AI-guided POCUS has a market access problem.

Point-of-care ultrasound has expanded rapidly across clinical settings, driven by device miniaturization, falling hardware costs, and the genuine clinical value of bedside imaging. That expansion has not been matched by investment in the reimbursement infrastructure required to capture its value. AI-guided imaging is now entering that same environment with a stronger clinical case and the same unsolved market access problem.

POCUS already has a reimbursement problem

Before addressing what AI-guided imaging can do, it is worth establishing what the broader POCUS market has not done. The segment is one where clinical utilization has outrun billing discipline, documentation practice, and payer policy alignment at the same time.

The bring-your-own-device problem

A significant share of POCUS in practice today follows an informal workflow that generates no reimbursable claim and leaves no defensible record. A clinician with a handheld device images a patient at the bedside. The images are reviewed in real time, inform a clinical decision, and then disappear. They are not stored in a PACS or VNA. They are not attached to the encounter in the EHR. No formal interpretation is documented. No CPT code is dropped.

This workflow is common enough to have a recognized name in the industry. Bring-your-own-device POCUS has normalized a practice pattern where the clinical act occurs but the administrative and billing infrastructure around it does not. The consequences are twofold.

Lost revenue. Every undocumented POCUS encounter is a missed billable event. The professional fee for an interpreted POCUS study is real and recoverable, but only when three elements are present. The image is stored in a retrievable system, a CPT code is applied that corresponds to what was assessed, and a formal interpretation is documented that defends that code. Remove any one of those elements and the claim either cannot be filed or will not survive a payer audit. Most BYOD workflows are missing all three.

Medico-legal exposure. An image that informed a clinical decision but was never stored or documented does not exist from a legal or compliance standpoint. If that decision is later questioned, in a malpractice proceeding, a payer audit, or a quality review, there is no retrievable artifact to support it. The clinical act occurred. The record does not reflect it. That gap creates liability that is largely invisible until it is not.

The prior authorization layer

For the clinicians and practices that do execute the compliant workflow, a second barrier emerges in the outpatient setting. POCUS is not uniformly covered without prior authorization across payer types. In many outpatient commercial and Medicare Advantage contracts, certain ultrasound studies require prior auth before the service is rendered.

When prior auth is not obtained and the study is performed, the claim is denied regardless of clinical appropriateness, documentation quality, or CPT accuracy. The clinician did everything right at the point of care and still does not get paid. That dynamic discourages investment in the documentation infrastructure that would make compliant billing possible, because the return on that investment is uncertain when payer policy can eliminate reimbursement upstream of the claim entirely.

The prior auth exposure is most acute in the outpatient environment. Inpatient and emergency POCUS generally operates under different coverage rules. But as POCUS expands into primary care, ambulatory specialty, and community settings, which is precisely where AI-guided imaging is most promising, prior auth becomes a front-line market access problem rather than an edge case.

What the AI evidence actually shows

Against that baseline, a peer-reviewed study published in European Heart Journal Digital Health in May 2026 demonstrated something the field has been building toward for years. Undergraduate students with no clinical training and no ultrasound experience completed a four-hour workshop, then used an AI-guided handheld device to acquire cardiac ultrasound images sufficient for diagnosing left ventricular dysfunction. The accuracy was not marginal. It was clinically meaningful.

The Mayo Clinic study enrolled 496 adults referred for diagnostic echocardiography. Three undergraduate operators performed focused cardiac ultrasound using the Philips Lumify handheld device guided by UltraSight AI image acquisition software. A Mayo Clinic AI algorithm interpreted the images, with expert cardiologist review reserved for flagged or uninterpretable scans.

The results validated a two-step screening workflow. AI read every scan first. Only 15.1% of scans required cardiologist review. The combined workflow achieved sensitivity of 96.2%, specificity of 95.4%, and a negative predictive value of 99.8% for detecting left ventricular ejection fraction below 40%. A prospective validation cohort of 344 additional patients produced sensitivity of 100% and specificity of 99.4%.

The clinical problem this addresses is substantial. More than half of LV dysfunction in community populations goes undetected, with higher rates in underserved communities. Guideline-directed medical therapy exists that prevents progression to symptomatic heart failure and reduces mortality. It cannot be initiated in patients who have not been identified.

