Cleerly

The product is AI analysis of coronary CT angiography. Cleerly Labs quantifies and characterizes atherosclerotic plaque and stenosis, and Cleerly ISCHEMIA derives a machine learning estimate of vessel level ischemia from the same imaging. There is no scanner or services layer; the company sells the algorithmic reading of a scan the customer already acquired. AI is the whole offering.

Explicitly clinical decision support. Cleerly ISCHEMIA is described as a diagnostic aid to be used with the entirety of relevant patient specific data by an interpreting healthcare provider, and results are reported by a physician. The plaque analysis produces quantitative measures the clinician incorporates, rather than issuing a standalone diagnosis. Prescription only and physician mediated, with the human retaining interpretation.

The clinical validation is published and specific: Cleerly ISCHEMIA was derived and validated using data from the CREDENCE and PACIFIC trials, and a PACIFIC sub-study reported high diagnostic accuracy against PET, FFR-CT, SPECT, and invasive FFR. The company states the ISCHEMIA prediction is based on coronary lesion morphology and plaque characteristics, and cites a training base of millions of images from more than 40,000 patients. What is less transparent is the model architecture itself and independent head to head benchmarking against other plaque analysis vendors. Strong on clinical derivation, lighter on model internals.

The published notice excludes patient data by design and a separate clause establishes the answer anyway, which makes this knowable by inference rather than by statement. The notice deliberately does not cover patient data, so it offers no retention schedule, deletion commitment or training terms for the asset the product actually handles, which is the coronary study.

The clause that does reach patient data points in one direction: de identified or aggregated patient information may be used alongside customer personal information to perform analytics and gather product insight. De identification is a mitigation rather than an exclusion, so submitted studies do contribute to product improvement in some form, and a buyer who reads only the scoping sentence would conclude the opposite. Two commitments deserve credit and are unusual enough to name.

There is a stated undertaking not to attempt re identification of de identified data, which is a real constraint rather than a description, and services are confined to the United States and its territories, which settles residency without requiring a separate assurance. What remains open is duration and scope. Ask how long submitted studies persist after analysis, whether they inform model development as distinct from analytics, what deletion can be required on termination, and for a sub processor list.

Among the strongest evidence bases in cardiac imaging AI. Validation draws on the multicenter CREDENCE and PACIFIC trials plus the international CONFIRM2 registry, with late breaking CONFIRM2 presentations reporting a twelve fold risk gradient based on total plaque burden.

The company and independent cardiologists frame Cleerly as having invested in outcome oriented clinical studies rather than detection accuracy alone, which is precisely the evidence standard the field says is missing from most imaging AI. Buyers should note much of the foundational trial data was generated by the company's clinical network, but it is peer reviewed and multicenter.

The public privacy notice deliberately excludes patient data, so it offers no retention schedule, no deletion commitment and no model training terms for the asset the product actually handles, the coronary CT study. The one statement that does reach patient data points the other way: de identified or aggregated patient information may be used alongside customer personal information to perform analytics and gather product insight.

De identification is a mitigation rather than an exclusion, so submitted studies do contribute to product improvement in some form. Credit is due for a stated commitment not to attempt re identification of de identified data, and for confining the services to the United States and its territories. Worth establishing how long submitted studies persist after analysis, whether they inform model development, and what deletion can be required on termination.

The privacy notice draws an unusually clean regulatory boundary, stating explicitly that it does not cover protected health information processed on behalf of customers, while a separate consumer health privacy policy addresses the Washington, Nevada and Connecticut consumer health data regimes. Processing PHI on behalf of a customer is the substance of business associate status, and HITRUST certification is built on HIPAA safeguards.

Held below the top of the band because neither the term business associate nor a business associate agreement appears anywhere in the public materials. A buyer will find the BAA only inside the master services agreement and should confirm its terms before assuming coverage.

