Written and maintained by CASRAI Editorial Board
Last updated
Viz.ai is a clinical artificial intelligence care-coordination and radiological triage platform best known for Viz LVO (originally cleared as ContaCT), which in February 2018 received the first U.S. Food and Drug Administration (FDA) De Novo authorization (DEN170073) for computer-aided triage and notification software (CADt) and in September 2020 became the first medical AI software awarded a Centers for Medicare & Medicaid Services (CMS) New Technology Add-on Payment (NTAP). Together with Digital Diagnostics’ LumineticsCore (formerly IDx-DR)—which received FDA De Novo authorization DEN180001 in April 2018 as the first autonomous AI diagnostic system—Viz.ai and IDx-DR established the two foundational regulatory, liability, and reimbursement archetypes for commercial clinical AI in the United States.
What Is Viz.ai? Architecture, Viz LVO, and Multi-Disease Modules
In acute ischemic stroke caused by a large vessel occlusion (LVO) of the internal carotid artery (ICA) or middle cerebral artery (MCA M1/M2 segments), clinical outcomes depend heavily on rapid endovascular mechanical thrombectomy. Traditional hub-and-spoke hospital stroke workflows suffered from serial communication bottlenecks: a patient arriving at a primary stroke center (“spoke” hospital) underwent a head and neck computed tomography angiography (CTA) scan, waited for a general radiologist to read the queue, waited for an emergency physician to telephone a comprehensive stroke center (“hub”), and waited for images to be burned to disc or transferred via PACS before a neurointerventionalist could make a transfer decision.
Viz.ai re-architected this workflow through an automated cloud-and-mobile pipeline:
- DICOM Edge Gateway and Automated CTA Ingestion: As soon as the CT scanner completes acquisition, an on-premises hospital DICOM router forwards the non-contrast CT, CTA, or CT perfusion (CTP) series over TLS to the HIPAA-compliant Viz.ai cloud without waiting for PACS archival or human technician intervention.
- Deep Learning Inference (`Viz LVO` Algorithm): A 3D convolutional neural network analyzes the arterial phase CTA volume in minutes to detect suspected intracranial large vessel occlusions in the anterior circulation.
- Parallel Specialist Notification (Mobile CADt Alerting): When a suspected LVO passes the operating threshold, Viz.ai simultaneously dispatches a push alert to the smartphones of both the spoke emergency team and the hub neurointerventionalist—bypassing sequential phone trees.
- Zero-Footprint Mobile DICOM Viewer and Secure Messaging: Within the FDA-cleared mobile and web application, clinicians view synchronized 3D maximum intensity projection (MIP) CTA reconstructions, color-coded CT perfusion maps (
Tmax > 6sand cerebral blood flowCBF < 30%viaViz CTP), and HIPAA-compliant care-team chat threads tied to the patient’s EHR integration encounter.
Following its initial stroke clearance, Viz.ai expanded its platform through subsequent FDA 510(k) premarket notifications across neurovascular, cardiovascular, and trauma service lines—including Viz ICH (intracranial hemorrhage), Viz SDH (subdural hematoma), Viz AN (cerebral aneurysm), Viz PE (pulmonary embolism with automated right-to-left ventricular RV/LV diameter ratio measurement), Viz Aortic (suspected aortic dissection), and Viz HCM (12-lead ECG screening for suspected hypertrophic cardiomyopathy).
How Viz.ai (CADt) Compares with IDx-DR / LumineticsCore (Autonomous AI)
In clinical AI regulation and clinical decision support system design, FDA CDRH draws a sharp boundary between computer-aided triage (CADt) software—pioneered by Viz.ai and peer radiology platforms like Aidoc—and autonomous diagnostic AI, pioneered by Digital Diagnostics’ LumineticsCore (originally cleared as IDx-DR):
- Viz LVO (`DEN170073` — Assistive Triage / CADt): Classified under 21 CFR § 892.2080 (Product Code
QAS, Class II). Under FDA special controls, Viz LVO is strictly a parallel workflow prioritization tool. It notifies a specialist to review a scan early, but it cannot remove a study from the standard radiology reading queue, cannot mark a scan as “normal,” and cannot render a standalone diagnostic conclusion without a qualified physician interpreting the original diagnostic workstation images. - IDx-DR / LumineticsCore (`DEN180001` — Autonomous Diagnostic AI): Classified under 21 CFR § 886.1100 (Product Code
PIB, Class II). LumineticsCore analyzes macula- and optic-nerve-centered fundus photographs captured by a non-specialist operator in a primary care clinic using a Topcon NW400 non-mydriatic retinal camera. Using two biomarker detector networks (evaluating hemorrhages, microaneurysms, exudates, and neovascularization), LumineticsCore outputs an autonomous diagnostic directive—either “More than mild diabetic retinopathy detected: refer to an eye care professional” or “Negative for more than mild diabetic retinopathy: rescreen in 12 months”—without any physician or ophthalmologist overreading the retinal image.
