’s deep learning AI spots suspected aortic dissections with high accuracy: study

Though aortic dissections—in which the inner layer of the body’s central blood vessel tears—are rare, affecting around three out of every 100,000 people per year, they carry a disproportionately high mortality rate. An estimated 20% of people with acute aortic dissection die before even reaching the hospital, and half of all those who go untreated for the condition die within 24 hours.

Hoping to somewhat counteract those grim statistics is, which has developed an artificial intelligence algorithm that aims to catch suspected aortic dissections as early as possible. Study results presented earlier this month demonstrated the technology’s accuracy in doing just that.

Viz developed the AI in a partnership with Avicenna.AI that kicked off last year. It’s part of the larger Viz Aortic suite, which also includes tools to help detect three types of aneurysms. Not only does the software automatically alert clinicians when it spots a potential dissection or aneurysm, but it also allows them to examine the CT scan of concern right on their smartphones and begin wrangling necessary team members using the Viz platform’s built-in communication tools.

The Viz Aortic module was cleared by the FDA in 2021.

In the study of the aortic dissection-spotting tool—the results of which were presented at the VEITHSymposium last week—researchers applied the deep learning AI to just over 1,300 CT angiography scans that had been collected from across the U.S.

When put to the test, the algorithm was able to correctly detect aortic dissections with about 94% sensitivity and to exclude negative cases with more than 97% specificity.

It also registered a positive predictive value of 80%, denoting a high likelihood that the scans flagged by the AI included actual cases of aortic dissection, rather than false positives. The algorithm’s negative predictive value—measuring how likely it is that a negative reading is indeed negative for the condition—was even higher, at more than 99%.

Highly accurate AI-powered detection of aortic dissections “may have a positive downstream effect on patient triage, leading to accelerated care coordination, earlier diagnosis, timely initiation of life-saving interventions and better patient outcomes,” the study’s authors concluded, per Viz.

The promising results for the dissection-detecting algorithm—one of a handful of AI tools Viz has developed to help speed up triage of CT scans and other imaging data—come as the company continues to expand its reach.

Last week alone, it inked two new partnerships. The first pairs Viz with Illuminate, a developer of natural language processing software and other AI tools that sift through electronic medical records to pinpoint high-risk patients, then also help assess disease severity and continue to monitor patients after they’ve been diagnosed. Through their team-up, Illuminate’s technology will be added into the Viz Aortic module to improve care for people with aortic aneurysms.

A day later, Viz unveiled a partnership with Vastrax. It’s joining forces with the clinical research organization in an attempt to speed up clinical trial enrollment for studies focused on new neurovascular therapies. The company's software identifies trial-eligible patients by scanning images and then automatically notifying research teams.