A new AI approach produces virtual PET/CT images that may eliminate the need for low-dose CT and consequently, reduce radiation exposure.

AI produces virtual PET/CT images, eliminates need for low-dose CT

June 21, 2022
by John R. Fischer, Senior Reporter
Researchers in Maryland have come up with a new approach utilizing AI to develop virtual PET/CT images that do away with the need for CT acquisitions and cut radiation exposure in half.

The scientists from the National Cancer Institute presented their method at the Society of Nuclear Medicine and Molecular Imaging 2022 Annual Meeting in Vancouver, Canada. Using virtual attenuation-corrected PET scans, they were able to reduce or eliminate low-dose CT, which, they say, exposes patients to the most radiation in PET/CT scans, despite being largely redundant.

With further development, the technique could enable more frequent and safer use of PET imaging for disease and treatment monitoring. “This opens opportunities for increasing the frequency and number of PET scans per patient per year, which could provide more accurate assessment for lesion detection, treatment efficacy, radiotracer effectivity and other measures in research and patient care,” said Kevin Ma, a postdoctoral researcher at the NCI, told HCB News.

The model was trained on 185 18F-DCFPyL PSMA PET/CT studies. Another 60 were used to validate it and an additional 60 were used for testing. The researchers used a 2D Pix2Pix generator to produce synthetic attenuation-corrected PET scans (gen-PET) from the original non-attenuation-corrected PET.

Two nuclear medicine physicians then reviewed 40 PET/CT studies in a randomized order without knowing if the image was from original-attenuation-corrected PET or gen-PET. Each recorded the number and locations of lesions detected with PET and qualitatively reviewed overall noise and image quality. They were successfully able to find lesions on the gen-PET scans, with reasonable sensitivity.

Virtual imaging is a growing interest in radiology, with a study earlier this month from Sydney, Australia discussing the benefits of virtual MR. With more refinement, it says that such scans could one day replace invasive biopsies for determining the likelihood of rejection in heart transplant patients. The approach it talks about uses MR to assess heart oedema levels, which are closely linked to heart inflammation, and has accurately detected signs of rejection.