Visiopharm

The APPs are the product. Each is a deep learning algorithm performing a specific scoring or detection task on whole slide images, sold individually or as a diagnostic package. No scanner or assay hardware is manufactured; the alliance with Agilent validates the algorithms against Agilent's assays and instruments rather than bundling Visiopharm hardware. Remove the models and nothing remains.

The vendor is explicit about where the machine stops. The APPs run fully automated with no manual interaction up to the point of pathologist review, and the company describes the output as objective decision support delivered for review rather than a diagnosis. The PD-L1 validation study tested both stand alone algorithm analysis and algorithm assisted interpretation as distinct conditions, which frames the autonomy question directly. Automation is deep but terminates before sign out.

Individual APPs are described with real specificity, including what each measures and how output is ordered, for example metastasis detection ranking by largest diameter or total area depending on tumor type. Validation designs are disclosed at a useful level: the PD-L1 study compared manual assessment against stand alone and assisted analysis across three European sites with a multi reader reference score. Training data scale, model architecture, and head to head comparison against competing pathology AI are not published, so transparency is strong on validation design and lighter on model internals.

A structurally short chain where the software runs on premises, with the rest unnamed. Visiopharm builds its own APPs, and the diagnostics page says the software runs on the web, on premises or in a hybrid setup, standalone with a slide scanner or inside the laboratory's image management system. An on premises installation means slide images need not leave the laboratory at all. The portfolio's prostate APP is supplied by a third party, and the APPs are sold with Agilent as the Visiopharm Diagnostic Package. For the web and hybrid options, no host, region or subprocessor is named: the privacy policy says external suppliers act as processors under data processing agreements without naming them.

A chain that can be structurally short, with the web option's parties unnamed.

A prospective, peer reviewed trial at one center, plus multi site validation described by the company. CONFIDENT-B (Nature Cancer, 2024) ran the Metastasis Detection APP in routine diagnosis at UMC Utrecht on 190 consecutive breast cancer sentinel lymph node specimens (September 2022 to May 2023), alternating two week blocks between AI assisted and standard reading. AI assistance cut the adjusted relative risk of immunohistochemistry use per detected metastasis to 0.680 (95% CI 0.347 to 0.878), shortened H&E review from 6 minutes 4 seconds to 3 minutes 41 seconds while pathologists looked at more slides, and raised pathologists' overall sensitivity from 55.8 to 60 percent (negative predictive value 78.9 to 88.2 percent), with estimated savings of about 3,000 euros over the trial. The company links the paper and reports the same site's positive business case. For the PD-L1 APP it describes a study across three European sites against a multi reader reference, and metastasis detection verification on breast and colorectal data from three sites.

The trial is single center and nonrandomized, and the biomarker APPs' studies are described without published figures.

The result is a named deployment with dated outcomes and published research, without prospective validation across several sites.

Nothing published describes what happens to slide images. The privacy policy (version 13, updated 14 September 2026) covers personal data of website visitors, newsletter, webinar and event registrants, customers and suppliers, which is business contact data; it sets out legal bases, standard contractual clauses for transfers outside the EEA including to the United States, and deletion when data is no longer needed, all for that data. It says Visiopharm concludes a data processing agreement where it processes personal data on a customer's behalf, which is the right instrument for clinical images, but no agreement, template or terms are published. No retention period, deletion position, de-identification practice or statement on whether customer images train Visiopharm's models is published for the diagnostic APPs, and no safety engineering is described for handling a wrong output as a safety event.

Nothing published on how patient material moves through the product; a policy covering business contacts does not reach it.

A Danish company that does not invoke HIPAA, with its processor role acknowledged and nothing published behind it. The privacy policy (version 13, 14 September 2026) is written to the GDPR and covers business contact data of website visitors, registrants, customers and suppliers; it states that Visiopharm concludes a data processing agreement where it processes personal data on a customer's behalf, and that transfers to third countries including the United States run on standard contractual clauses. No HIPAA statement, business associate position or agreement appears, and the CE-IVD APPs are not for sale in the United States; the research APPs are. Under the non US route, the processor position is stated, but no data processing agreement or document reaching the product's clinical data is published.

The status rests on statements alone. Visiopharm does not say whether it will sign a business associate agreement with a United States research customer, and the data processing agreement covering clinical images for a European laboratory is not published.

Nothing independent stands behind the information security of the product. No ISO 27001 certificate, SOC 2 report, HITRUST certification, trust center, penetration testing statement or vulnerability reporting route is published. The site footer's legal section lists a legal notice, a cookie policy, a privacy policy and an ESG policy; the site's page listing contains no security or trust page; and the privacy policy (version 13, 14 September 2026) describes no security certification. What the company does publish is its EN ISO 13485:2016 quality system certificate as a downloadable document with its validity stated (to 27 December 2027), which is good disclosure practice for what it covers, but ISO 13485 is a medical device quality standard, not an audit of how the company operates information security, so it is not a security credential. The software can be installed on premises, where the laboratory's own controls apply, but it is also offered as a web service with no controls described.

Nothing independent published on information security.

