Patient Positioning AI

Cut retakes on CT, MRI, X-ray, and mammography, and add markerless SGRT on every linac.

Try it now for free Talk to an AI Expert

Patient Positioning AI Across Radiology, Radiation Therapy, and Every Modality

Deploy Anywhere, Run Everywhere
Run patient positioning on CT and MRI gantry cameras, linac SGRT depth cameras, X-ray and mammography suite cameras, PET/SPECT, and interventional radiology, on the edge, on-prem, in a hospital-controlled VPC, or air-gapped.
One Platform, Full Adoption
Tools every hospital team can adopt, from radiologic technologists, mammographers, and radiation therapists to medical physicists, no separate ML team required to ship and own positioning models.
HIPAA-Ready, Audit-Ready, Reimbursement-Ready
Data stays safe with HIPAA compliance, SOC 2 Type II, encrypted data, hospital-controlled deployment, and an uptime SLA on every modality.

“We wanted to develop models using data we owned, have the ability to scale usage, leverage state-of-the-art AI capabilities, and along the way ensure the privacy of our users’ personal health information. That’s what we found in Roboflow.”

Alec Zopf

Co-founder and Chief AI & Automation Officer, Wellth

Read the full story See customer stories

Talk to a Vision AI engineer who's shipped patient positioning.

A single mispositioned patient on a CT chest scan, a missed ISO center at treatment planning, or a repeat mammography compression can mean a retake that doubles patient radiation dose, a linear accelerator fraction that misses the tumor and hits organs at risk, or an HCAHPS hit that costs the hospital reimbursement points. Bring us your toughest patient positioning problem and we'll map a working solution.

Ask us about:

Solution architecture for HIPAA, FDA 510(k), IEC 60601-1, ISO 13485, DICOM, HL7 v2/FHIR, IHE Radiology profiles, ACR accreditation, NRC/IAEA, and Joint Commission survey documentation
Live demo on your CT gantry, MRI suite, linac SGRT camera, mammography, X-ray, PET/SPECT, or interventional footage with de-identified imagery
Deployment options: edge, on-prem, hospital-controlled VPC, or air-gapped, with integration into PACS, EMR/RIS, TPS, and oncology information systems through DICOM, HL7 v2, FHIR, and IHE profiles
ROI modeling against retake radiation dose, scanner-hour throughput loss, linac fraction accuracy, HCAHPS impact, staff musculoskeletal load, and payer medical-necessity documentation

Over 80,000 organizations build with Roboflow.

Rivian
Pella
Chobani
USG Corporation
BNSF Railway
American Woodmark

Start where you are

Bring Real-Time Intelligence to Every Patient on Every Scanner and Every Linac, with Vision AI

Add a real-time positioning layer to every patient on every modality with Vision AI for patient positioning. Roboflow extends positioning coverage to every scan and every treatment fraction, on the depth and RGB cameras your suites already run.

Radiology Positioning: CT, MRI, X-Ray, Mammography, PET/SPECT:

Detect patient landmarks (head, sternum, ISO center, anatomic markers) markerlessly with RGB, depth, and structured-light cameras, and align to the modality's planned protocol before the first exposure
Cut retakes on CT (5-15% historical), X-ray (15-25%), and mammography (5-10% recompression), reducing patient dose, staff musculoskeletal load, and scanner-hour throughput loss that hits HCAHPS
Coach patient motion in real time (breath-hold prompts, positional-drift alerts, comfort adjustments) with the technologist in the loop and the patient in view of the console

Radiation Therapy: Markerless SGRT, DIBH Gating, and ISO Alignment:

Add markerless surface-guided radiation therapy (SGRT) on any linac (Varian TrueBeam, Elekta Versa HD, Elekta Unity MR-linac, ViewRay MRIdian), with sub-millimeter surface tracking against the planned reference from the planning CT and TPS
Coach and gate deep-inspiration breath-hold (DIBH) for left breast SBRT and other cardiac-sparing treatments, with a beam-gating signal the linac controller can act on if the patient drifts out of tolerance
Reconcile positioning against the TPS (Varian Eclipse, Elekta Monaco, RaySearch RayStation, Brainlab Elements) and push per-fraction logs into the oncology information system (Varian ARIA, Elekta Mosaiq) via HL7 and IHE RO profiles