| Dimension | Viz.ai (Viz LVO / ContaCT) | LumineticsCore (formerly IDx-DR) |
|---|---|---|
| Developer | Viz.ai, Inc. | Digital Diagnostics Inc. (formerly IDx LLC) |
| Landmark FDA Authorization | De Novo DEN170073 (February 13, 2018) |
De Novo DEN180001 (April 11, 2018; rebranded via 510(k) K203629) |
| CFR Regulation & Product Code | 21 CFR § 892.2080 — Product Code QAS (Class II) |
21 CFR § 886.1100 — Product Code PIB (Class II) |
| AI Autonomy Category | Computer-Aided Triage & Notification (CADt) — clinician interprets original CTA | Autonomous Diagnostic AI — makes clinical referral decision without physician overread |
| Pivotal Trial Performance | Sensitivity 87.8%, Specificity 89.6%, and ~52-minute median reduction in time-to-notification | Pivotal primary care trial (N = 900, Abramoff et al., 2018): Sensitivity 87.2%, Specificity 90.7%, Imageability 96.1% |
| Primary Medicare Reimbursement | Inpatient IPPS New Technology Add-on Payment (NTAP) (FY 2021–2023, up to $1,040) | Outpatient MPFS CPT Code 92229 (Point-of-care autonomous retinal imaging) |
How Is Viz.ai Used in Clinical Research and Healthcare Operations?
Beyond acute emergency triage, Viz.ai and autonomous diagnostic platforms are deployed across three clinical research and health-system domains:
- Door-in-Door-Out (DIDO) and Door-to-Puncture Stroke Quality Metrics: Health systems and stroke registries measure Viz.ai implementation using spoke hospital Door-in-Door-Out (DIDO) transfer intervals, Door-to-Groin-Puncture times, and 90-day modified Rankin Scale (mRS 0–2) functional independence rates. Pragmatic health-services studies evaluate how automated parallel alerting reduces inter-facility transfer delays across rural and community hospital spokes.
- AI-Driven Clinical Trial Patient Screening (`Viz Recruit`): Time-sensitive neurovascular and cardiology clinical trials (such as acute ICH evacuation trials, novel thrombolytic trials, or hypertrophic cardiomyopathy trials) chronically under-enroll because research coordinators cannot manually screen 24/7 emergency CT and ECG feeds. Viz.ai’s research module automatically screens incoming imaging and ECG studies against trial inclusion/exclusion criteria (such as hematoma volume thresholds or LVO location) and alerts the on-call site principal investigator and clinical research coordinator in real time.
- Reimbursement Precedents for Software as a Medical Device (GMLP / SaMD): Viz LVO and LumineticsCore serve as the primary health-economics case studies for the two main CMS payment pathways for clinical AI: inpatient bundled add-on payments (IPPS NTAP, which requires demonstrating newness, cost above the MS-DRG threshold, and substantial clinical improvement) versus outpatient Physician Fee Schedule national coding (CPT 92229, the first Category I CPT code created specifically for autonomous artificial intelligence analysis without physician interpretation).
Frequently Asked Questions
Why did IDx-DR change its name to LumineticsCore?
In 2020, IDx LLC rebranded the company as Digital Diagnostics and subsequently renamed the IDx-DR product LumineticsCore to reflect a broader multi-disease autonomous diagnostic platform strategy while maintaining the same underlying FDA De Novo (`DEN180001`) and 510(k) (`K203629`) regulatory clearance for diabetic retinopathy screening.
Can Viz.ai rule out a stroke on a CT angiogram?
No. Under its FDA Class II special controls (21 CFR § 892.2080, Product Code QAS), Viz LVO is cleared solely as a triage and notification aid for suspected positive large vessel occlusions. A negative (non-alerted) Viz.ai scan does not rule out a stroke, distal vessel occlusion, or non-vascular pathology, and every CTA must still undergo full diagnostic interpretation by a radiologist.
How does CMS New Technology Add-on Payment (NTAP) work for AI software like Viz.ai?
Under the Medicare Inpatient Prospective Payment System (IPPS), hospitals normally receive a single bundled MS-DRG payment for an inpatient stay. When CMS grants an NTAP designation (as it did for Viz LVO in FY 2021 with ICD-10-PCS code 4A03X5D), Medicare pays the hospital an additional add-on payment (up to 65% of the eligible technology cost, capped at $1,040 per eligible case for Viz LVO during its active 3-year NTAP window) when total case costs exceed the MS-DRG charge threshold.
