Brainomix: AI Imaging Analysis in Drug Development

Dr. Michalis Papadakis, founder and CEO of AI Imaging company Brainomix shares how the company’s AI Imaging software is supporting drug development in both retrospective analysis of trial data to more precisely identify clinical meaningful impact, and in prospective trials as AI biomarker endpoints.

Key Points – Brainomix has expanded its AI imaging services to support the clinical development of numerous novel therapies; both in stroke as well as lung fibrosis

Brainomix, a company founded in Oxford over 15 years ago as a spin-out from the University, began its journey by developing AI-powered imaging solutions for acute stroke assessment. By now, the Brainomix 360 Stroke platform boasts an extensive clinical footprint, supporting physicians in more than 350 hospitals across 20+ countries to detect and characterise stroke faster and more reliably, thereby significantly improving patient outcomes.

Since its inception, Brainomix has expanded its services to support the clinical development of numerous novel therapies; both in stroke as well as lung fibrosis, which are diseases with high mortality and significant unmet medical needs, with an opportunity for improved diagnosis and management through accurate imaging.

With data-trained and robustly validated biomarkers, a flexible service offering, strategic partnerships with leading pharmaceutical and device companies as well as imaging CROs, Brainomix has positioned itself as a leader at the forefront of digital innovation in the drug development sphere.

Q: Brainomix originally focussed on clinical adoption of AI imaging software in stroke. For the past few years, you have steadily increased your presence in the clinical development space. How would you say your clinical success has translated to clinical trials?

A: Yes – our foundation was always focussed on optimised care and outcomes for patients through improved medical imaging analysis. This remains our overall purpose, but this is clearly aligned with Life Sciences organizations, and working more closely with these partners through their entire development lifecycle has allowed us to realize our vision more effectively.

Our Life Science partners can also benefit from the data and evidence we have built over the years in the real-world clinical setting using our software.

For example, the success of our clinically implemented stroke software was recently highlighted by a study published in the Lancet Digital Health, which showed that Brainomix Stroke AI technology enabled more patients to access life-saving stroke treatments, much earlier.

Use of our stroke AI tool recently yielded successful results through our analysis of data from Argenica’s Phase II trial of a neuroprotective agent in acute ischemic stroke: our digital biomarkers were able to confirm a significant treatment effect in severe stroke patients, supporting Argenica’s continuation of its development program, and providing the opportunity for optimised future trial design in regards to specific patient selection, characterisation and efficacy determination.

We recently also announced two additional partnerships with Boehringer Ingelheim: a prospective multicenter study ( PROGRESS-PPF ) to enable earlier diagnosis of progressive pulmonary fibrosis; and another on their DROP-FPF study, a Phase III trial investigating the efficacy of Jascayd ® in people with a family background of pulmonary fibrosis. This is a pivotal moment for the field as this study marks the first time a quantitative, digital HRCT biomarker is used as a co-primary endpoint in a Phase III pulmonary fibrosis trial, highlighting our partners’ confidence in our imaging biomarkers, and the impact AI imaging can have for under-served therapeutic areas.

Q: Where do you see the greatest value of AI imaging analytics in guiding clinical development?

A: Many trials rely on accurate detection of clinically meaningful changes in relevant tissue – but studies show that even highly trained specialists struggle to recognize subtle signals of disease in CT images, and opinions often vary, making clear diagnosis very challenging. Our software, which has been trained on thousands of diverse real-world cases, provides quantitative assessment of even subtle changes, and at much earlier timepoints than the human eye. This reduces inter-physician variability, and through its reliable, highly sensitive biomarkers provides greater statistical power than traditional methods. For instance, our Brainomix 360 Stroke platform is able to make quantitative assessments of net water uptake as a characteristic of the severity of stroke damage, and the complex blood flow analysis we are able to do with perfusion imaging in stroke is not possible by radiologist visual assessment alone.

Q: One of your services includes retrospective analysis of historical data from completed clinical trials. How do you provide value in this area?

A: In clinical trials for example, for partners such as AstraZeneca, we performed a post-hoc analysis of their previously completed clinical trial in IPF; demonstrating our ability to extract new insights from existing data. The ability to almost go back in time with more advanced modern approaches is something that is very exciting now we have these types of AI technologies.

