Techcyte

The models are the engine of the clinical pathology suites, on a platform that also sells without them. Techcyte's clinical pathology products (parasitology, bacteriology, hematology, cervical cytology) are built around convolutional neural networks that find and pre-classify ova, parasites, cells and bacteria for a technologist to review, developed in-house with laboratories such as ARUP and Mayo Clinic. Fusion Anatomic Pathology, by contrast, is an image management system with a worklist, viewer, case routing and synoptic reporting, deployed across Mayo Clinic's pathology practices, which a laboratory can use without any algorithm, and which also hosts third party AI.

The model is the engine of core modules on a platform that carries other value.

The review workflow is described for the parasitology suite, with the negative side left open. For the wet mount product the algorithm finds ova, parasites and other objects and presents them for technologist review, grouped by class and sorted by confidence; technologists confirm objects and, where needed, prevalence and size, can request a consultation, and manually confirm positive samples on the microscope before reporting in their LIS.

The company describes the tool as AI-assisted screening and markets quick elimination of negative slides, with validation read times of 15 to 30 seconds. What a technologist does with a slide the AI calls negative is where the risk sits, and the company publishes no threshold, no proportion of negatives reexamined and no drift monitoring; the other suites' review designs are not described at this level.

Training data provenance is specific: a long standing Mayo Clinic collaboration providing access to the Safe Harbour dataset of more than 17 million deidentified slides paired with pathology reports. With the institutional source and the scale named, representativeness can be judged.

No accuracy, sensitivity or specificity figure is published for any classifier, which matters more given the Research Use Only status, since a laboratory cannot rely on regulatory review to have examined performance.

The host is named and the models are in-house, with the partner layer open. The platform page says Fusion is deployed on AWS, and that the clinical pathology algorithms are built in-house with laboratories such as ARUP and Mayo Clinic; the company has disclosed training on Mayo Clinic's Safe Harbor dataset of more than 17 million de-identified slides and pathology reports. Fusion also runs third party algorithms through a unified AI API and the Fusion Partner Program, and a partner that brought a research copilot to the platform was later acquired by AstraZeneca, so the parties that can process a laboratory's slides are set by the partner roster and change as partners change hands. No subprocessor list, partner list with processing terms, or customer data training position is published.

Host named and model layer in house, with the partner layer and its terms unenumerated.

Substantial adoption and one peer reviewed validation, with the flagship customer also a partner in the business. Fusion runs across Mayo Clinic's pathology practices, announced jointly; the clinical pathology algorithms were built with ARUP and Mayo Clinic; BD collaborates on the SureView cervical cytology system; and the company reports its veterinary segment profitable. The Journal of Clinical Microbiology published a retrospective clinical evaluation of the wet mount parasitology product used alongside operator review, reporting 96.62 percent positive and 93.33 percent negative percent agreement for protozoan and helminth detection, and the product page describes a single site study. Mayo Clinic is also an investor, a licensor of intellectual property, a co-developer of the platform and the source of a training dataset of more than 17 million slides, so the Mayo deployment is not an arm's length adoption decision.

Named deployments and published research short of independent, multi site validation.

General compliance claims, and a large training corpus whose handling is not described. The platform page says Fusion is HIPAA and GDPR compliant, built on AWS with role-based security. The company has disclosed access to Mayo Clinic's Safe Harbor dataset of more than 17 million de-identified slides and pathology reports for model development; whole slide image files routinely embed a photograph of the slide label carrying an accession number and often a name, so how the label and macro layers were stripped at that scale is the specific question, and the method applied is not published. Nothing states whether slides processed by customer laboratories contribute to model improvement, how long they are kept, or how they are de-identified. Safety engineering, meaning how a wrong output is handled as a safety event, is not published.

General assurances that leave retention, training use and de identification unanswered for customer data.

HIPAA compliance is claimed, with no business associate position or document behind it. The platform page and compliance line say Fusion is HIPAA and GDPR compliant; no business associate agreement, statement of business associate status, contracting entity or subprocessor list is published. Three relationships need separating: as a platform supplier to human laboratories the ordinary business associate analysis applies; partner algorithms that receive a laboratory's images would also handle protected health information, and whether they are covered by Techcyte's agreement or need their own is not stated; and the de-identified Mayo Clinic training corpus sits outside HIPAA, governed by the licensing agreement between the two organizations.

Compliance claimed without a business associate position or service document.

