes with its BI-RADS management: category 2 (benign) returns to routine care, category 3 (probably benign) leads to short-interval follow-up, and categories 4 and 5 lead to tissue diagnosis. Non-mass lesions are described with the terms that v2025 adds to the ultrasound lexicon.

Related models

• us/gynecology reads adnexal masses with O-RADS.