The authors note appropriate caveats. The study is single-center, the three novice operators represent a narrow sample, and the population was already presenting for diagnostic echocardiography rather than being drawn from a true community screening cohort. Further validation is warranted.

None of that diminishes the core result. AI-guided image acquisition by genuinely novice operators, combined with AI interpretation, can produce diagnostically actionable results. The technology stack works. The deployment environment, as the baseline above establishes, does not.

Three structural barriers to deployment at scale

Companies developing AI-guided imaging tools have generally treated regulatory clearance as the primary commercialization milestone. FDA 510(k) clearance is necessary. It is not sufficient. Three distinct structural barriers sit between a cleared product and a reimbursed clinical workflow, and all three remain largely unaddressed in the AI-guided POCUS segment.

State licensure and scope of practice

State scope of practice frameworks were built around human credentialing. They define who can perform imaging procedures, under what supervision, and with what documented training. None of those frameworks were written with an AI system as the functional operator.

The UltraSight and Philips workflow demonstrated that AI does the work that training previously did, guiding probe placement in real time and reducing acquisition to a task a novice can complete in minutes. State law does not recognize AI guidance as a substitute for clinical credentialing. Three questions remain unanswered across most state lines.

Who is supervising the acquisition? A licensed clinician is typically required to supervise. The credential required for that supervision, and whether it must be synchronous or can be remote, is not settled.

What counts as the qualifying operator? Most state statutes require imaging to be performed by a physician, a credentialed sonographer, or under direct physician supervision. A novice operator guided by AI does not fit any of those categories in their current form.

Does real-time AI guidance constitute training? No state has formally answered this. Until they do, a deployment model that relies on non-credentialed operators carries legal and compliance exposure regardless of the clinical performance data.

The path forward requires engaging state medical boards, working with professional societies to define AI-supervised acquisition standards, and in some cases pursuing legislative or regulatory clarification. That policy strategy has not been built by most imaging AI companies.

CPT coding and coverage architecture

POCUS reimbursement is structured around what was assessed and documented, not how the image was acquired or who acquired it. The professional fee requires a qualified provider to interpret and sign the study. The technical fee, where it exists, typically requires image capture by a physician, a credentialed sonographer, or in an accredited imaging lab.

Those requirements do not accommodate a workflow where a non-credentialed operator acquires images under AI guidance. Even a clinically validated AI-guided POCUS workflow has no clean CPT path to reimbursement under the current coding structure.

Without a reimbursement mechanism, deployment is limited to contexts where costs are absorbed outside fee-for-service, such as direct-to-employer contracts, capitated value-based arrangements, or grant-funded pilots. Those are real near-term markets, but they are not the path to population-level scale.

Building the reimbursement pathway requires analyzing existing CPT codes for applicability to AI-interpreted POCUS under a supervising provider model, constructing the case for new or revised codes through the AMA RUC process where needed, and engaging CMS on coverage determination in parallel. That is a multi-year workstream that needs to begin before deployment, not after. It has not been started systematically in this segment.

Health economics and outcomes research

The Mayo Clinic study does not contain a health economic argument. It does not model cost-effectiveness against standard of care, quantify the downstream cost offset from earlier HFrEF identification, or define the risk segments where a population screening investment generates the best return. That is appropriate for a clinical feasibility study. It is a gap for commercialization.

Payers do not make coverage decisions based on clinical accuracy alone. They require a value demonstration framed in their own terms, meaning cost per quality-adjusted life year, budget impact model, and downstream utilization effect. Without that package, a coverage request arrives as a compelling clinical story with no economic anchor.

The HEOR case for AI-guided cardiac screening is constructable. The clinical evidence for preventing progression from Stage A or B heart failure to symptomatic disease is well-established. The cost of heart failure hospitalization is well-documented. The incremental cost of an AI-guided screening workflow at scale in a primary care or community setting is low relative to the downstream offset. That argument has not been assembled and presented to payers in the AI-guided POCUS context. It remains implicit in the clinical literature rather than explicit in a payer submission.

Why the gap persists

Companies building AI-guided imaging tools are predominantly product and regulatory organizations. Their core competency is algorithm development, FDA clearance, and device integration. Those are hard problems and solving them is the right priority in the early stages of a technology platform.

Market access strategy, HEOR, payer engagement, and scope of practice policy work are different disciplines requiring different expertise, different stakeholder relationships, and a different timeline. A payer medical policy engagement process takes 18 to 24 months from initial submission to coverage determination. A CPT code application cycle runs on a similar timeline. State scope of practice engagement is slower.

