TeraRecon

Platform not algorithm, and this is the radiology counterpart of the case this index already documents in digital pathology. TeraRecon's core asset is Intuition, advanced visualisation software with roughly 1,900 installed clinical sites that functions entirely without AI, and its second asset is vendor neutrality across imaging equipment and PACS.

Eureka Clinical AI is a distribution and workflow layer whose distinguishing feature is that it carries other companies' models: Riverain, ClariPi, Coreline Soft, ImageBiopsy Lab and Imaging Biometrics all supply the intelligence. TeraRecon's own contribution is integration, consolidated management and reach, which is genuinely valuable and is not a model. The same reasoning applies to Indica Labs.

The grade describes the mechanism, and a buyer should understand they are purchasing an AI delivery channel with a visualisation platform underneath rather than an algorithm developer.

The described loop is human confirmatory: the company positions Eureka as clinical decision augmentation, multi specialty teams receive results and mobile alerts to confirm AI findings, and clinicians can interact with the output. That is a reasonable posture for interpretation support.

The structural issue is specific to marketplaces and it is the thing to press on. Autonomy is a property of each algorithm rather than of the platform. A site running five third party models is running five different intended uses, five sensitivity profiles and five sets of failure modes, presented through one consistent interface that makes them look alike.

No unified confidence framework, no cross algorithm governance layer, no published admission standard for what a model must demonstrate before it appears on Eureka, and no aggregate performance monitoring across enabled algorithms was located. The interface consistency that makes the platform valuable also flattens differences a radiologist needs to keep in view.

TeraRecon does not build the models, so the transparency question routes through to each partner and the platform publishes little of its own. Credit where it is due: the company names its algorithm partners openly and specifically, which is better practice than the unnamed original equipment supplier arrangements this index has repeatedly found elsewhere, and it means a buyer can trace any given capability to the company that actually built it.

What is absent is any published standard for admission to Eureka, any validation TeraRecon performs before listing a model, or any description of how algorithm versions are managed and communicated to sites when a partner updates a model.

The partners are named, which is the useful half, and nothing governs what happens to data once it reaches them. This platform sits between a large installed base of clinical sites and several third party model providers, so a buyer needs to know what leaves the institution, which partner receives what, and under whose agreement, and none of that is published: no retention position, no de identification standard, no per partner terms and no statement on whether imaging routed to a partner algorithm is retained after the result returns.

A certified information security management system does not fill the gap, and the distinction is the one this index applies to attestations generally: it establishes that controls over confidentiality and access are documented and audited, states no retention period, takes no position on training, and names no de identification standard. A control framework is not a stewardship position.

One further question is specific to the corporate structure and should be put in writing rather than inferred: whether the parent group, whose principal business is real world evidence and data products for life sciences, has any pathway to imaging or derived outputs flowing through this platform. That is not an accusation and there may be a clean answer, and the context makes it a question a buyer owns rather than one they can assume away. Ask it, plus per partner retention and the training position.

Two genuinely third party signals, which is more than most of this category can show. TeraRecon was named KLAS Category Leader for Advanced Visualization in 2020, 2021 and 2022 and is a Best in KLAS provider, which is an independent buyer surveyed assessment rather than a vendor claim.

More interestingly for this index, TeraRecon appears among the eight FDA approved plaque analysis systems benchmarked head to head in the Nature Reviews Cardiology quantitative coronary imaging consensus statement. Submitting to an independent comparative benchmark is itself creditable, and it is worth recording that the vendor with the largest evidence base in that field, Heartflow, was unable to provide the analysis required for inclusion in the same exercise.

Held at B rather than A because scale and awards are not outcome evidence, no peer reviewed study of the platform itself was located, and the clinical performance evidence that does exist largely belongs to the third party algorithms rather than to TeraRecon.

No published statement on imaging data handling, retention, de identification or secondary use was located for the platform. That absence carries more weight here than for a single algorithm vendor, because Eureka is a layer sitting between a large installed base of clinical sites and several third party model providers, so a buyer needs to know what leaves the institution, which partner receives what, and under whose agreement.

The ISO 27001 certification credited on the security row does not fill this gap, and the distinction is the same one this index makes about attestation reports generally. A certified information security management system establishes that controls over confidentiality and access are documented and audited. It states no retention period, takes no position on whether customer imaging contributes to model development, and names no de identification standard. A control framework is not a stewardship position.

Two questions are specific to this vendor's structure and neither is answered. Whether imaging routed to a third party algorithm is retained by that partner after the result returns, and on what terms. And whether the parent group, whose principal business is real world evidence and data products for life sciences, has any pathway to imaging or derived outputs flowing through this platform. That second question is not an accusation and there may be a clean answer, but the corporate context makes it one a buyer should put in writing rather than infer.

No statement or business associate agreement terms were located. A platform intermediating between health systems and multiple third party algorithm vendors raises a specific contractual question that public materials do not address: whether the site executes one agreement with TeraRecon that covers downstream partners, or separate agreements with each algorithm provider. Establish the chain before enabling any third party model.

The company's ISO 27001 certification does not reach this question and should not be read as though it does. That standard certifies an information security management system. It says nothing about who the covered entity contracts with, which party carries the obligation when patient imaging is routed to a third party algorithm, or what happens to that data once it reaches a partner's environment. A buyer that sees a certification badge and stops asking will have answered a different question from the one that matters here.

The practical ask is a map rather than a document. For each algorithm a site intends to enable, establish who holds the agreement, whether imaging leaves the platform to reach that vendor, what that vendor may retain, and who is responsible for notifying the health system if a partner suffers an incident. The value of an open platform is that a site can run several vendors' models through one integration. The cost is that the contractual chain multiplies behind a single interface, and nothing published describes its shape.