A European regulatory position stated plainly, with the United States position stated beside it. The diagnostics page says Visiopharm's CE-IVD APPs, for biomarker assessment in breast, lung and prostate cancer and lymph node metastasis detection, are certified under the IVDR for in vitro diagnostic use in the EU and UK and are not for sale in the United States, where translational APPs for research are offered instead; the site disclaimer says all products are research use only unless otherwise stated. The company calls this the largest portfolio of IVDR certified APPs and says its first IVD APP was released in 2014; it describes the PD-L1 APP as its ninth IVDR certification. FDA's 510(k) database lists no clearance under Visiopharm. The prostate APP in the portfolio is supplied by a third party and carries the older IVDD mark rather than IVDR. The EN ISO 13485:2016 certificate is published with its validity.

A non US regime named precisely, with the marketing matching it; no US clearance.

Results within the group where the metastasis APP is most at risk are published, by an outside trial. CONFIDENT-B (Nature Cancer, 2024; UMC Utrecht; one author on Visiopharm's advisory board) reports the Metastasis Detection APP's false alerts separately for patients who had neoadjuvant treatment, whose nodes carry treatment changes: more alerts per slide in that group at every suspicion level (yellow 10.0 against 5.4, orange 2.1 against 1.5, red 3.1 against 0.9). It also reports standalone sensitivity by deposit size, 95.8 percent for macro and micrometastases against 44.4 percent for isolated tumor cells. These are results on the product's own failure mechanism, with counts, from a single site on one scanner setup, and cover one APP.

No results by scanner, stain protocol, site or patient demographics are published for the biomarker scoring APPs, which the company says were validated across three clinical sites, and no governance framework or monitoring commitment is named.

Results within the at risk group, short of results compared across the groups the product serves.

The company links an independent trial that reports the algorithm's standalone sensitivity with its denominator, and nothing stands behind the output. CONFIDENT-B (Nature Cancer, 2024), a prospective, single center, nonrandomized trial at UMC Utrecht that Visiopharm links from its integrations page, enrolled 190 consecutive sentinel lymph node specimens from breast cancer patients (100 in the AI arm, 90 in the control arm) and ran the Metastasis Detection APP on both arms against the final diagnosis with immunohistochemistry. Standalone, the algorithm found all 17 macrometastases and all but one micrometastasis, 95.8 percent sensitivity for the two combined, and 44.4 percent of isolated tumor cell deposits; the missed micrometastasis sat in a cauterized area visible only on the immunohistochemistry slide. False alerts were mainly blood vessels, histiocytes, follicle centers, nerves and capsular nevi. One author sits on Visiopharm's advisory board. The figures cover one APP at one site, not the biomarker scoring APPs.

No warranty, indemnity or remediation commitment attaches to a result.

A published error figure with its method and denominator, adopted by the vendor, with nothing contractual behind it.

Named integrations with the systems a pathology laboratory runs, one of them documented in production. The integrations page names Sectra IDS7 and Proscia Concentriq AP, describes analysis triggered automatically from laboratory information system metadata, and states that the Sectra integration uses Sectra standards with no custom changes, shows overlays and results in the Sectra primary viewer, and can upload on-slide controls for internal quality testing. UMC Utrecht runs it in routine diagnosis in its Sectra PACS, as the CONFIDENT-B trial (Nature Cancer, 2024) documents. The Agilent alliance validates the APPs against Agilent assays and the Omnis platform. No integration documentation or marketplace listing is published, and the LIS side is described as a metadata trigger rather than a named system. For a pathology product the LIS and image management system, not the patient chart, are the correct surfaces.

Named systems with read access and results written to the viewer.

Deployment options are stated, residency is committed only where the laboratory hosts it, and isolation is open. The diagnostics page says the software runs on the web, on premises or in a hybrid setup, either standalone with a slide scanner or inside the laboratory's existing image management system or PACS. An on premises installation places analysis and slide images on the laboratory's own infrastructure. The integrations page describes the reference deployment at UMC Utrecht: a slide is scanned, analyzed automatically in the background from LIS metadata, and the results are waiting in the Sectra viewer when the pathologist opens the case.

For the web and hybrid options no host, region or tenant isolation is described, and nothing says which processing leaves the laboratory in a hybrid setup. The privacy policy's transfer clause covers business contact data, not slide images.

Options stated with residency committed for the on premises route, and isolation and the processing path of the other options left open.

No pricing is published. The per APP packaging does imply a modular purchasing model, and availability through Agilent as the Visiopharm Diagnostic Package gives a second procurement route, but no rates, bundle structure, or per case economics are disclosed. Nothing says whether a license covers individual APPs or a portfolio, and because the clearances are narrow, coverage depends on exactly which algorithms are included.

Anatomic pathology biomarker assessment, with the diagnostic scope set by market and assay. The CE-IVD APPs cover breast (HER2, Ki67, ER, PR, metastasis detection), non-small cell lung cancer (PD-L1) and prostate, certified for diagnostic use in the EU and UK and not sold for diagnosis in the United States, where research APPs are offered. The APPs are validated to specific assays, including Agilent's under the alliance, so real coverage depends on the tumor type, antibody clone and assay a laboratory runs. Validation is published for lymph node metastasis detection in breast cancer at UMC Utrecht (CONFIDENT-B), and described by the company for PD-L1 and metastasis detection across three sites, though no validated scanners are listed. A large research APP library serves translational and preclinical work separately.

Coverage is named, and validation stands behind part of it.