A new deep vein thrombosis (DVT) diagnostic pathway incorporating non-expert artificial intelligence (AI)-guided compression ultrasound could reduce workload and costs for healthcare systems while providing a quicker diagnosis and improving patient care.
Efthymios (Makis) Avgerinos, MD, of Athens, Greece, shared this key conclusion from a pilot study of the ThinkSono guidance system (ThinkSono) at the 2024 European Society for Vascular Surgery (ESVS) annual meeting (Sept. 24–27) in Kraków, Poland.
“The diagnosis of DVT in a tertiary setting is part of a complicated clinical pathway including clinical probability scoring, D-dimer testing and a full duplex scan, often leading to long waiting times,” Avgerinos—who is co-director of the Clinic of Vascular and Endovascular Surgery at Athens Medical Group and visiting professor of vascular surgery at the University of Athens in Greece—told Venous News ahead of the ESVS meeting. He added that the need to have trained radiologists or sonographers available 24/7 incurs high costs and represents a “suboptimal use of resources”.
The ThinkSono system, Avgerinos explained, is based on the point-of-care ultrasound concept and eliminates the need for an expert operator and the absolute need to perform a standard duplex. “It is an AI-based software that you can install on your cell phone or tablet and will guide a non-expert on how to perform a compression ultrasound using a wireless probe,” he continued.
In their pilot study of the ThinkSono system, Avgerinos and colleagues set out to evaluate its clinical utilisation in real-world practice—having previously demonstrated its ability to generate valid images—hoping to show that the system could spare the need for standard venous duplex.
The researchers prospectively enrolled 53 patients who were suspected to have a DVT. These patients underwent a ThinkSono scan, performed by a nurse or resident in the emergency room, with the resulting images then reviewed remotely by the on-call radiologist. D-dimer testing was also performed as a safety measure.
Avgerinos shared that all 53 scans generated were of appropriate quality to be read by the radiologist. “This is a great milestone for an AI technology in the hands of non-experts,” he told Venous News, going on to reveal that 89% of the scans were negative for DVT. The team managed to discharge one-third of the patients without performing any standard duplex scan, which Avgerinos commented represents significant cost and time savings.
Six out of eight suspected DVTs were confirmed on standard duplex to be real, with the ThinkSono technology not missing a diagnosis. “Interestingly enough,” Avgerinos added, “all positive D-dimer testing patients who had a negative ThinkSono scan were also eventually negative, indicating D-dimer has a very low specificity and is not really a powerful testing method to rule in DVT.”
Avgerinos continued that it took on average six and a half minutes in total for scan and review. The median length of time between scan and review was 30 minutes, although there was a wide variation in this figure. According to Avgerinos, this “reflects unoptimized hospital logistics—mainly the availability of the qualified clinician to review the data”. He clarified that the expert clinicians who took part in the pilot study were volunteers who had various other clinical lab duties, going on to suggest that careful resource planning or the use of dedicated reviewers could help to maximize the efficiency of the AI-assisted pathway.
Avgerinos summarized: “Our study demonstrated that in DVT diagnostic workflow, using AI and remote expert review could safely decrease the need for a full duplex ultrasound and even D-dimer testing for a significant proportion of the patients who are suspected to have a DVT in the emergency room.”
Looking ahead, he noted that a large trial is needed to validate the results of this pilot study, and that clinical commercialization will soon launch in the UK and Germany, while a Food and Drug Administration (FDA) trial is underway in the U.S.
Finally, Avgerinos considered the wider application of this technology. While the pilot study focused on tertiary centers, he hinted at its potential future use in community hospitals and ambulatory centers. “Not a lot of hospitals have radiologists who can perform standard duplex scans 24/7,” he said. “With this technology, patients would not need to travel to the reference centers. A nurse at the community hospital, a nurse at the GP’s office or at the ambulatory centre could perform the scan and an expert could review it remotely from the main hospital. The application eventually expands to any remote or rural underserved area, essentially democratizing ultrasound—anybody can do it, anywhere.”
