1. Match the reading role and inputs

MammoScreen 5 K260714 describes a concurrent reading and reporting aid for screening mammograms using compatible full-field digital mammography and tomosynthesis systems, with compatible priors where available. Its analysis alone must not determine patient management. This is one exact-record example, not a description of every mammography AI tool.

  • Is the proposed use concurrent interpretation, another reading role, or a worklist change? Match each requested role to current labeling.
  • Which acquisition systems, images and prior examinations are supported by the proposed software version? Identify unsupported inputs explicitly.
  • What will readers see when priors are absent or input requirements are not met?

Source context: FDA K260714: MammoScreen 5

2. Test the complete display path

Review both ends of the intended interface. DICOM documentation can guide the integration discussion, but broad conformance does not prove the requested display behavior on Merge, Visage, Sectra, AGFA or IntelePACS at a particular site.

  • Which exact PACS or viewer release, AI version, output objects, licenses and hanging protocols will be tested together?
  • Can readers identify the AI result's patient, examination and software version and return to the underlying images without ambiguity?
  • How will delayed results, corrected studies, prior matching, downtime and rollback appear to the reader? Assign an owner to each test.

Source context: DICOM PS3.2: Scope and Field of Application

3. Keep scope, evidence and the offer separate

The Transpara 2.1.0 K241831 record provides another exact submission to inspect alongside MammoScreen. A comparison is an evidence collection task; it does not establish equal inputs, equivalent results or a preferred supplier. Written offer scope and local acceptance remain separate checks.

  • Do the studies describe the same reading role, examination type, population and endpoint that your team plans to evaluate?
  • Who will review discordant results and assess workflow effects before deciding whether the proposed use is acceptable?
  • Does the quote identify all required interfaces, workstations, modules, support and upgrade obligations? Take unresolved items to the committee packet.

Source context: FDA K241831: Transpara 2.1.0 · FDA list of artificial intelligence enabled medical devices

4. Work a fictional evaluation

A fictional screening service asks whether an AI reading aid could fit its existing mammography reading workflow. Its scanner, viewer and installed versions remain unspecified. Synthetic buyer inputs only. No product winner, clinical approval, offer or local acceptance is implied. Use the worked links below.

  1. Define the task and reading role before selecting records. Use K260714 and K241831 as separate source-bound examples, not interchangeable offers or a product ranking.
  2. Inspect each exact FDA identity, label constraints and evidence-context card. Keep supported inputs, reader requirements and tested study populations with their own record.
  3. Compare scope before performance. FDA decision materials do not establish a current offering. CE context does not expand US authorization, and a larger paper count is not stronger applicable evidence.
  4. Follow the integration path and leave installed compatibility open. Ask about acquisition, routing, processing, output, viewer/display and result review. A distribution platform is not a tested local pairing.
  5. Build the private committee draft. Carry both exact IDs into comparison/checklist and packet, add unresolved clinical and IT questions to the private action register, and leave sample/threshold choices blank until local reviewers decide.

Source context: FDA K260714: MammoScreen 5 · ACR ARCH-AI: governance and implementation context · ACR Assess-AI: clinical performance-monitoring context

Take the next step

Inspect the exact example records

Examples illustrate the questions above. They are not a ranking, equivalent alternatives, confirmed offers or the complete market.

Documented family navigation: MammoScreen reading aid · Transpara reading aid. Confirm each record's own scope.

Explore a private interpretation-capacity and cost scenario · Read a fictional committee packet · Check system compatibility scope

The worksheets and scenario use your assumptions. Save or attach private inputs explicitly; public evidence does not turn those inputs into verified results. The public sample uses fictional facility values.

Sources and scope

Sources anchor the stated regulatory or interoperability context. The questions are original editorial prompts; a citation does not validate a financial assumption or establish local fit. Source dates below apply to this guide, not the whole directory.

  1. FDA K260714: MammoScreen 5

    PDF pages 7-8: indications and predicate comparison. Checked .

    Exact K260714 reading role and compatible-input context; no installed pairing is established.

  2. FDA K241831: Transpara 2.1.0

    Decision letter: device name and submission identity. Checked .

    Exact-record identity for an example comparison; no comparative performance claim.

  3. DICOM PS3.2: Scope and Field of Application

    Chapter 1; page identifies edition 2026d at review. Checked .

    Conformance statement context; no viewer-specific compatibility conclusion.

  4. FDA list of artificial intelligence enabled medical devices

    About this list and public decision summaries. Checked .

    Regulatory record context only, not purchasing or local validation evidence.

  5. ACR ARCH-AI: governance and implementation context

    Requirements At a Glance: governance team, inventory, testing and workflow integration. Checked .

    General governance resource only. Linking does not certify this directory, a device or a local site; no affiliation or endorsement.

  6. ACR Assess-AI: clinical performance-monitoring context

    Real World Data and monitoring: model inputs, versioning and concordance. Checked .

    General post-deployment monitoring resource only. No participation, coverage for an example device, certification, affiliation or endorsement is claimed.

Dataset and maintenance dates · Methods · Submit a source-backed correction