Patient ID, Workflow Integration, and Compliance:

Verify patient identity at modality setup with wristband OCR and opt-in facial-biometric confirmation (HIPAA-compliant, hospital-controlled), tying the physical patient to the order in the EMR/RIS (Epic Radiant, Cerner Millennium, Meditech Expanse) and preventing wrong-patient wrong-study events
Integrate with PACS (Sectra, Merge/Change Healthcare, Agfa, Fujifilm Synapse, Philips IntelliSpace, GE Centricity) via DICOM, EMR/RIS via HL7 v2 and FHIR, and radiation oncology systems via IHE RO profiles (RO-ILS, RO-SWF), with REST, MQTT, and OPC UA for hospital IT
Deploy on-prem, on the edge at each scanner or linac, in a hospital-controlled VPC, or air-gapped, with SOC 2 Type II, HIPAA, encrypted data, and audit-ready records for FDA 510(k), IEC 60601-1, IEC 62366, ISO 13485, ACR accreditation, NRC/IAEA, ASTRO, CMS documentation, and Joint Commission surveys

Bring intelligence to every patient today. Stop retakes, mispositions, and wrong-patient events from becoming radiation dose to the patient, linac fractions that miss the tumor, HCAHPS complaints, or HIPAA reportable events.

‍

More About Patient Positioning

What is patient positioning with Vision AI?

Patient positioning with Vision AI uses computer vision models (opens in a new tab) on RGB, depth, and time-of-flight cameras to detect patient landmarks (head, sternum, ISO center), track patient surface in real time against a planned reference from the planning CT and TPS, coach breath-hold and motion management, verify patient identity with wristband OCR, and reconcile positioning across CT, MRI, X-ray, mammography, PET/SPECT, interventional radiology, and radiation therapy on linear accelerators.

Can Vision AI hit sub-millimeter accuracy for markerless SGRT and DIBH gating?

Yes. Roboflow continuously reconstructs a patient-surface point cloud from depth and structured-light cameras above the couch, aligns it to the planned reference surface from the planning CT and TPS at every frame, and exposes a beam-gating signal the linac controller (Varian TrueBeam, Elekta Versa HD) can act on if the patient drifts out of the physicist-defined tolerance (typical 2-5 mm translation, 1-3 degree rotation).

Does patient positioning support HIPAA, FDA 510(k), IEC 60601-1, DICOM, HL7, and ACR accreditation?

Yes. Roboflow models can be trained against your specific HIPAA and HITECH protections, SOC 2 Type II, FDA 510(k) premarket notification, IEC 60601-1 and 60601-1-2 (medical electrical equipment safety and EMC), IEC 62366 (usability engineering), ISO 13485 (medical device QMS), ISO 14971 (risk management), DICOM (including DICOM RT and Structured Reporting), HL7 v2 and FHIR, IHE Radiology profiles (SWF, PIR, XDS-I, RO-ILS, RO-SWF), ACR accreditation for CT, MRI, mammography, and radiation oncology, NRC and IAEA for radiation therapy, ASTRO clinical practice guidelines, CMS reimbursement documentation, and Joint Commission survey requirements.

Can it integrate with our PACS, EMR, RIS, TPS, oncology information system, and hospital IT?

Yes. Roboflow Inference (opens in a new tab) exposes a standard API and supports common hospital and radiology protocols. Customers integrate with PACS (Sectra, Merge/Change Healthcare, Agfa, Fujifilm Synapse, Philips IntelliSpace, GE Centricity, Carestream Vue), EMR/RIS (Epic Radiant, Cerner Millennium, Meditech Expanse, Allscripts), TPS (Varian Eclipse, Elekta Monaco, RaySearch RayStation, Brainlab Elements, Accuray Precision), oncology information systems (Varian ARIA, Elekta Mosaiq), image-guided surgery navigation (Medtronic StealthStation, Brainlab Curve, Stryker Q Guidance), and interventional radiology systems (Philips Azurion, Siemens ARTIS, GE Innova) through DICOM (including DICOM RT), HL7 v2, FHIR, IHE Radiology profiles, REST, MQTT, and OPC UA.

Patient Positioning AI