Aiosyn

The models are the product. Every Aiosyn offering is AI image analysis: slide quality control, mitosis detection, kidney biopsy outlining and scoring, ready made preclinical readouts, and custom models built on a client's images. Aiosyn sells no scanner, viewer or laboratory service that would stand without them.

Ask which model version produced a given readout, and whether models are retrained for your stains and scanners.

Pathologist review is built into the laboratory products, but no threshold is published. Aiosyn Mitosis marks detected mitoses in the slide viewer before case review, so the pathologist confirms them. AiosynQC flags slides with focus, fold, scan or color problems for rescanning or review. NephroPath outlines and classifies structures in the viewer for the pathologist to assess. Preclinical readouts were validated against pathologist scoring.

What is missing is the operating point: the confidence at which a mitosis is marked, the quality score at which a slide is flagged, and what happens when a pathologist disagrees with the output.

Ask for the detection thresholds, whether they can be tuned per laboratory, and how disagreements are logged and fed back.

Aiosyn describes what each product detects, but identifies nothing about the models. It names the structures and measurements each product outputs, and builds on more than 20 years of computational pathology research at Radboudumc.

No model architecture, training data, version or update practice is published for any product.

Ask for the model architecture, the training data by source and size, the version in use, and how updates are announced.

Aiosyn builds its own models, from Radboudumc research and its own development, and no outside model provider is named or implied. Cloudflare is named as a processor for the website under a data processing addendum.

The infrastructure and subprocessors that handle customer slide images are not described. Nor is the role of partner platforms such as Sectra and HistoWiz in processing them.

Ask for the subprocessors and hosting that handle slide images, and whether partner platforms run Aiosyn's models in their own environments.

Aiosyn publishes named preclinical work with figures and method, and nothing for the laboratory products. In a validation with TNO on 80 aortic root cases, Aiosyn reports that its AI lesion area readouts matched expert pathologist scoring with a Pearson correlation of 0.96. A diabetic kidney disease study with Physiogenex, published in the European Journal of Pharmaceutical Sciences (2026), used NephroPath to detect the effect of dapagliflozin in rats. Aiosyn's academic partners include Radboudumc, Amsterdam UMC and Leiden University Medical Center.

No performance figures are published for AiosynQC, Aiosyn Mitosis or NephroPath on human slides. No study shows an effect on laboratory workflow or diagnosis.

Ask for agreement with pathologists for mitosis counts and kidney scores on human biopsies, and for the detection rates of AiosynQC on your scanners.

Aiosyn's privacy terms are general, and do not answer the questions about images or training use. Its privacy policy says data is kept no longer than necessary, and that product support exchanges may be kept for up to ten years for post market surveillance.

Nothing published says how long customer slide images are kept, whether they are used to train or improve models, or how training data from partner hospitals was de identified. Custom models are built on a client's images, with no statement of who may reuse them. No safety engineering for the models is published.

Ask whether client or laboratory images train any model beyond the client's own, how long images are kept, and how partner training data was de identified.

Aiosyn's privacy policy reaches part of the service, and takes no position on patient data. The policy of Aiosyn B.V., effective 24 August 2026 and written under the GDPR, covers website visitors and its preclinical research services. It names Cloudflare as a processor under a data processing addendum, with standard contractual clauses for transfers outside the European Economic Area.

It does not address patient slide images processed for laboratory customers and says nothing about HIPAA. No data processing agreement for customer images is published.

Ask whether Aiosyn acts as a processor for laboratory slide images, for the data processing agreement, and for a business associate position if you are a US customer.

Aiosyn publishes no independent security attestation. It holds ISO 13485 certification from BSI, which audits its quality system for medical device development, not how it runs information security. No ISO/IEC 27001 certificate, NEN 7510 certificate, SOC 2 report, penetration testing statement or trust center appears on its site.

Ask for any information security certification, its scope and date, and how laboratory images are protected.

Aiosyn makes no device claim and states each product's position, though one product is presented beyond its label. It states that AiosynQC is not a medical device under the EU IVDR or UK MDR 2002, and is for research use only in the United States. Aiosyn Mitosis is for research use only. NephroPath is for research use only, not CE marked and not for diagnostic procedures. Aiosyn holds ISO 13485 for its quality system, and the FDA's databases list no Aiosyn device.

Aiosyn Mitosis is presented on the clinical page to diagnostic laboratories as marking mitoses before routine case review. That sits uneasily with its research use label, since the mitotic count is part of breast cancer grading.

Ask how Aiosyn Mitosis is kept out of diagnostic grading in practice, and whether a CE IVD or FDA submission is planned for it.

Aiosyn describes validation with experts, but publishes no results compared across groups. It works with a scientific advisory board of three named nephrologists and pathologists for NephroPath, validates preclinical models against pathologist scoring, and runs a pathologist network.

No results are published by scanner, stain, laboratory, species or patient group. No governance framework or monitoring after release is described.

Ask for performance by scanner and stain protocol, and how models are monitored when a laboratory changes either.

Pathologists can review each output, but nothing stands behind it. Detections and outlines are shown in the pathologist's viewer for review, and quality control flags slides rather than removing them, so a pathologist can challenge each output.

No error rate for the laboratory products, warranty, remedy or correction route is published, and the products are labeled for research use.

Ask for false positive and false negative rates for mitosis detection and quality control, and what Aiosyn commits to when a preclinical readout is found wrong.

No clinical record integration exists. The products reach laboratories through named image platforms instead. NephroPath is listed on Sectra's Amplifier marketplace and connected to HistoWiz's PathologyMap. Aiosyn has announced real time kidney biopsy processing with Pramana's scanners, and names Pathomation and Argos among its integration partners. Results appear in the pathologist's existing viewer, flowing one way from the AI into the image platform.

Ask which of these integrations run in production at clinical laboratories, and whether results can be written back to the laboratory information system.

Aiosyn runs its analysis as a service, and where it runs is implied, not stated. It analyzes preclinical images and returns the results, and the laboratory products run through partner viewers and platforms. Aiosyn is a Dutch company, and its privacy policy names standard contractual clauses for processing outside the European Economic Area.

Nothing published says where customer images are processed or stored, which cloud is used beyond Cloudflare for the website, or whether the products can run on premises.

Ask where slide images are processed and stored, whether any processing leaves the EU, and whether the laboratory products can run on premises.

No price is published, but how each route is sold can be worked out. Drug company and contract research customers buy analyses from existing models, with readouts in about a week, or custom model development on their own images. Laboratory products reach customers inside viewers and platforms, including Sectra's Amplifier marketplace and HistoWiz's PathologyMap.

No rate per slide, per study or per license is published.

Ask for the price per preclinical study or per slide, the cost of custom development, and the license terms through each marketplace.

Aiosyn names coverage product by product, with validation behind part of it. Preclinical analyses cover atherosclerosis, brain, liver and kidney tissue, with a TNO atherosclerosis validation and a published rat kidney study. Laboratory products cover quality control of H&E and immunohistochemistry slides, mitosis detection in breast cancer, and kidney biopsy assessment including the Banff ci and ct scores. Those are for kidney pathology laboratories and transplant centers, under research use.

Aiosyn sells preclinical readouts and custom models to drug companies and contract research organizations, which is why it is also listed under drug discovery. No validation on human slides is published for the laboratory products.

Ask which species, stains and scanners each preclinical model is validated for, and for validation of the laboratory products on human tissue.