MediView XR

Artificial intelligence is claimed in the company's news and cannot be located in the product. The claim lives in MediView's announcement of its $24 million Series A, surfaced on its home page in September 2026, which said the company would transform image guided surgery through augmented reality and AI.

It does not live in the product: the current home page uses no AI, machine learning or algorithm language, and the 510(k) summary for XR90 (K223125) describes a commercial off the shelf augmented reality headset wirelessly connected to a streamer that interfaces with a GE Vivid iq ultrasound system and an electromagnetic field generator, receiving 3D volumetric data from CT, with no machine learning, AI or neural network mentioned.

What XR90 does is spatial registration and rendering, aligning a prior CT volume to live ultrasound and a tracked instrument and displaying the result in anatomical position; that is a hard engineering problem, and it is not a learned one on anything published. The 3D models may be built on GE HealthCare's Allia platform, which the home page quotes a user describing, but no learned step is named for MediView's product.

The constraint is written into the clearance. XR90 is cleared (K223125) for 3D visualization and interaction in conjunction with traditional imaging and monitors, for pre operative planning and heads up intra operative display during ultrasound guided needle procedures, so it supplements rather than replaces standard imaging, and the clinician keeps the instrument and the decision throughout. The system displays; it does not act, advise or move anything.

The safety question in navigation is accuracy rather than autonomy, and the summary reports registration error from phantom, cadaver and animal testing (means of 2.3 to 2.9 mm, upper bounds up to 3.4 mm). What is not published is what the system does when registration degrades, when the patient moves, or when ultrasound and CT disagree; a navigation display that is confidently wrong by a few millimeters matters more than one that declines to display.

The approach and the component suppliers are named in the clearance, with no model to disclose. K223125 describes what produces the display: a commercial off the shelf augmented reality head mounted display (Microsoft HoloLens 2), wirelessly connected to a streamer that interfaces with a GE Vivid iq ultrasound system and an electromagnetic field generator for instrument tracking, receiving CT volumetric data and registering it to live ultrasound; it reports target registration error from phantom, cadaver and animal testing.

The user facing components are CT derived holographic anatomy, live ultrasound projected in anatomical position, and a holographic guide along the instrument path. The company does not publish how the CT is turned into the 3D model (manual, semi automatic or learned, and whether GE HealthCare's Allia platform does it), the registration algorithm itself, or software versions and update practice.

The hardware chain is named, and there is no model provider to name. K223125 identifies the components in the chain: a commercial off the shelf headset (Microsoft HoloLens 2), a streamer, a GE Vivid iq ultrasound system and an electromagnetic field generator; no learned model is described, and no cloud host, subprocessor list or remote collaboration provider is named.

The chain question that matters here is data leaving the room: remote collaboration sends procedural imaging, and possibly headset camera and spatial mapping data, to another location, through infrastructure the company does not describe.

The regulatory record is non clinical, and the clinical record is early feasibility work. The 510(k) summary (K223125) reports non clinical testing only: target registration error on a phantom (mean 2.543 mm), cadavers (2.293 mm) and animals (2.9 mm), a usability evaluation with intended users in simulated conditions, and a GLP porcine study whose results are not in the summary. The company's Publications page lists clinical case reports (prostate biopsy, splenic mass biopsy, cryoneurolysis), early clinical feasibility studies of holographic guidance for percutaneous tumor ablation, one in the Journal of Vascular and Interventional Radiology, a cadaver accuracy study, a comparison of goggle based augmented reality CT and ultrasound fusion against conventional fusion navigation for needle insertion, and review articles. First clinical use is documented with a named operator and institution: a mediastinal lymph node biopsy by Dr Bradley B. Pua, Chief of Interventional Radiology at NewYork-Presbyterian / Weill Cornell Medical Center, in November 2023, and a trial of navigation accuracy was reported in March 2024 without its numbers, cohort or method on the company's pages.

No study shows that augmented reality navigation changes procedural success, complication rates or radiation exposure against conventional guidance.

One fact about the data path is published, and nothing about stewardship. The 510(k) summary (K223125) describes the system as a head mounted display wirelessly connected to a streamer beside the ultrasound system in the procedure room, receiving the patient's CT volume. The home page also offers remote collaboration, consultation, proctoring and support, which means procedural imaging, and whatever the headset shares, can leave the institution during a live procedure.

Nothing published says what is transmitted, over what path, whether sessions are recorded at either end, or what a remote participant's institution retains. A headset of this type also builds a spatial map of the room and its outward cameras see the team and the patient; nothing states whether any of that is retained or used for support, training or product improvement, or whether staff are told what the headset captures. The privacy policy covers the website only. Safety engineering, meaning guardrails and how a safety event is handled, is not published.

No statement of status, and the only privacy document covers the website. MediView's privacy policy (updated 10 March 2026) describes personal information for which the company is a data controller, giving website visits as the example, and does not mention XR90, patient data, procedure imaging, headset data, remote sessions, HIPAA or business associate agreements. No other HIPAA or data protection statement is published.

The deployment involves more parties than a vendor and a hospital: the headset platform is Microsoft's; the company works with major imaging manufacturers; and remote collaboration lets a clinician at another organization take part in a live procedure, whom an agreement between the hospital and the vendor does not by itself govern.

