Dr Alex Novak Interview about X-ray Critical Care Suite

January 18, 2021

image/svg+xml

GE Healthcare is a global medical technology and digital solutions innovator, as well as being an NCIMI partner. 

They recently posted a video featuring Dr Alex Novak discussing the X-ray Critical Care Suite. 

You can watch the full video below.

Video transcript for X-ray Critical Care Suite: Dr Alex Novak

My name is Alex Novak I’m a Consultant in emergency medicine and ambulatory care at Oxford University Hospitals, and I’m the research lead for the emergency department. And in terms of NCIMI the chief investigator for GE Critical Care Suite project.

What is NCIMI?

NCIMI is The National Consortium of Intelligent Medical Imaging, which is an organization that tries to bring together a whole host of stakeholders around the development of AI in imaging. So clinicians researchers industry stakeholders patient groups, and then try to bring all the voices into the mix and try and accelerate the innovation pipeline for AI imaging from bench to bedside. Essentially from conception and into clinical practice.

What is the X-ray Critical Care Suite? 

This is one of the main projects that we’re working on at the moment, and I’m the chief investigator for this project. It’s is based around chest x rays into two key areas. 

One is image quality control so trying to make sure that you capture the best image you can. And also, in detection of varies pathologies, and to helping radiographers and clinicians to do that. The particular area of focus, in terms of pathologies, is pneumothoraces. so punctuated lungs on chest X-rays. That’s been the focus on the interest from me as an acute clinician because we’re routinely asked to detect this from a chest X-ray.

This is perhaps our main area of interest at the moment.

Can you tell us about the market evaluation at OUH?

Market valuation took place over about three months from October last year and we captured around 2,500 images. This is the first time I think this is the first time we’ve had AI algorithms embedded in acute clinical practice. And this was particularly based around the ICU and thoracic wards where often imaging is taken and we don’t get immediate radiologists reports. This is used by radiographers. Essentially the Critical Care Suite alerts you when things are the probability of pneumothorax and flags the reader to have a look and provides an overlay of where it thinks it might be.

Obviously, the clinical decision rests with the radiographer and clinicians themselves to act on it but the idea is that that should help speed up the workflow.

Who has benefited from x-ray critical care suite? 

There was a number of political stakeholders because, like most of the thing that actually is relevant to a whole host of things, especially radiographers, radiologists, ICU, cardiothoracic and Ed doctors like myself, all engaged in the project. 

We conducted a small informal reader study I guess to compare against senior clinicians. I’m still annoyed it beat me on one of the images. 

Broadly speaking it was a good chance for us to get engaged with AI in a broader sense.

It was the first time a lot of us encountered an AI imaging algorithm in a clinical practice. 

How can AI help in a clinical setting?

AI imaging has huge potential. A lot of clinicians in a lot of different sectors are very very excited about the prospect of having AI-assisted imaging in their practice. 

Two key ways it can have a big impact. One is Improving our current practices, enhancing our abilities to identify what we’re already looking for, whether it is pneumothorax or its air under the diaphragm which can indicate valve perforation. Whether it’s picking up cancer perhaps we weren’t necessarily looking for on the X-ray or CT.

The other way, of course, is extending those capabilities further, to think about imaging biomarkers which hadn’t really detected as being relevant. The NCIMI X-ray project covered both these aspects.

What are the barriers to adopting AI?

I think there is a lack of familiarity at the moment, in terms of where these may lie in the wider clinical thing there’s lots of excitement, but you have to think carefully about where the actual use cases are and what’s going to make a difference.

 I think there is sometimes a misconception amongst radiologists and radiographers that this is in some way going to replace clinicians I think that’s certainly not the case if anything its the opposite. A synergy between physicians in the algorithm.

 I think in terms of barriers. The other thing is, is the, I guess the hardware, making sure that needs the right level of tech is in place to cope with sort of the large data flows that are needed to do this. And I guess, in terms of other barriers is, making sure that all the systems are able to talk to each other, there’s a potential that I think a lot of fragmentation. If you have lots and lots of different competing ai algorithms, is one of the longer-term barriers might be trying to make sure that leads have a shared language and shared set of values and I guess a shared set of metrics for performance, and this will all develop over time, over the next few years or so.

What’s the timeline for adopting AI?

I’m always amazed at the rate at which this is developing. I think one of the exciting things for me about AI, it feels as though the pipeline is a lot shorter to getting a usable product, a usable piece of tech actually on to the shop floor. And, of course, a lot of the delivery systems and virtually ready we’re ready to use. X-ray is pretty universally done using digitized images. I think, I hope the pipeline is very short to this. it’s really within a few years we can’t start to see these algorithms being deployed, perhaps 4 or 5 years.  

Once the regulatory approvals are through. There’s a space for this in clinical practice already we have something called the red dot, in the UK, which is traditionally you’d put a red dot or a red sticker on X-ray if the radiographer suspected there was an abnormality on it. Where AI fits in, is essentially an intelligent red dot. An automatic red dot. 

So, there’s already space in the clinical practice, to respond to these kinds of triggers.

What’s next?

Next for NCIMI in the short term, we’re just finishing up plans for the reader study. And this is where we’re going to take the algorithm and see how it improves the performance of clinicians, not just at a senior level but right across, in about six different specialities. Consultants, to middle grade, to junior, reporting new x rays on the shop floor. It’s important to not just focus on the activities of consultant clinicians and radiologist who are always going to do really well against the algorithm, but also at an earlier stage, people who are working in the middle of the night in a busy department. So we can see what impact the algorithm will have on their ability to pick up Pneumothoraxes. We hope to have completed that by spring next year.
We’re excited about that, that’s going to run across the Thames Valley, so across 5 different hospitals, various stakeholders of NCIMI.

What’s exciting about AI and the future of NCIMI?

I think what’s exciting about it, is it’s potentially a powerful way of enhancing the performance of clinicians and, ultimately, that should translate into better patient care

All of this boils down to essentially what is going to be better for patients. By increasing the acceleration of a particular product or particular algorithms but, actually the scope for creating an environment where it’s possible to really accelerate the innovation and development in this area as a whole. 

It’s not just about the products that are in existence but the scope for how many more. We’re at the tip of the iceberg here, really. And that’s what’s really exciting for someone like myself a clinician-researcher, to be at that interface of these stakeholders. 

The thing about NCIMI that is exciting is that it brings all those other key aspects together so you’ve got the patient group you’ve got industry players. You’ve got Researchers in the academic background and the cynical sphere all coming together to really define the questions and to really set out of scope and where AI Imaging can go. 

image/svg+xml