Summary

• PXN Ventures led the £10 million Series A, with existing investors Northern Gritstone and Parkwalk participating.
• Adsilico's Virtual Chimera platform creates synthetic anatomical populations from clinical scans and applies physics-based simulations to medical implants.
• The company plans to extend its cardiovascular work into neurovascular devices, with clinical validation and regulatory acceptance remaining central to adoption.

University of Leeds spinout Adsilico has raised £10 million to expand technology that allows medical device manufacturers to test implants against virtual patient populations, combining generative artificial intelligence with engineering simulations to identify potential problems earlier in development.

The Series A round, disclosed on 26 October, was led by PXN Ventures, with existing investors Northern Gritstone and Parkwalk participating. The investment follows a £3.5 million seed round completed in 2024 and will support further development of the company’s Virtual Chimera platform.

Adsilico has concentrated initially on cardiovascular devices, including heart valves and stents, and intends to extend its work into neurovascular applications involving implants used to treat conditions affecting blood vessels in the brain.

The company is developing simulation tools intended to supplement established testing and clinical evidence rather than eliminate the need for trials involving patients. Its commercial proposition depends on whether virtual experiments can reliably reveal design risks before manufacturers commit to more expensive development stages.

Anatomical variation creates design uncertainty

Medical implants interact with the anatomy and physiology of individual patients, whose characteristics can vary substantially. A device that performs satisfactorily within one anatomical configuration may encounter different conditions when used in another.

Manufacturers assess these effects through laboratory experiments, engineering calculations, preclinical studies and clinical trials. Each method contributes useful evidence, but practical constraints limit the number of variations that can be examined before a product reaches wider use.

Clinical trials may involve patient populations that do not fully reflect the diversity of those who could eventually receive an implant. Differences in anatomy, sex, age and underlying conditions can influence device performance, while some groups may be difficult to recruit in sufficient numbers.

Adsilico seeks to investigate that variation by creating large sets of synthetic anatomical models derived from anonymised clinical scans. Its simulations then examine how devices behave within those virtual environments.

For a heart valve, engineers may want to understand how differences in cardiac anatomy influence device placement, blood flow or the mechanical behaviour of surrounding structures.

A stent or other vascular implant presents different technical questions, including how the device conforms to the geometry of a vessel and how its presence affects fluid movement.

Virtual testing allows developers to examine additional scenarios without requiring a separate physical experiment for every anatomical configuration. However, the results are useful only where the models and engineering assumptions accurately represent the conditions being studied.

Generative AI needs physical evidence

Virtual Chimera combines generative AI with physics-based simulation. The AI component is intended to create realistic variations in anatomy, while the engineering models calculate how devices and biological structures may interact.

Producing a plausible anatomical image does not necessarily establish that the resulting model is physiologically representative.

Developers need to understand whether synthetic populations reproduce the important characteristics of real patients, including combinations of anatomical features that may be uncommon but clinically significant.

Training information also influences the range of outcomes a generative model can represent. If particular patient groups are under-represented within the underlying scans, the resulting synthetic population may reproduce rather than correct existing gaps.

Physics-based simulation introduces its own assumptions about material properties, blood flow, tissue behaviour and device interaction. These assumptions must be tested against reliable experimental or clinical evidence.

An apparently precise simulation result can still be misleading if important physical conditions have been simplified or omitted.

Manufacturers therefore need to distinguish between predictions that have been validated for a defined purpose and exploratory findings that require further investigation.

A simulation may be suitable for comparing alternative designs while remaining insufficient to support a formal conclusion about clinical safety.

Adsilico’s platform is intended to help engineers make those comparisons earlier, when changing a product design may be less expensive than after tooling, testing and regulatory submissions have progressed.

Neurovascular devices add new complexity

Adsilico plans to develop applications for neurovascular implants alongside its existing cardiovascular work.

Although both fields involve blood vessels and implanted devices, the relevant anatomy and clinical conditions differ substantially. The geometry of cerebral blood vessels, the behaviour of blood flow and the consequences of an unsuccessful intervention can create distinctive engineering challenges.

Devices used to treat aneurysms or other neurovascular conditions must be evaluated against the requirements of those applications rather than assessed through assumptions derived entirely from cardiac implants.

Extending simulation technology therefore requires new datasets, appropriate physical models and validation against relevant clinical or experimental observations.

The company has identified neurovascular device manufacturers among its existing development partners, providing a starting point for investigating those requirements.

However, an established relationship with a device developer does not prove that the resulting simulations have been accepted as a substitute for evidence required by regulators.

Regulatory use depends on validated models

Computer modelling and simulation already contribute to medical device design, while regulators and standards organisations continue developing approaches for assessing when simulation results can support safety and performance evidence.

Acceptance depends on the intended use of a model, the consequences of an incorrect prediction and the evidence supporting its reliability.

For manufacturers, the potential commercial benefit lies in selecting better designs for further testing, identifying vulnerable configurations and reducing the number of unnecessary development iterations.

Yet simulation also creates costs associated with data preparation, specialised computing, software validation and expert interpretation.

These expenses must be considered when comparing virtual testing with established methods. A model that requires extensive bespoke engineering for each product may offer different economics from a platform capable of supporting repeated, validated workflows.

Documentation and traceability are important because an engineering decision may depend on a particular model version, dataset and set of assumptions. Developers need records that allow findings to be reproduced and reviewed as a product moves through its development process.

Where generative AI is used to construct synthetic patient populations, manufacturers must also be able to explain how the populations were formed and whether they reflect the intended patients.

Financing supports commercial development

Adsilico was established in 2024 following research at the University of Leeds. Its technology draws on academic work in computational modelling and medical device engineering, supported by commercial and clinical expertise.

The new investment provides resources to develop the platform and expand its engagement with manufacturers. However, the company has not disclosed recurring revenue, profitability or enough commercial information to establish the average value of customer deployments.

Investors are backing the prospect that simulation can become a more routine part of implant development, particularly where manufacturers need stronger evidence about performance across varied patient anatomies.

The technology’s commercial relevance will depend on whether its results help customers make better decisions and whether those results can be incorporated into established engineering and regulatory processes.

Adsilico’s £10 million round supports the next phase of development, while broader adoption will require reproducible evidence that virtual patient populations improve device evaluation without overstating what simulation can establish about clinical outcomes.