Any post-hoc findings of course need to be translated into prospective pre-specified analyses, which makes our work with companies like Argenica more impactful, where our blinded analysis and imaging quantification of their stroke study allowed them to make critical decisions regarding the viability of their asset.

Q: What are your other services, particularly for prospective trials?

A: We have a comprehensive suite of capabilities to support a long-term partnership for sponsors across all stages of drug development, and also in the delivery of treatments to patients in the real-world.

Where we can provide significant value for drug development is in the preparation and design of studies, using inputs into statistical design, selection of endpoints and thresholds for enrollment criteria. The high repeatability and sensitivity of our biomarkers also allows for reduced cohort sizes, which is particularly relevant for rare disease indications, and together this approach allows us to position our sponsor partners to have confidence and clarity in their results.

In prospective studies, we can unlock unique opportunities through the use of our regulatory cleared software and biomarkers for clinical use, allowing us to provide relevant sites with the required software directly to inform enrollment decisions, including with automated flagging of eligible cases to investigators.

The use of clinical-grade software distinguishes Brainomix from other AI-imaging tools in the field, and the availability of Brainomix software in routine clinical practice also offers the unique opportunity to connect clinical trial findings with market access strategies for new and existing treatments in the realworld setting.

Q: How does Brainomix ensure its analyses are aligned with regulatory standards?

A: Given our deep roots in clinical practice, we are well-used to working in a secure and heavily regulated environment. For real-world clinical use, regulatory agencies consider our AI software as a medical device, and as such, we have an extensive Quality Management System (QMS) framework which is needed for us to achieve and maintain authorisation for provision of our software. To that end, our software and biomarkers are FDA cleared, and CE marked for clinical use.

In the context of clinical trials, these clearances by regulatory agencies have allowed us to give our partners confidence in our biomarkers, and we are seeing increased adoption in clinical trials, and the agencies appear to be open to further inclusion of these types of assessments to support clinical development.

In order to continue to make progress, we are partnering with like-minded clinical trial sponsors which are incorporating imaging endpoints into their trials. In this way we are working together to support their clinical development, but also to generate data which can support improved design for future trials across the field.

Q: What are Brainomix’s strategic priorities over the next few years?

A: Ultimately, our vision will always have patients at the heart. Everything we do strives to improve patient outcomes, whether it be through adoption of our software in a clinical setting, or by supporting the development of treatments that address unmet medical needs. To this end, we are aiming to deliver value in existing and new partnerships with pharmaceutical companies developing novel treatments, and to provide them with tailored support for their clinical development.

We’ll also be extending our current partnerships with clinical trial sponsors and CROs into new disease areas, and we are always open to working with new partners which are looking to integrate smart AI imaging into their development and service strategy.

We see the value we add in clinical trials as a continuum to what translates into clinical practice. The imaging criteria used in clinical trials define the guideline criteria when these new treatments translated into clinical adoption.

In the future, we aim to focus even more on clinical data integration for our biomarker development; this will boost the refinement of prediction and composite markers and it will expand our role in the digital innovation of the clinical trial sphere, for example through integration of digital control arms for trials.

Q: What should clinical trial sponsors consider when integrating AI imaging into clinical development programmes?

A: While we can help at all stages of clinical trials, early integration yields the greatest value. This allows imaging biomarkers and decision-support tools to be aligned with desired trial objectives, endpoints, and patient stratification strategies from the outset. Sponsors should also consider AI solutions that are thoroughly clinically validated, transparent, and developed under appropriate QMS processes.

To optimize the use of quantitative AI imaging tools, sponsors should also consider longitudinal imaging, with baseline, early, and late-time follow-up timepoints. This allows ideal tracking of disease progression, and highly sensitive AIbiomarkers can detect changes even at early timepoints, giving an indication of treatment trends throughout study duration.

The other consideration is that AI is a tool which at this stage should still sit within a human-reviewed environment and as one component of an overall development strategy. Sponsors should ensure that they work with a partner that understands the strengths and limitations of any algorithms, biomarkers or software that is incorporated into drug development strategies.

Finally, sponsors should consider scalability of AI solutions, including whether they have sufficient evidence, readiness and regulatory status for adoption and translation into clinical use.