Q&A: AI For All Radiologists

October 27, 2017
Whitney J. Palmer

Removing the cost barrier of AI in radiology.

Since 2013, Zebra Medical Imaging has been using artificial intelligence (AI) to streamline diagnostic imaging, and today it took a step to make its tools more widely affordable.

Operations began in early 2016, bringing a radiology assistant to the industry that helps identify clinical abnormalities before you even open a scan to read. Now, according to Zebra co-founder Elad Benjamin, the company has taken steps to make its technology a tool nearly every hospital or practice can use regardless of location or budget.

Through the new AI1 initiative, Benjamin said, Zebra will now make all of its current and future algorithms available for $1 per scan for all existing and new customers.

Diagnostic Imaging talked with Benjamin about this initiative and what it could mean for radiology worldwide.

DI: What are the details of the AI1 initiative?

Benjamin: We’ve tried to make things as simple as we can. We’re going to offer our entire analytics solutions for $1 per scan. It’s always been Zebra’s mission to provide accessible, transparent care in the health care arena. Now, we have a product that can actually help radiologists diagnose a number of different diseases automatically by using our AI technology. We decided to make it accessible to as many people as we can. No matter how many insights, how many images, or how many algorithms we have, everything will be $1 per scan.

DI: Why is this type of initiative necessary? What factors create this situation?

Benjamin: There are big differences between countries and their health care systems. We recently read a study that there are 4 billion people in radiology scarce zones. That’s measured by how close any individual is to a radiologist or care. If you think about it, that’s nearly half the planet that doesn’t have access. That touches on the main vision and purpose of why Zebra was founded.

A lot of these people can get to a place where a scanner is, but there’s not access to a quality professional who can read a report on the images. And, even in developed countries, radiologists are getting squeezed and pushed to procure more and more throughput without improving quality. So, we see a lot of burnout and fatigue. From our perspective, AI1 evens the playing field. We didn’t create complex typing schemes where hospitals have to think six times about budget and whether they should get the tools. If you think about it, the way radiology is going and the amount of information and data that are being generated - the only way to save radiology is to democratize it in some form. We feel that’s the only way.

In radiology, on the one hand, we have radiologists being pressured to produce more throughput without compromising quality. You have them reading more data. Even if they read the same number of scans per day - 50 or 60 - those same 50 scans generate four or five times more data than 50 scans from a decade ago just because the equipment is advanced and produces more output. Something is going to start to break. It’s very difficult for radiologists to spend equal amounts of time on each patient. They need assistance, and we make that happen.

DI: Is there anything institutions can do to better prepare themselves for using AI technology?

Benjamin: I think we’ve put a lot of effort into making this very simple to deploy and integrate the radiology assistant into the flow of radiology. In terms of preparation, hospitals or radiologists don’t need special technology or products or IT preparation. And, they don’t need to be willing to accept another product into their fold.

I think today, it’s more of radiologists being willing to accept that tools like Zebra’s can really empower them to become better physicians. They can provide more comprehensive reports. They can miss fewer findings. I think it’s more about that than any kind of technology preparation.

DI: Who or what type of institution will be the initial targets for AI1?

Benjamin: Interestingly enough, we’re having an equal amount of pull from two edges of the spectrum. One is in developing countries like India, Brazil, and China. We’re seeing a tremendous pull for these tools. They feel the pressure daily and see the lines of people waiting 2 to 3 weeks to get a diagnosis. From their perspective, this is a basic tool to help provide care.

On the other edge of the spectrum, we see physicians who are super busy in developed countries saying if this can help them not miss findings or create better reports or help a referring physician treat a patient better, then they’re all for it. We’re seeing a lot of willingness to adopt on both ends. At the end of the day, it’s radiologists everywhere who are our primary targets, and we’re offering all existing customers and new ones the opportunity to convert to this model.

DI: What will be the benefits to institutions that enroll?

Benjamin: Essentially, it provides algorithms for the entire suite of tools that we fold under the radiology assistant product. We have a certain number of algorithms available today, and that continues to grow. Today, we’re able to provide and analyze CT of the chest and abdomen, and we have algorithms in the pipeline for mammography, CT head, X-rays, and other things. The tool provides the radiologist a second pair of eyes for nearly every CT of chest and abdomen.

We also have algorithms for cardiovascular disease, liver, lung, or bone disease. We have specific tools that highlight the reads if they have found abnormalities. If so, where it has been found is pinpointed to help radiologists procure better output. The output is the report that helps guide patient care downstream, and the radiology assistant is geared to help create reports that are as comprehensive and quantitative as possible.

At a dollar a scan, anyone can afford it. The return on investment is very quick. So, there should be no budgetary constraints that block hospitals or providers from using it.

DI: What services are currently available through the program? What ailments are included?

Benjamin: There are several algorithms that are ready. There is a long list of ones that will be ready over the next six to nine months. Right now, we have compression fracture, bone density, liver density, emphysema, breast cancer, brain trauma, hypertension, and coronary calcium scores.

DI: What additional services are on the horizon to be included in AI1?

Benjamin: All our future algorithms will be folded into the same initiative. Six to nine months are just on the forefront of what’s going on.

DI: What type of impact are you hoping AI1 will have?

Benjamin: We have a relatively ambitious goal from my perspective. In five years, I want tens of hundreds of millions of completed automated diagnostics for any scans. We want to create software engines that analyze any scan with high accuracy. I know that’s a high bar, and that’s the vision of the company.

In the short term over the next 12 to 24 months, we hope that physicians will start using the tools. We hope to be able to show them how they can improve the quality of their work and improve the throughput of their work without compromising on anything. That’s really the goal in the short to medium terms.

DI: When does the initiative launch?

Benjamin: It will be available immediately. From our perspective, although the model that we’re talking about sounds like a financial model, from what we’ve seen, it comprehensively changes the way hospitals and clinics think about adopting AI. It removes barriers and allows them to step into the world of AI and medical imaging in a non-threatening way that helps them audit the technology.

A lot of hospitals and doctors are thinking about it and having a simple, affordable mechanism to experiment with will greatly impact the adoption among radiologists and others.

We’re blazing a trail here. I think this model has never been tried before. It’s transparent and open to everyone. I’ve been in the imaging space for a long time, and it’s never happened before. I hope that it prompts others to think about their products and design them in a way that they can be affordable and accessible because it will benefit society.

Related Content:

News | DI Executive