Lifeng Yu PhD, Presenter: Nothing to Disclose

Joel Garland Fletcher MD, Abstract Co-Author: Grant, Siemens AG

Maria Shiung, Abstract Co-Author: Nothing to Disclose

Kristen Barry Thomas MD, Abstract Co-Author: Nothing to Disclose

Jane Sexton Matsumoto MD, Abstract Co-Author: Nothing to Disclose

Shannon Nicole Zingula MD, Abstract Co-Author: Nothing to Disclose

Cynthia H. McCollough PhD, Abstract Co-Author: Research Grant, Siemens AG

To evaluate the radiation dose reduction potential of a novel image-based denoising technique in pediatric body CT exams and compare it with an iterative reconstruction (IR) method.

Fifty pediatric CT exams (25 chest, 25 abdominopelvic (AP)) acquired using a weight-based low-kV protocol were included in this retrospective study. For each case, we used a validated noise-insertion tool developed in our lab to simulate half-dose images. A novel denoising technique, adaptive non-local means (aNLM) filter, which was developed in our institution, was applied to the half-dose images. An IR method (SAFIRE, Siemens) was also used to reconstruct the half-dose images. Three pediatric radiologists evaluated 4 sets of images for each of the 50 cases: (1) full dose + filtered-backprojection (FBP), (2) half dose + FBP, (3) half dose + IR, and (4) half dose + aNLM, in a randomized and blinded fashion. The overall image quality and the diagnostic confidence for each organ (chest: lung and mediastinum; AP: liver, kidney, and small bowel) were rated using a five point scale. For each case, each reader ranked dose/reconstruction method preference using a side by side comparison. Image sharpness for AP exams was rated.

The original CTDIvol was 5.3±2.1 mGy for AP exams and 2.4±1.1 mGy for chest exams. At half dose, both IR and aNLM improve the overall image quality over the FBP for both chest and AP exams (p<0.01). In AP, there was no significant difference between aNLM denoised images at half dose and the original full dose images (3.61±1.01 vs. 3.55±0.86, p=0.54), and aNLM performed better than IR (3.61±1.01 vs. 3.33±0.89, p<0.01). In chest, there was no significant difference between IR at half dose and the original full dose images (4.12±0.61 vs. 4.16±0.58, p=0.66), but IR performed better than aNLM (4.12±0.61 vs. 3.68±0.69, p<0.01). The organ-specific diagnostic confidence and preference order were consistent with the overall image quality evaluation.

The use of a novel image-based denoising technique resulted in a 50% radiation dose reduction in pediatric AP CT while maintaining the same diagnostic quality as in the full dose FBP images. IR image quality was worse than aNLM in the abdomen, but better in the chest.

A novel denoising technique, which can be implemented across all scanner platform, can preserve diagnostic image quality despite a 50% radiation dose reduction in pediatric AP CT.

Yu, L, Fletcher, J, Shiung, M, Thomas, K, Matsumoto, J, Zingula, S, McCollough, C, Radiation Dose Reduction in Pediatric Body CT Using a Novel Image-based Denoising Technique.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14013883.html