Nothing independent is published on security. The site links no security page, trust center or vulnerability disclosure policy, the privacy policy says only that the company has adopted security measures designed to protect personal information and that no system is perfect, and the 510(k) summary as published documents no cybersecurity testing. A cleared device carries quality system obligations, which govern design control rather than information security.

The headset's hardware security features, device encryption and platform certifications are Microsoft's, not MediView's. This is also a cleared device dependent on a single third party headset line that Microsoft has stepped back from, so hardware availability and support horizon are security and continuity questions at once.

The clearance is identifiable and its indication is legible. FDA records list XR90 (XR90-SYS) as K223125, decided 13 July 2023, under 21 CFR 892.2050, product code LLZ, with HOLOSCOPE-i (K210072) as predicate. Its indication is 3D image visualization and image interaction in conjunction with traditional imaging and monitors, for pre operative planning and the heads up intra operative display of images during ultrasound guided needle based procedures; the company describes the uses as biopsies and tumor ablations in soft tissue and bone, and says this was the first 510(k) for an augmented reality device combining live imaging with 3D extended reality visualization across pre operative and intra operative use, which is its own claim.

The clearance authorizes a display used alongside standard imaging, not any autonomous determination. The summary describes a commercial off the shelf headset, a GE Vivid iq ultrasound interface and electromagnetic tracking, so a change of headset or ultrasound system may need its own clearance. OmnifyXR, marketed on the home page, is not separately listed.

No model is described to govern, and nothing is published on how model behavior is governed or tested. The question that matters for this device is registration accuracy across patients: accuracy between a prior CT and live ultrasound depends on tissue characteristics, body habitus, respiratory motion and acoustic window quality, all of which vary across populations, so a navigation system that registers less accurately in larger patients would be quietly less safe for them.

The 510(k) summary (K223125) reports target registration error on a phantom, cadavers and animals, with no human or per population results. Mayo Clinic and Cleveland Clinic are investors in the company, and Cleveland Clinic is also the origin of the underlying intellectual property and hosted early clinical work, so the institutions closest to validating the technology also hold an interest in it. That arrangement is common in device commercialization.

A review mechanism is built into the label, and nothing stands behind the output. XR90 is cleared (K223125) for 3D image visualization and interaction in conjunction with traditional imaging and monitors, so the clinician keeps the standard ultrasound image as the reference a hologram can be checked against.

The summary reports target registration error, a mean of 2.543 mm on a phantom (95 percent upper bound 2.726 mm), 2.293 mm in cadavers (2.825 mm) and 2.9 mm in animals (3.4 mm); those are figures for a registration function the filing does not describe as learned, so they measure the device rather than an AI component. No warranty, indemnity, accuracy commitment or remediation route is published.

The stakes are specific to navigation: a detection aid that is wrong produces a flag a clinician can disregard, while a navigation overlay that is wrong can place a needle in the wrong place, and the error is invisible at the moment it matters because the display shows the instrument where the system believes it to be; registration also degrades with patient movement, respiration and tissue deformation.

The imaging systems it reads from are named in the clearance, and the record is not reached. Interoperability for this device means imaging systems: K223125 describes XR90 interfacing with a GE Vivid iq ultrasound system for the live feed and receiving 3D volumetric data from CT sources, and the home page quotes a user building the 3D model on GE HealthCare's Allia platform.

The company does not publish which other ultrasound systems are validated, how CT volumes arrive (DICOM from PACS or otherwise), or any EHR or procedure documentation path. GE HealthCare led the Series A, and nothing says whether integration is optimized for, or restricted to, GE imaging equipment.

On premises in the procedure room by design, with the remote paths undescribed. K223125 describes a commercial off the shelf augmented reality headset wirelessly connected to a streamer that interfaces with the ultrasound system and an electromagnetic field generator, so registration and display run in the room.

The headset is Microsoft's HoloLens 2, and that dependency is the most important deployment question: the clinical viability of a cleared device rests on hardware whose roadmap, availability and support lifecycle belong to a third party, and Microsoft has publicly stepped back from that product line; a change of display hardware may carry regulatory consequences of its own. The remote collaboration feature sends procedural imaging outside the room, and no hosting, region or residency detail is published for it.

No price, tier or commercial model is published for XR90 or OmnifyXR, and the home page routes buyers to contact the company and try on a headset. For a device platform the open questions are more numerous than for software: capital purchase or subscription; whether headsets are included, separately procured or customer supplied, and how many a procedure room needs; whether per procedure consumables, tracking markers or drapes apply; and what service, calibration and software update coverage costs.

Reimbursement is the other half of the model: augmented reality navigation used adjunctively during a needle procedure has no obvious separate payment pathway, and no position is published. The home page quotes a user building 3D models on GE HealthCare's Allia platform, and GE HealthCare led the company's Series A, but no sales route through GE is stated.

Narrow, clearly stated, and validated outside patients so far. The cleared scope (K223125) is pre operative planning and heads up intra operative display during ultrasound guided needle based procedures, used in conjunction with traditional imaging; the company names biopsies and tumor ablations in soft tissue and bone as the uses, and interventional radiology is the primary buyer. Named clinical settings are NewYork-Presbyterian / Weill Cornell and Cleveland Clinic.

The validation behind the clearance is phantom, cadaver and animal testing, with first clinical use in November 2023 and an accuracy trial reported in 2024. The company describes broader ambitions across image guided intervention, and OmnifyXR is marketed for interventional suites; those sit outside what is validated today.