An independent certification does exist and it is held in the right name. TeraRecon Inc. was certified to ISO 27001:2022, announced in May 2023, with the standard and its version stated rather than a vague claim of alignment. That is a certified information security management system assessed by an external body, and it is issued to the operating entity itself rather than to the parent group, which is the distinction this index most often has to make when a company sits inside a larger organisation.

Held at B rather than A for three reasons. There is no trust centre and no security page where a buyer could request the certificate, read its scope statement or see what else is held. The certification was announced in 2023 and certificates of this type run on a three year cycle with annual surveillance audits, so currency and current scope should be confirmed rather than assumed from an announcement. And no SOC 2 report or healthcare specific certification was located alongside it.

One piece of language to read carefully rather than credit. Material announcing the platform's software as a service transition states that automatic updates and security patching ensure enhanced compliance including HIPAA, data protection law and SOC 2. That describes patching as supporting compliance objectives. It is not an assertion that the company holds a SOC 2 report, and it should not be read as one. Ask for the certificate, its scope, its expiry and whether the Eureka platform is inside the certified boundary.

The platform has standing of its own, evidenced externally by its inclusion among the FDA approved plaque analysis systems listed in the Nature Reviews Cardiology consensus statement.

The more important point for a buyer is structural, and it is the rule this index already applies to open platforms in digital pathology: on a marketplace, regulatory status is per algorithm rather than per platform. Each third party model carries its own clearance, its own intended use and its own indication, and several may be displayed in one workflow with differing status. Coreline Soft's lung and cardiac algorithms are described as FDA cleared and CE marked, while other partners' positions vary and must be checked individually.

A site should maintain its own register of which enabled algorithms are cleared, for what indication, and which are research use only, because the platform's consistent presentation will not surface the difference.

No fairness, subgroup or performance variation disclosure was located, and no requirement that partner algorithms publish one before listing. The governance question that distinguishes this record, and it belongs on the record as a question rather than an allegation: TeraRecon's parent, ConcertAI, is a real world evidence and AI company whose commercial business is supplying data derived insight to life sciences customers.

TeraRecon sits inside roughly 1,900 clinical sites processing diagnostic imaging. Nothing published states where the boundary lies between imaging processed for clinical interpretation and data that may inform the parent's evidence products, at what aggregation, under what consent, or whether a site can decline that use and remain a customer.

This index has asked the same structural question of Proscia under a diagnostics manufacturer and of Proximie contributing surgical data to a chip maker's foundation models. It is the third instance of the same shape and it should be answered in writing rather than assumed either way.

This company does not build the models, so the transparency question routes through to each partner, and the credit is that it makes that routing possible. Algorithm partners are named openly and specifically, which is better practice than the unnamed supplier arrangements this index has repeatedly found elsewhere, and it means a buyer can trace any given capability to the company that actually built it and go and read that company's evidence.

A marketplace that hides its suppliers leaves a customer unable to diligence anything. What is absent is everything the platform owns rather than passes through. No published standard for admission to the catalogue, no validation the platform performs before listing a model, and no description of how algorithm versions are managed and communicated to sites when a partner updates a model.

That third one is the operational risk a marketplace uniquely creates: a hospital validated a partner's algorithm at a point in time, the partner ships a new version, and unless the platform tells the site, the tool in use is no longer the tool that was assessed.

This index has now recorded the same admission and monitoring gap on three separate platforms hosting third party clinical models, which makes it a structural feature of the marketplace model rather than an oversight by one vendor. Ask for the admission criteria, what validation precedes listing, and how version changes are notified to sites.

The strongest interoperability position of anything assessed on this sourcing list, and it is the product's actual thesis rather than a feature. TeraRecon is deliberately independent of any single imaging manufacturer or PACS vendor, which is the opposite of the scanner coupled architecture common in this space, and it reports roughly 1,900 clinical sites running on that basis.

Eureka extends the same principle to algorithms: a site integrates the platform with its PACS once and then manages multiple third party models through it, rather than negotiating a separate integration for every AI vendor it buys. For a radiology department already carrying integration debt, consolidating that surface is a substantial part of the value, and it is a genuine architectural advantage over point solutions.

Eureka is described as a SaaS platform and Intuition is an installed advanced visualisation product, so most sites will run a hybrid, but no hosting provider, region, tenancy model or data residency commitment is published. For a SaaS layer processing diagnostic imaging across roughly 1,900 sites internationally, residency and the location of third party algorithm inference are standard procurement questions and neither is addressed publicly.

No price, tier or licensing model is published. The marketplace structure makes the commercial question more layered than for a single product: a buyer needs to establish what the Intuition platform costs, what Eureka costs on top of it, and critically whether third party algorithms are priced by TeraRecon or contracted separately with each partner, since that determines both the negotiation and where liability sits. Whether algorithm pricing is per study, per site or per subscription is also unaddressed, and per study economics behave very differently at screening volumes.

Among the broadest in the index and evidenced by installed base rather than asserted. Named specialty coverage spans radiology, oncology, cardiology, neurology and vascular surgery, and the algorithm partnerships map onto that breadth concretely: chest and lung interpretation, musculoskeletal radiograph measurement, neuro oncology MR perfusion, coronary calcium scoring and emphysema quantification.

Buyer types include hospitals, imaging centres and independent radiology practices, at roughly 1,900 clinical sites globally. The marketplace model is what produces the breadth, since coverage expands by adding a partner rather than by building a new model, which is also why depth in any given indication belongs to the partner rather than to TeraRecon.