Aignostics applies AI to tissue analysis, pathology research and visual case search

Aignostics develops AI for analyzing digital pathology images and extracting information about tissue, cells and their spatial relationships. Its offer spans HETA tissue analysis, Atlas foundation models, biomarker work and PathoSearch visual retrieval. These products share a pathology focus but have different access and usage boundaries. In particular, the current HETA Clinical and PathoSearch pages identify research-use-only products; their names and workflow examples should not be read as general diagnostic authorization.

01 / ProductTissue analysis and case retrieval are separate jobs

The life sciences offering combines foundation model licensing, HETA analysis, pre-analyzed datasets and custom biomarker services. A biopharma team can therefore approach the company at different levels: it might want measurements from a study, a modeling foundation for its own development or a collaboration around a specific biological hypothesis. Those are different projects, with different evidence and commercial requirements.

HETA works on whole-slide hematoxylin and eosin images. Its public description includes quality assessment, cell classification, tissue segmentation and spatial readouts. The output is richer than a single case label: it can describe where features occur and how they relate. That makes it relevant to research where the arrangement of tissue and cells matters alongside their overall abundance.

PathoSearch performs a different task. A user supplies a region-of-interest screenshot, and the system retrieves diagnosed reference cases with similar morphology. The returned references can be inspected rather than accepted as an unexplained generated answer. The page currently offers a waitlist and states that the product is intended for research use only. A login link is not evidence of open, unrestricted enrollment.

HETA Clinical introduces workflow-oriented products for slide quality, tumor detection, mitosis measurement and tumor cellularity. Although the page discusses pathology laboratory tasks, it explicitly says the products are not for diagnostic procedures. That qualification is central to the current offer. A research evaluation can investigate workflow utility without presenting these products as cleared tools for routine diagnostic decision-making.

02 / AudienceResearchers should choose the level of service they need

A biopharma researcher examining tissue cohorts may want standardized measurements that can be connected to clinical or molecular data. A computational pathology team may want to license a foundation model and develop its own application. An academic researcher may be interested in the company's research access routes. Each should state whether the desired output is a dataset, a model capability or a validated biomarker project before requesting access.

PathAI provides an adjacent comparison in computational pathology. Owkin offers another perspective on AI and biomedical research. The meaningful comparison is the particular analytical task, data access and permitted use. Similar pathology terminology does not establish that the companies supply equivalent outputs, have identical validation scope or allow the same deployment model.

A pathology department seeking a production diagnostic tool needs a more specific answer than the broad product category supplies. The reviewed Aignostics pages do not authorize diagnostic use of HETA Clinical or PathoSearch. Likewise, a developer cannot assume that academic access to a model permits commercial use. The proposed projects below remain research evaluations, with the appropriate agreement and data permissions established before any upload.

03 / WorkflowA proposed HETA study preserves the link to each slide

Choose a bounded research question and a dataset whose provenance and permitted use are understood. For example, a team might explore whether a spatial tissue feature varies across previously characterized research cohorts. Define the relevant feature and cohort criteria before processing. This keeps the project from becoming an unfocused search through many measurements for an attractive association that was never independently evaluated.

Inspect input quality with a pathologist and identify which organs, preparation methods and image formats are within the supported application. The platform documentation describes whole-slide inputs and standard result formats, including measurements and visual outputs. Match the actual installed application and version to its current requirements. A format being accepted does not establish that every tissue type or acquisition condition is appropriate for the analysis.

Run an initial approved set through the agreed access route, download the results for authorized retention and preserve the relationship between each source slide, analysis version and output. Review visual results alongside quantitative measurements. If a region is excluded by quality control, examine whether the exclusion changes the biological question being asked. A seemingly complete table can conceal a different analyzed tissue area from the one the researcher originally intended.

Next, inspect disagreements or surprising results with the relevant specialists. Record whether the issue appears related to staining, scanning, tissue composition or the interpretation of a metric. Keep those explanations separate from any final conclusion about the cohort. A proposed pilot should reveal where automated measurement is useful and where manual review remains important, rather than suppressing difficult examples to improve an aggregate summary.

