Subtle Medical has received 510(k) clearance for a deep learning algorithm designed to cut the time it takes to perform positron emission tomography (PET) scans.
Shortening the duration or lowering the dose of PET scans leads to the loss of information and a lower signal-to-noise ratio, rendering the resulting image of less diagnostic value.
Subtle Medical created the algorithm, SubtlePET, to reconstruct standard-quality images from truncated scans, thereby enabling imaging centers to boost throughout while reducing the burden on patients.
The potential for algorithms to cut the time it takes to capture usable medical images has attracted considerable interest. That idea is at the heart of a collaboration between Facebook and NYU School of Medicine that is targeting a 90% reduction in the time it takes to perform magnetic resonance imaging (MRI) scans. The idea is also a big part of why Subtle Medical exists.
Enhao Gong and Greg Zaharchuk spun Subtle Medical out of Stanford University to build on their work to enhance PET scans. In a paper published 12 months ago, Gong, Zaharchuk and collaborators presented a deep-learning method for reconstructing standard-quality PET images from scans that used 0.5% of the regular dose, albeit it from a very small study. At that time, the focus was on reducing the dose to limit radiation exposure to patients.
Now, Subtle Medical has received 510(k) clearance and a CE mark for its PET-enhancing algorithm. The algorithm is an add-on to existing scanners and picture archiving and communication systems.
Subtle Medical is pitching the algorithm as a way for hospitals and imaging centers to generate more scans without investing in extra imaging equipment. Scans presented by Subtle Medical to support its case show SubtlePET making PET images gathered at one minute per bed comparable to scans that took four times as long to generate. It remains to be seen whether SubtlePET can match that performance in widespread, real-world use at a diversity of centers with different imaging protocols.
For now, the algorithm is in pilot clinical use at some university hospitals and imaging centers in the U.S. and overseas. Subtle Medical has received positive feedback from at least one early user of the system.
"SubtlePET technology allows us to scan a patient four times faster than normal, yet maintain equal image quality, not otherwise impacting work flow," Michael Brant-Zawadzki, of Hoag Hospital in California, said in a statement. "This creates immediate ROI benefit for our hospital and a compelling value proposition."
The algorithm is at the forefront of a field that is likely to get busier in the years to come. Subtle Medical itself is advancing algorithms to accelerate MRI scans and reduce the gadolinium dose used in other imaging procedures, while researchers at Massachusetts General Hospital and elsewhere are also working on the problem.