A recognized attestation is named, loosely, without the artifact. The platform page's security section says Fusion uses OAuth and role-based security, is built on AWS, and is "SOC2 compliant, ISO 13485 certified, HIPAA/GDPR compliant"; the site's compliance line lists FDA, CE, HIPAA, ISO, GDPR and SOC2. SOC 2 is an audit of how a company operates security, but "compliant" does not say which report type or period, and no report, scope, trust center or penetration testing statement is published. ISO 13485 is a medical device quality standard and HIPAA and GDPR are laws, so none of them is a security attestation; the compliance line mixes credentials, laws and regulators.

A recognized attestation named in the vendor's own material without the artifact.

A European position and a US position stated, with US submissions in process. The site states that Techcyte is CE-IVDD marked in Europe and for research use only in the United States, and announced CE marking under the IVDR for the Fusion AP digital pathology platform; the parasitology page labels the wet mount product research use only, not for diagnostic use in the US, with claims "not examined by the FDA".

The platform page's regulation section footnotes its compliance statement with "FDA submissions in process", which is a pathway stated and in progress; FDA's 510(k) database lists no clearance for Techcyte. The site's compliance line reads "Compliance: FDA, CE, HIPAA, ISO, GDPR, SOC2", but the "FDA" there is not a clearance, which the company's own labels rule out.

Per customer tuning is disclosed, and no evaluation by group is published. The company states that semi-quantitative ranges for the peripheral blood smear classifier can be adjusted per customer to match their protocols and populations. That is useful and honest to disclose, and it moves part of the clinical risk decision onto the laboratory: a threshold set for throughput behaves differently from one set for sensitivity, and no default rationale, guidance or validation requirement for customer tuning is published. Population adaptation is given as a reason without any analysis of performance across populations, scanners or sites, and no governance framework is named.

Tuning and adaptation asserted with no evaluation behind them.

Every AI finding is reviewed by a technologist, and nothing stands behind the output. In the parasitology suite the algorithm proposes objects for a technologist to confirm, and positive samples are confirmed on the microscope before reporting, so a wrong positive is caught by the technologist. The company lists a peer reviewed study of its wet mount algorithm in the Journal of Clinical Microbiology among its publications, which reports agreement for the AI with operator review; that measures the assisted workflow rather than the algorithm's standalone error, and the product page describes a single site study. The human products are research use only in the United States, so no regulator has reviewed their performance, which makes the vendor's own figures the only source a laboratory has. No warranty, indemnity or remediation commitment is published.

Review before reporting with nothing attached to the output.

A documented API and named production sites, with the connected systems unnamed. The platform page says Techcyte's complete, documented API integrates with laboratory information systems, EHRs, grossing stations and radiology systems, that Fusion supports all major scanner vendors, and that Fusion Anatomic Pathology offers a unified third party AI API for commercial and in-house algorithms. In the parasitology workflow technologists report results into their LIS.

Fusion runs across Mayo Clinic's pathology practices, and the clinical pathology algorithms were built with ARUP and Mayo. No LIS or EHR product is named, the API documentation is not public, and there is no marketplace listing.

A cloud service with its host named and its location unstated. The platform page says Fusion is browser based, built on microservices and deployed on AWS for global security and scalability, with customizable image storage and retrieval. No region, residency commitment, tenant isolation or on premises option is described, although the company sells in Europe under CE marking and in the United States. Partner algorithms run on the platform through a unified AI API, so which parties can reach a laboratory's images depends on the partner roster, which changes as partners are added or acquired.

A single hosted option with location implied rather than committed.

There is no public pricing. The company sells to human, veterinary, and environmental laboratories, with the veterinary segment reported profitable and human and environmental targeted for 2027. In the US the product is Research Use Only, and a product that cannot inform clinical diagnosis has a materially narrower internal use case than the same platform in Europe.

A wide scope, drawn precisely, with validation behind part of it. Techcyte serves human, veterinary and environmental laboratories; within human diagnostics the products are named individually (parasitology, bacteriology, hematology with a peripheral blood smear classifier, cervical cytology with the SureView system developed with BD, frozen section and organ donation suites, and Fusion Anatomic Pathology), and the regulatory boundary is stated beside them: CE-IVDD marked in Europe, with Fusion AP CE marked under the IVDR, and research use only in the United States. Validation is published for the wet mount parasitology algorithm (Journal of Clinical Microbiology) and described as a single site study on the product page; the company says the clinical pathology algorithms were built with ARUP and Mayo Clinic. The veterinary segment is the commercially mature one, and evidence there does not transfer to human diagnostics.

Coverage named with validation behind part of it.