Blackford founder Ben Panter on his journey from astrophysical researcher to AI entrepreneur, the challenges of implementation across different healthcare settings, and the importance of teamwork.

You started your career as an astrophysicist. What led you to become involved in medical AI?

While I certainly enjoyed my time in astronomy, I recognised that the academic lifestyle wasn’t for me – galaxies don’t really give much back! My team and I had developed a technology that had application in many fields, and I set out looking for opportunities to apply it to. While we looked at various industries, the one that really appealed to me was healthcare – applying what we had to improve patient outcomes was both compelling and gratifying. We worked with a local hospital for a few years prototyping solutions to help ease medical imaging challenges in the lab and formally launched Blackford in 2010.

Blackford now has over 750 clinical sites using its platform. What are the main challenges you encounter when implementing solutions?

I think the real challenge is that every facility or clinic is unique: what works in one setting is unlikely to work in another site you implement – and the best way to learn is to experience solving these problems across a wide range of environments.

One dimension of this challenge is finding the combination of AI products that best fit the clinical and business needs of the customer. The reality is that the market for AI solutions is complex and evolving, and without careful analysis it’s easy to get swayed by whichever vendor shouts loudest rather than which best meets your needs. We’ve found that a given product might delight one customer, while for another it is rejected. Often this comes down to the environment of the customer – a busy radiology practice might put higher value on turn-around time, while an academic environment seeks the ability to gain deeper insight from a study, even though that may be more time consuming.

Another dimension is the technical environment the product needs to operate within. While a standalone hospital, in complete control of its imaging modalities and protocols, is reasonably straightforward, the reality is that those sites are rare. A large health system that has grown through consolidation may have several different IT solutions across different sites and needs to deliver a consistent service offering across all users. A large radiology practice may receive images from hundreds or even thousands of individual scanners, with no control over image acquisition protocols – it’s essential to be able to build these into a common workflow and deliver at the pace required for radiologists and referrers.

Operationalizing and scaling of innovation is a universal challenge in healthcare. How does your organization approach this?

As I mentioned before, a lot comes down to experience. We were the first to conceive and deliver an AI marketplace and platform in the market, and have more and deeper experience than any other vendor of deploying AI products through a platform into clinical workflows. Along the way we’ve learnt how to properly assess what solution best suits the needs of a particular customer, working with their clinicians on their data and measuring the benefit for their patients to ensure that both clinical and business goals are met. It’s essential that the process can be as efficient as possible – we’re heavily focussed on making sure that the impact of such a trial to the multiple stakeholders in the hospital is minimized

On the technical side, our original product required us to build much of the infrastructure that is needed to integrate AI into clinical workflow, and over the years we’ve encountered a huge variety of vendors, architectures and environments that allow us to recognise up front what is likely to work for a given setting, and how best to deliver it. Rather than a fixed technical asset, we have a collection of components and integrations that combined will deliver a customised solution, tailored to the needs of the site.

In what areas do you see the next big advances in health AI?

I think the underlying technology is now mature – almost commoditised. The big advances now are how the results of AI are integrated into clinical care. We learnt a long time ago that you can’t just throw an AI result at a clinician and hope they can make use of it – you need to provide the information at the correct time and in the correct context. That might be using AI to measure lesion sizes in a radiology setting that, when aggregated with other information, changes the choice of treatment for a patient in the neurology department; or perhaps a background analysis of archive images that uncovers that a patient requires intervention for osteoporosis. In both these examples it’s not the act of AI finding something that’s important, it’s the downstream impact of that discovery when appropriately integrated.

I think the next big advances will be formed as we connect the dots throughout the patient care cycle to improve both patient outcomes and overall healthcare economics.

What advice would you give someone starting their career in medical AI?

The field of medical AI is exciting, innovative, and fast paced – it has tremendous potential to improve the lives of patients and provide healthcare-economic savings that impact us all. The key is making sure that your efforts are productive and adoptable.

On that issue, I think there is no substitute for on-the-ground experience. Experience of solving real problems with a clinical site, and the rich complexity of the challenges they face, is the only way to really understand how the solutions that you develop will help – or hinder – clinical users. From my own career, it was sitting in reading rooms, modality suites and IT offices where I learnt the reality of the challenges faced by healthcare.

Who’s been the biggest influence on your career?

I think the biggest influencers have been the teams we’ve worked with in our early deployments on the hospital side – the clinicians who had the patience to help us understand the real clinical challenge and how it was addressed by various specialists, the technologists who spent time sharing their workflow and the IT teams that helped us learn how to either address or work around the numerous challenges a real-world radiology department faces. We’re in debt to them for the quality and breadth of the solutions we are now able to offer.

If you could return to the past, what would you change or do differently?

Honestly, I struggle to think of what I would change – there have been areas where our product didn’t quite work first time out, or deals that we didn’t win for whatever reason – but every experience, positive or negative, has contributed to both our understanding of our products and the wider healthcare context that we operate in. This has made us better able to meet our customers’ needs, and ultimately delivered the business that we have today.

Dr Panter is a speaker at AIMed’s Global Summit, taking place live and in person in Laguna Niguel, CA on January 18-20, 2022. Book your place now.