Aferente

Managed clinical AI for radiology

Tell us what clinical problems you want solved. We handle everything else.

Aferente is a managed clinical AI service, not a marketplace. One integration with your PACS, one contract and one point of contact. Results arrive in the viewer your radiologists already use.

Book a demo Try the interactive demo (synthetic data)

What we commit to

  • 5 business days

    from the VM being ready to going live, with the PACS connected. That is our target for every new hospital.

  • 0 IT tickets

    for a new capability to reach the hospital. No new on-site software, no configuration.

  • 0 new screens

    for radiologists. The result is one more series in the same study, in the usual viewer.

  • 0 identifiers

    of patients leave the hospital. De-identification happens inside your network.

The problem

Today, every AI vendor is a new IT project.

Imaging AI is already useful. Putting it to work is the hard part. Every product brings its own path to production:

  • Another install

    A server or VM per vendor, with its own operating system, its own updates and someone in IT to look after it.

  • More firewall rules

    New ports, VPNs and exceptions for every product, each one reviewed by information security.

  • Another contract

    Another legal review, another data processing agreement (LGPD), another security questionnaire.

  • Another screen

    A new login and a new viewer for radiologists, outside the reporting workflow they already use.

The outcome is predictable: the hospital adopts one or two products and stops. Aferente replaces all of that with a single integration.

How it works

Three steps, done once.

  1. One VM

    The hospital provides a virtual machine. We install and operate the Aferente gateway on it. In the PACS, the administrator adds one DICOM destination and one forwarding rule.

  2. One outbound connection

    The gateway de-identifies each study inside the hospital and uses a single outbound HTTPS connection to the Aferente cloud, which takes the study to the vendor fulfilling the approved capability.

  3. Results back in your PACS

    The result returns to the gateway, is re-identified inside the hospital and stored as a new series in the same study. Radiologists see it in their usual viewer and report as they do today.

PACSHospital gatewayAferente cloudVendor

If anything fails (vendor down, network, PACS busy), the original study is untouched and the radiologist reports as usual. Nothing waits on AI.

Follow one study end to end in the interactive demo

Capabilities

You choose the clinical problem. We take care of the vendor.

Each capability is fulfilled by one or more vendors that we select, validate and operate. The hospital approves the capability once and sees which product fulfils it.

Examples. These are examples of what can be offered. They are not approved products and are not available today. Each capability depends on the vendor's ANVISA registration for the intended use, clinical validation and a contract.
  • ExampleX-ray

    Chest X-ray findings

    Flags findings such as pneumothorax, pleural effusion and consolidation, to help prioritise the worklist.

  • ExampleCT

    Intracranial haemorrhage

    Flags suspected haemorrhage on non-contrast head CT for priority reading.

  • ExampleCT

    Lung nodules

    Detects and measures nodules on chest CT, with comparison to the prior study.

  • ExampleX-ray

    Extremity fractures

    Marks possible fractures on extremity X-rays in the emergency department.

  • ExampleCT

    Organ segmentation

    Organ outlines and volumes on CT, delivered as a DICOM series for review.

  • ExampleMG

    Breast density

    Classifies breast density on mammography to support the report.

Privacy

Patient identity never leaves the hospital.

  • De-identification at the edge

    Before anything is sent, the gateway builds a copy with only the pixels and the technical fields needed for analysis. Name, record number, dates, institution, physicians and free text are left out.

  • Re-identification only on site

    Identifiers are replaced with random codes known only to the gateway, in memory, for that one study. Neither the Aferente cloud nor the vendors know who the patient is.

  • LGPD from the contract up

    The hospital is the controller and Aferente is the processor. We give the hospital's DPO (encarregado) and legal team the processor agreement, the record of processing, input for the data protection impact report (RIPD) and the sub-processor list.

  • Everything auditable

    Every transfer, result, approval and change goes into a chained audit trail that the hospital can verify and export.

For hospital IT

Built to pass the security review.

  • One outbound HTTPS connection

    A single outbound connection on port 443. No ports open to the internet, no VPN, no inbound access.

  • Signed updates

    Every gateway release is signed and verified before it is applied, and rolls back automatically to the previous version if anything fails.

  • Four-eyes approval

    A capability is switched on only when two hospital administrators approve it, once, with a declared scope: modalities, body parts, intended use and vendors.

  • Vendor changes under control

    For each capability the hospital chooses: be notified, with a window to object, or approve before any vendor is swapped or added.

  • Standard DICOM

    The PACS sends over DICOM to one more destination and receives results as new series. Nothing changes in the PACS, the RIS or the reporting workflow.

  • No patient data kept

    The gateway keeps no copy of studies. Configuration, usage and the audit trail are available to IT in the hospital console.

Book a demo

Tell us what you want solved.

We will reply with a demo shaped around your institution and the next steps to bring the first capability live.

Prefer to look around first? Open the interactive demo with synthetic data, in your browser.

For example: prioritising chest X-rays in the emergency department.

We use these details only to reply to you. No cookies or tracking on this site.