[Influence of Deep Learning Reconstruction on the Detection of Susceptibility Effect by T<sub>2</sub>*-weighted Imaging].
Authors
Affiliations (3)
Affiliations (3)
- Division of Radiology, Department of Medical Technology, Hirosaki University Hospital.
- Department of Radiation Science, Graduate School of Health Sciences, Hirosaki University.
- Department of Radiology, Graduate School of Medicine, Hirosaki University.
Abstract
The purpose of this study is to evaluate the influence of DLR on the detection of susceptibility effects by T<sub>2</sub>*WI. The imaging sequence was T<sub>2</sub>*WI using a 3.0 T MRI system. The phantom, assuming white matter, was made of diluted gadolinium contrast agent solidified with agar, and the susceptibility effect of bubbles in the phantom was evaluated. The slice thickness was varied from 1 to 5 mm, and images with and without DLR were acquired. The imaging was performed under conditions in which the number of averages was changed to adjust the SNR at each slice thickness, and also at 1 NEX separately. Binarization was performed on all slices, and the number of detections of the susceptibility effect was compared. The number of detections of the susceptibility effect increased with DLR for slice thicknesses of 2-5 mm and did not differ by DLR for slice thicknesses of 1 mm. Under 1 NEX with DLR, the number of detections increased for slice thicknesses of 3-5 mm and decreased for slice thicknesses of 1 mm. The results suggest that the number of detections of the susceptibility effect is higher when DLR is used. However, the behavior may vary depending on the SNR of the conventional image, so the image should be taken under guaranteed SNR conditions.