Public trust center at trust.cleerlyhealth.com, carrying both HITRUST and SOC 2 alongside a publicly posted redacted information security incident response plan and a security document packet. Control coverage is enumerated across twelve domains including access control, network security, data security, application security, security monitoring, and response and recovery, with sub domains for internal audits, risk assessments and regulatory compliance.

HITRUST is the most prescriptive healthcare specific security framework and very few vendors in this category hold it. Two honest limits sit alongside the grade rather than driving it: the SOC 2 type is not specified on the public page, and the HITRUST assessment level is not named. Both are worth requesting, with current certificate dates.

Cleerly holds multiple FDA 510(k) clearances, including Cleerly ISCHEMIA cleared K231335 and the earlier plaque analysis clearance dating to 2020 as the first company cleared for AI quantification and characterization of coronary plaque types. Clear, current, multi product US authorization with documented clearance numbers.

Unusual for this category in the direction that matters: performance across patient subgroups has been examined in peer reviewed work rather than asserted. Published analyses include diagnostic performance of the ischemia algorithm stratified across sex and age subgroups, a CONFIRM2 registry analysis reporting that quantitative plaque features carry a higher relative risk in women, and quantification distributions stratified by age, sex and self reported race across a cohort of roughly thirty thousand patients.

Held below the top of the band on three counts. There is no published AI governance framework, bias policy or responsible AI statement in the company's own materials, and the trust center's control domains are security domains rather than AI ones. The studies are largely company sponsored, several carrying employee or advisory board authorship, though competing interests are disclosed in the papers.

And race appears in distribution work describing how plaque presents across groups rather than in validation showing the algorithm performs equally well across them, which is a different question. Worth asking for performance stratified by race and by image quality.

The derivation and validation are published and rest on named trial datasets rather than on internal cohorts, which is the difference between evidence a reader can locate and evidence they must accept. The strongest feature is the comparator: a sub study reported diagnostic accuracy against multiple reference standards including invasive measurement, and invasive measurement is the definitive answer in this domain rather than another imaging estimate.

A model validated against the thing it is trying to predict, rather than against a cheaper proxy for it, has been tested where it could most clearly have failed. The company also states what the prediction is based on, naming coronary lesion morphology and plaque characteristics, and cites a training base of millions of images from more than forty thousand patients, so a reader knows the input and the scale. Held below the top grade on two counts.

Model architecture is not described, so the derivation is transparent while the internals are not. And no independent head to head benchmarking against other plaque analysis vendors was located, which this index has credited elsewhere in this cluster and marked another vendor down for avoiding. No warranty, indemnity or remediation commitment attaches. Ask whether the company will enter a multi vendor comparison, and for performance in patients unlike the trial populations.

Cleerly operates as a CCTA analysis service that returns a physician report, sitting in the cardiac imaging and reporting pathway rather than integrating deeply with the EHR. Specific PACS, cardiology information system, and EHR connectors are not enumerated in the materials reviewed, so integration is described at the workflow level rather than through a published connector list.

Delivered as a cloud analysis service on submitted CCTA studies. Specific hosting, tenancy, and data residency terms are not published, which matters given cardiac CT studies contain identifiable imaging data.

More visible than most imaging AI on the economics that decide adoption, because the reimbursement pathway is public and specific. The plaque analysis carries Category I CPT code 75577 effective January 2026, replacing earlier Category III codes, and ISCHEMIA is billable under Category I code 75580; Medicare coverage plus commercial payer policies from UnitedHealthcare, Cigna, Humana, Aetna and others are reported to span more than 86 million covered lives. Per study product pricing to the provider is not published, but the reimbursement structure a buyer actually needs to model the return is unusually well documented.

Deliberately concentrated on coronary artery disease evaluation from CCTA, spanning plaque quantification and ischemia estimation across the diagnostic pathway for known or suspected CAD. Deep within cardiovascular imaging rather than broad across radiology, which is the source of both its evidence depth and its reimbursement wins, but it does not extend beyond cardiac.