Viz.ai’s Stephanie Vasquez-Mostofi speaks with MD&T about recent advancements and applications of AI-powered technology for abdominal aortic aneurysm.
The FDA recently granted 501(k) clearance for an artificial intelligence (AI) algorithm designed by Viz.ai to detect suspected abdominal aortic aneurysm. This is just the latest example of the life sciences industry embracing the use of ai-based algorithms.
Stephanie Vasquez-Mostofi, senior product marketing manager at Viz.ai, spoke with Medical Device & Technology about Viz.ai’s new algorithm, along with the impact that this and similar technologies may have on the industry moving forward.
(MDT:) How do algorithms like the Viz.ai Aortic Solution algorithm work?
Vasquez-Mostofi: When a patient receives an abdominal CT scan, that imaging study is automatically sent to the secure Viz Cloud. Our cloud-based abdominal aortic aneurysm (AAA) algorithm is constantly on and searching. If the algorithm finds a suspected AAA, the study is flagged for the appropriate member or members of the care team to review. That could include vascular care specialists, radiologists, or emergency room physicians. The goal is to enroll the patient in the most appropriate form of care, whether that is surveillance, intervention, or medical management.
(MDT:) Can these processes be used to diagnose other conditions?
Vasquez-Mostofi: The Viz Enterprise platform is composed of several AI algorithms that can detect a variety of suspected diseases. We have algorithms that can detect pulmonary embolisms, along with automated ratios for the right ventricle to left ventricle in order to determine heart strain, which is a critical factor for pulmonary embolism. Patients with elevated right heart strain are eight-times more likely to die of PE than those who don’t.
(MDT:) Are algorithms going to become more commonly used in diagnosis?
Vasquez-Mostofi: I think of artificial intelligence as being similar to electricity, which allows people to see the unseen. We’re able to use electricity to invent entirely new ways of thinking about the world. New technologies are emerging on a daily basis. The way we’re using AI can vary across institutions. Ten years from now, I don’t think there will be a single center in the U.S. that isn’t using AI in order to improve patient outcomes. We’ve been live with many early adopters since our first clearance in 2018, and now over 1,300 hospitals have moved in our direction across the globe. ThisThere’s several early adopters who are moving in that direction across the U.S. now and this very much has the potential to become standard of care in the future.
The American Heart Association just issued a position highly recommending that all stroke centers should have an AI powered stroke triage software in place.
(MDT:) What are the challenges in creating an algorithm like this one?
Vasquez-Mostofi: At Viz, we’re always looking to create the best possible algorithms and incredibly rich data sets which are composed of diverse sets of patients populations. This allows us to generalize the use of these algorithms. One challenge is taking the time to ensure that the data is of the upmost quality. The second challenge is understating the workflow. We work directly with local care teams to understand their current state, determining how patients are being diagnosed and followed up with, and where the leaks are in that workflow. Based on what we learn, we create a product that prevents leakage and also is able to be generalized across the population.
(MDT:) Will these algorithms fully replace traditional diagnosis methods?
Vasquez-Mostofi: I don’t believe that artificial intelligence will ever replace the expertise of a radiologist or a clinical care team. I do think the radiologists and care teams that use AI will quickly outpace those who aren’t. It’s not about AI versus humans, it’s about humans and AI working together in order to ensure that we’re providing the best possible care to patients.
John Varaklis, chief strategy officer for PhysIQ, discusses the company's machine-learning based algorithms for the detection and early detection of physiological conditions to inform both clinical care and clinical drug development.