Radiology reporting standard

BI-RADS

BI-RADS is the American College of Radiology system for breast imaging that assigns every mammogram, breast ultrasound or breast MRI a final assessment category from 0 to 6 tied to a management recommendation, plus a breast density category from a to d.

BI-RADS through RadVector API BI-RADS in RadVector Reporting

Validated packValidated criteria pack (2026.1, validated 2026-08-18); verdicts may be marked usable for output.

At a glance

What BI-RADS is

Full name: Breast Imaging Reporting and Data System
Edition RadVector applies: ACR BI-RADS Atlas, 5th edition
Published by: American College of Radiology (ACR)
Applies to: Mammography, breast ultrasound and breast MRI.
Modality: Mammography, breast ultrasound, breast MRI

Official source: ACR BI-RADS

BI-RADS categories

0: Incomplete: additional imaging or prior comparison needed.
1: Negative.
2: Benign.
3: Probably benign: short-interval follow-up.
4 (4A, 4B, 4C): Suspicious: tissue diagnosis recommended, with low, moderate or high suspicion.
5: Highly suggestive of malignancy: tissue diagnosis.
6: Known biopsy-proven malignancy.
Density a to d: Almost entirely fatty, scattered fibroglandular, heterogeneously dense, extremely dense.

A short summary in our own words. Use the official BI-RADS documents for the complete criteria.

How RadVector applies it

BI-RADS, applied the same way every time.

The same deterministic engine grades BI-RADS inside RadVector Reporting and through RadVector API, and a test keeps the two answers identical.

Validated criteria pack: a BI-RADS verdict can be marked usable for output.
Keeps screening and diagnostic rules apart: category 3 is not used at screening, and additional imaging with no stated category reads as category 0.
Checks that the category, management and density in a report agree, including an institution's coded BI-RADS line, and that a recommended workup was not already done today.
Carries the BI-RADS 3 follow-up schedule so a conditional recommendation is decided from the prior mammogram dates instead of copied.

Radiologists and imaging centers

RadVector Reporting supplies the BI-RADS category and recommendation to the report draft from the radiologist's own findings, cites the guidance applied, and keeps a per-report governance record. The radiologist signs.

radcloud.net

Software teams and AI companies

RadVector API returns the BI-RADS verdict with pack version, validation status and a trace id, as JSON or FHIR R4, with a sandbox key and synthetic fixtures to start.

Developer guide

Hospitals and enterprise imaging

One versioned BI-RADS implementation across reporting, QA and other AI systems' results, with an audit trail of which pack version graded each report.

Talk to us

BI-RADS questions

Is RadVector's BI-RADS pack validated?

Yes. BI-RADS (ACR 5th edition) is one of three validated RadVector criteria packs, alongside Fleischner 2017 and DEXA. A validated pack names its validator and validation date on every verdict, and only a validated pack with a confident decision may be marked usable for output.

Can software get a BI-RADS category from an API?

Yes. RadVector API evaluates breast findings against the BI-RADS pack and returns the category, management language and provenance, as JSON or as a FHIR R4 Bundle. It grades findings a radiologist or another system reported; it does not read images.

Why does BI-RADS 3 not appear on a screening mammogram?

Under BI-RADS, a finding that needs further evaluation at screening is called back as category 0; probably benign (category 3) is assigned after diagnostic workup. RadVector applies that rule and flags a screening report that assigns category 3.

Other reporting standards

RadVector is clinical decision support for qualified professionals, not a diagnostic device; the radiologist makes the final determination.