At the 2022 Charing Cross (CX) International Symposium (26–28 April, London, UK), Tom Carrell (Cydar Medical, Barrington, UK) delivered a Podium 1st presentation titled “Making endovascular aortic repair outcomes more predictable: Artificial intelligence [AI] takes on a 20-year-old challenge”.
“Given any individual patient with some anatomic complexity, do we really know—particularly with the ever-expanding range of treatment options— what is going to be the outcome for them?” According to Carrell, this is the key question at the centre of the 20-year-old challenge the speaker referred to in the title of his talk. Twenty years ago, he detailed, the pioneers of endovascular surgery recognised that anatomic severity was a major determinant of outcomes and therefore probably required reporting standards. “A scoring system would need to strike a balance between having enough granular detail to be useful and also being simple enough to be usable in everyday practice,” he said. A scoring system was built, with the expectation being that subsequent clinical investigations to test the system would be used to modify it.
In the 20 years since then, Carrell highlighted that a number of papers on the anatomic severity grade scoring system have been published, showing that it does indeed predict outcomes and complications. However, “there has not been quite so much progress in terms of using that data to modify the schemes,” Carrell pointed out. The speaker noted that there are probably a number of reasons for this, one being that “some of the things that go into the scoring schemes turn out to be rather complicated”. For example, he said, measuring infrarenal neck angulation is “really contentious,” with “very high intra-observer variability”.
“Can AI offer both that detail and simplicity?” Carrell asked, highlighting a key question in 2022. He explained that Cydar Medical is developing an extended capability to operate with what the company calls Intelligent Maps. “The concept is that when you plan a case, you are being informed by the outcomes of previous patients with similar anatomy and disease,” he explained.
The company’s current product, which is used for planning, guiding, and reviewing endovascular surgery, uses “virtual guidewires”. Carrell elaborated: “We use [virtual guidewires] for planning, but the main function is to identify where you are going to be operating, so that your planning is rendered in the form of a map and overlaid on a live fluoroscopy. As your real guidewires deform the anatomy, the virtual guidewires deform the map so that you have an overlay that reflects the real-time anatomy.”
Carrell provided some details on where the software is headed: “What we are doing with Intelligent Maps is taking that capability with the virtual guidewires and the AI that we have in the system and the data from all the thousands of patients who have been treated with Cydar EV Maps and building tools to analyse anatomy.” He added that these tools are a combination of deterministic algorithms using the virtual wires and also using simulated neural networks, or deep learning, to not just segment the vessels, but also to label the aorta and the iliacs according to what branches are coming off at each level. “For each patient, you bring these things together so that you have your geometry, but you have it in the context of where you are right the way along the aorta and the iliac system. This is highly deterministic, so you put the CT scan in and these measures come out, put the same CT scan in or a similar CT scan in and you get the same things coming out.”
One of Carrell’s key takeaway messages was that this software has relevance for helping inform decision-making. “You take one patient that produces the analytics and then we use it to match to other patients in the database already, patients who have similar anatomy. It is literally as simple as entering in the CT scan and finding those matching patients. We can then find out what treatment those patients had, what type of approach they had.”
Next, Cydar will be working on bringing in the “final steps” of the technology—the outcome metrics for the surgical strategy that was used. “We want to close that 20-year ambition to have the feedback loop in there to modify the scoring system,” Carrell concluded.