Q and A article

Where Old Meets New: How AI and X-ray Are Building a Future State of Radiology

Traditionally, when a radiographer or someone capturing an X-ray image suspected something suspicious in a lung X-ray they would print out their report and place what they call the “red dot” or a small red sticker on the report to indicate they’ve found an abnormality in the lung X-ray image.

Today, artificial intelligence embedded onto X-ray technology is serving as an “intelligent, automatic red dot” to let clinicians know that the patient may need attention more immediately. Critical Care Suite, the industry’s first AI algorithm collection embedded on an X-ray for triage, does just this by leveraging AI to prioritize critical chest X-ray review.

A key step to validating these technologies is collaborating across the industry and academia to show their value in the clinical practice. University of Oxford’s National Consortium of Intelligent Medical Imaging (NCIMI) enables this by bringing together stakeholders around the development of applying AI to medical imaging with the goal of accelerating AI innovation and bringing these technologies to clinical practice faster.

Last July, NCIMI and GE Healthcare announced their plan to apply this collaborative model to help develop software tools that can help analyze medical imaging, laboratory, and clinical data to flag critical information. The partnership is now performing research to understand how GE Healthcare Critical Care Suite on-device artificial intelligence can provide value to clinicians and prioritize the most critical patients in the ICU and Emergency Room.

Insights spoke with Dr. Alex Novak from the Oxford University Hospital Trust about NCIMI’s collaboration on AI* and his experience evaluating GE Healthcare’s X-ray Critical Care Suite.**

How can AI-assisted imaging help in a clinical setting?

AI imaging technology has huge potential. I think a lot of clinicians in a lot of different places are very excited about the prospect of having sort of AI-assisted imaging in their practice. I’d say there’s two key ways it can have a big impact. One is improving our current practices, enhancing our abilities to detect what we know we’re already looking for. Whether it's a pneumothorax or its air under the diaphragm, which can indicate about perforation, or whether it's picking up cancer that we weren't necessarily looking for on an X-ray or CT. And the other way, of course, is extending those capabilities a little further to think about imaging biomarkers we haven't really detected as being relevant for a pathology. The NCIMI project is looking to cover both of those aspects.

What are the barriers of adopting AI?

I think that is a lack of familiarity at the moment. There's lots of excitement, but you have to think carefully about where the actual use cases are and what is going to make a difference. I think there is sometimes a misconception amongst radiologists and radiographers that AI is going to replace clinicians. I think that's certainly not the case, and if anything, it's the opposite. It’s about synergy between clinicians and the algorithms they’re using.

Can you tell us about NCIMI’s study on applying Critical Care Suite?

It's based around chest X-rays and two key areas: one is image quality control, so trying to make sure that we capture the best image quality we can when taking a chest X-ray; the other is assisting in the detection of various pathologies and helping radiographers and various sorts of clinicians to do that. The particular area of focus in terms of pathologies is pneumothoraxes, which are punctured lungs on chest X-rays. That's been the big focus and interest for me as an acute clinician because this is something we are routinely asked to try and detect on chest X Ray. 

Can you tell us about the evaluation of Critical Care Suite at Oxford University Hospital?

The market evaluation took place over about three months from October last year and we captured around 2,000 images. This is the first time Oxford University Hospital really has had AI algorithms embedded in acute clinical practice. This was particularly based around the ICU and cardiothoracic wards where often imaging is taken, and we don't immediately get a radiology report. Essentially the Critical Care Suite alerts you when it thinks that there's a probability of a pneumothorax and flags the radiology reader to have a look at it. The clinical decision rests with the radiographer and with the clinicians themselves to act on it, but the idea is that Critical Care Suite should help speed up the workflow.

Which clinical specialties might benefit from Critical Care Suite?

There are a number of clinical stakeholders because, like many of our projects, it’s relevant to a whole host of clinical specialties. There were radiographers, radiologists, ICU, cardiothoracic, and ED doctors like myself all engaged in in the project. We conducted a small, informal reader study just to compare Critical Care Suite against some of the senior clinicians. I'm still annoyed that it beat me on one of the images. Broadly speaking, it was a good chance to get engaged with not just the product itself but AI imaging in a in a broader sense.

What’s next for the NCIMI study and its use of Critical Care Suite?

In the short term, we're finishing up plans for the reader study for the Critical Care Suite. 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 levels in about six different specialties, from the middle grade to juniors, who are actually reporting these X-Rays on the shop floor. It's important not to just focus on the activities of the most experienced clinicians and some radiologists who are always going to do well against the algorithm. We want people sort of earlier stage. They’re going to be working in the middle of the night in a busy department and we’re hoping to see what impact the algorithm potentially has to pick up pneumothoraxes.

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

I’m most excited about how enhancing the performance of clinicians can help translate into better patient care. Virtually all of this eventually boils down to better care for patients. By increasing not just the acceleration of these particular products or particular algorithms, but actually expanding the scope and creating an environment where it is possible to really accelerate the innovation and development in this area as a whole… that is very exciting.

To listen to the full interview with Dr. Alex Novak, click here.

*NHS Foundation Trust. Factors that should be considered by clinicians include cleared and approved product labelling and guidelines provides by medically sources organizations. Dr. Novak specializes in Emergency and Ambulatory Care and receives financial support from GE healthcare.

 **Not all products or features are available in all geographies. Check with your local GE Healthcare representative for availability in your country.