RSNA 2022 CT Preview

AI takes low-dose chest CT dose even lower

By Erik L. Ridley, staff writer

Thursday, December 1 | 1:30 p.m.-2:30 p.m. | R6-SSCH10-2 | Room N228
Researchers have found that deep learning-based reconstruction can enable low-dose chest CT images to be reconstructed with a 75% lower radiation dose without negatively impacting image quality and lesion detection.

In a retrospective study, a team from South Korea used a commercial vendor-agnostic deep-learning algorithm to reconstruct low-dose chest CT images from 100 patients with only a quarter of the standard low-dose CT exam reconstructed with a dedicated iterative reconstruction algorithm. They then compared the image quality and detectability of lung nodules on the quarter-dose images with the full low-dose images.

The researchers found no subjective difference in noise, spatial resolution, and overall image quality between the two types of images. There was also no difference in detection rates for Lung-RADS 3-4 nodules.

“Deep-learning reconstruction may further decrease radiation dose needed for lung nodule evaluation,” wrote presenter Dr. Gyeong Deok Jo of Seoul National University School of Medicine and colleagues.

Take in this scientific presentation on Thursday afternoon to get all the details.

Copyright © 2022