The window between clinical validation and commercial deployment is not self-executing. It requires a parallel track of market access investment that most companies in this segment have not yet started. The result is a growing body of clinical evidence sitting ahead of the infrastructure needed to capture its value.

This pattern is not unique to AI-guided POCUS. The Caption Health experience, which demonstrated AI-guided cardiac ultrasound acquisition in primary care settings before its acquisition by GE HealthCare, surfaced the same dynamic. Regulatory clearance and clinical validation were achieved. The reimbursement and scope of practice environment required to deploy at scale was not yet constructed. That remains true for the segment as a whole.

What a market access strategy for this space actually requires

The AI-guided POCUS market access problem is solvable. It requires treating commercialization as a parallel track to clinical development and addressing the baseline POCUS reimbursement infrastructure alongside the AI-specific barriers. The components are well-defined even if they have not been assembled.

Workflow infrastructure first

Before AI-guided acquisition can be billed, the three foundational elements of a compliant POCUS encounter must be in place. Image storage in a retrievable system such as a PACS or VNA, a documented interpretation that is defensible under audit, and a CPT code that accurately reflects what was assessed. AI-guided imaging companies that do not address these elements as part of their deployment model are inheriting the BYOD problem rather than solving it.

Prior auth strategy

In the outpatient settings where AI-guided POCUS is most commercially promising, prior authorization is a front-line barrier. A deployment model that does not include a prior auth workflow, either through payer engagement to establish coverage without prior auth for defined indications or through administrative support for the auth process, will face denial rates that undermine the economics of the program regardless of clinical performance.

HEOR package

A cost-effectiveness model comparing AI-guided POCUS screening to standard of care in defined high-risk populations, such as Stage A heart failure, hypertension with LV dysfunction risk, and post-chemotherapy monitoring, is the analytical foundation for payer engagement. It needs to quantify the downstream cost offset from earlier HFrEF identification in language a payer medical director can act on. Budget impact modeling for a defined attributed population anchors the payer-specific conversation.

CPT and coverage pathway

A systematic review of existing CPT codes applicable to AI-interpreted POCUS under a supervising provider model identifies the nearest available coverage path. Where existing codes are inadequate, a new code application requires clinical evidence, a time-and-effort analysis, and a crosswalk to existing codes. Parallel engagement with CMS on coverage determination is distinct from the CPT process and needs its own workstream. Both need to begin now.

Scope of practice engagement

A state-by-state analysis of existing scope of practice statutes identifies where deployment is feasible under current law, where it requires regulatory guidance, and where it requires legislative action. Professional society engagement, particularly with the American Society of Echocardiography and relevant primary care and emergency medicine organizations, accelerates the credentialing standards conversation. That conversation needs to happen before deployment, not after.

Value-based contract design

While fee-for-service reimbursement pathways are being built, the near-term commercial opportunity sits in value-based arrangements. Accountable care organizations, direct-to-employer contracts, and capitated Medicare Advantage plans can absorb AI-guided screening costs when the downstream savings are clearly modeled. Those contract structures require the HEOR package as their foundation.

The opportunity in the gap

The Mayo Clinic study is a meaningful proof point. Undergraduate operators with four hours of training, guided by AI, produced diagnostic results that matched expert cardiologist performance on a clinically significant finding. That represents years of investment in algorithm development, device integration, and clinical validation by UltraSight, Philips, Mayo Clinic, and others working in this space.

What it does not represent is a commercialization strategy. The technology is ready to extend POCUS capability into primary care, community settings, and underserved populations where the undetected LV dysfunction burden is highest. The reimbursement environment it is entering has not solved the foundational billing and documentation problems that have characterized the broader POCUS segment for years, has not built the prior auth workflows that outpatient deployment requires, and has not assembled the HEOR and coverage strategy that payer adoption demands.

The companies that win in AI-guided imaging will be the ones that treat market access, HEOR, and payer engagement as core commercialization functions rather than downstream activities. That is where the value is sitting unrealized. It is also where the next phase of competitive differentiation will be decided.

All views, analyses, and frameworks presented here reflect independent professional judgment informed by more than two decades of experience across clinical strategy, commercial strategy, and product strategy. They do not represent the views or positions of any current or former employer or affiliated organization.