For a separate PathoSearch evaluation, use approved research material and compare retrieved reference cases with the source region. Examine why a visually similar result may or may not be biologically relevant. Morphological similarity is a retrieval signal, not a definitive diagnosis. The useful research question is whether inspectable cases help the expert's reference work within the permitted scope. Sequenced has not performed either evaluation or measured product performance.

04 / PricingBusiness access and early access have different terms

The stable platform documentation says organizational access requires a formal business agreement and registration by Aignostics. It also describes organizational email requirements, mandatory two-factor authentication and an administrator role. That makes the hosted analysis platform a managed access route, rather than an anonymous API that anyone can immediately use with a personal account. Confirm the current access process for the particular offering.

The public life sciences page invites inquiries about a free pilot, but that invitation does not establish unlimited free processing or a universal entitlement. Scope the dataset, application, output delivery and permitted use in writing. Foundation model licensing and custom biomarker work should be assessed as separate commercial discussions. Platform charges are based on processed slides, with rates defined in the agreement; the Test Application is free for registered users. Canceled runs incur charges only for slides already processed. No general numeric per-slide price was verified.

The PathoSearch launch announcement states that the service is free during early access. The current product page offers a waitlist. Treat those as two distinct conditions: no charge during the stated phase does not guarantee immediate admission, ongoing free availability or permission to use the service diagnostically.

A realistic comparison should also consider the form of the purchased result. Receiving analysis outputs, licensing a model and commissioning biomarker development leave different responsibilities with the customer. Ask what can be retained, reproduced or incorporated into later work, and how application updates affect a continuing study. Those questions define the value of access more clearly than a headline claim that one route is cheaper or more flexible.

05 / DistinctionsSpatial outputs make the model's work inspectable

Aignostics' focus on tissue and cell relationships is a substantive distinction from tools that only predict a single label for a slide. Researchers can formulate questions about where features occur and how regions differ. The editorial inference is that inspectable spatial outputs can make a computational result easier to interrogate, provided the output definitions and limitations are understood by the people conducting the study.

The combination of foundation models and application products also offers different entry points into the same technical field. A team that wants a research readout may not need to build its own model, while a team developing a new application may want more control than a hosted result supplies. Aignostics' offer makes that choice visible, but the appropriate route depends on the buyer's expertise and research objective.

The company background describes its origins at Charité and its interdisciplinary team. That helps explain the company's pathology specialization, without proving any particular application's performance. The product pages publish comparative validation claims. Those claims should be read against the stated tissues, reference methods and datasets rather than repeated as a universal statement that the software exceeds human performance.

06 / QuestionsProduct names cannot override explicit usage limits

The most consequential current qualification is the research-only status on HETA Clinical and PathoSearch. A product may be designed around a clinical laboratory task while remaining unavailable for diagnostic procedures. Organizations interested in future clinical adoption should request the precise applicable documentation and availability for their jurisdiction. A development pathway toward companion diagnostics is a roadmap and qualification process, not proof that a current research output is already a diagnostic product.

Another open question is how a proposed cohort relates to the validated scope. Broad coverage across tissues does not mean every specimen type, stain, scanner and disease context is equally represented. Ask for the relevant supported-input and validation details for the exact application. If a research question depends on an unusual population or preparation, that gap should be explicit before interpreting the output.

Finally, determine how exploratory findings will be evaluated independently. A large number of quantitative readouts can produce many plausible associations. A team should preserve the distinction between discovering a hypothesis and validating it in separate material. That distinction is a research-design responsibility; access to an advanced foundation model does not remove it or establish that a newly observed signal will generalize.

07 / DecisionMatch the access route to a research question

Aignostics is a relevant AI company because its products connect specialized models to concrete pathology research tasks. The best initial project has a defined tissue dataset, a narrow analytical purpose and a review process capable of investigating unexpected outputs. Choose the hosted analysis, licensing or collaboration route according to the result the team needs, while keeping usage permissions explicit.

For reference retrieval, evaluate PathoSearch within its early-access and research boundaries. For tissue analysis, establish supported inputs and inspect both visual and quantitative outputs. For a proposed diagnostic application, require the specific evidence and authorization that the present research label does not supply. That approach gives the company's technical strengths a fair evaluation without turning future possibilities into current capabilities.