The Diagnostic Power of Artificial Intelligence

In healthcare, new technology isn’t adopted just for the sake of it. It only gets adopted where a difference can be made in patient care, efficiency, or in improving access. There needs to be a positive, measurable outcome or real issue to be tackled; otherwise, it’s very hard to justify the expense within an already stretched environment. Innovation for the sake of being ‘cool’ has a hard time getting traction.

At Zebra we’re tackling head on the mismatch in global supply and demand of radiologists. There simply aren’t enough radiologists to handle the workload. Our mission, from the very beginning, has been to create software that can bridge that gap.

Our process is to use deep learning and convolutional neural networks, which is nomenclature for a particular software architecture that allows a computer to learn how to identify visual objects through repetition.

We’ve applied the principles of deep learning to medical imaging, feeding thousands, tens of thousands, and even millions of imaging cases into neural networks that we’ve developed in-house. By feeding the network enough examples of congestive heart failure or lung nodules or breast cancer, along with image samples that are benign, we’re teaching the software to recognise conditions that require attention. When the software has reached a level of accuracy that’s high enough to meet clinical thresholds, we release a new application into the market.

Over time, as our analytics engine improves and capabilities grow, we will be able to increase throughput, consistency, and efficiency for radiologists. We integrate our solution in a way that doesn’t intrude their workflow but is rather embedded into it, elevating patient care.

As a co-founder and CEO, I enjoy being part of a company that is providing benefit to people. I see medical technology as an amazing enabler, potentially providing high-end but affordable diagnoses for the billion or so people who are joining the middle class over the next decade and who otherwise might not have access to great healthcare. There is a lot of satisfaction in releasing an application that can provide an early diagnosis of breast cancer for women in developing countries and do that at very low cost. We’ve announced a $1 per scan business model which was previously unheard of. At that price point we’re enabling access to critical diagnoses for literally millions of people.

Right now we’re focused on growing the company and releasing great products. We’re seeing the market for AI and imaging grow at a very fast pace. We’re fortunate to be one of the leaders in that space so we’re putting our heads down to make sure we succeed in this market.

As leader of the business I constantly think about whether we have the right DNA in order to make the next leap. I think about whether we’re critical enough of ourselves and asking the right questions to make sure we’re seeing the market in the right way and being sure of where we can add value.

I often reflect on my time as a competitive ultra-marathon runner; there are a lot of similarities between running a startup and running an ultra-marathon. You need to be prepared for the long haul. You need to go into it knowing there’s going to be pain. You need to focus on logistics and planning, to understand how you’re performing to make sure you’re not burning yourself out too quickly or sandbagging and going too slowly. And you need assistance along the way – you can’t do an ultra-distance race without assistance, and a company is never about just one person.

Mainly it’s a mindset that accepts you’re entering something that is long and difficult, filled with uncertainty, stumbles and setbacks along the way. If you accept that from the beginning, knowing also that the destination is worth it, knowing that you are potentially helping millions of people, it makes the whole journey much easier to handle.