Deep Learning Based Noise Reduction for Brain MR Imaging: Tests on Phantoms and Healthy Volunteers
Problem
Brain MR images acquired with fewer acquisitions suffer from noise, degrading diagnostic image quality.
Why it matters
Denoising could enable faster MRI scans, reducing scan time and improving patient throughput.
What the AI did
A deep learning-based reconstruction (dDLR) denoised brain MR images, tested on phantoms and 15 healthy volunteers and compared against other methods.
Results
PSNR and SSIM for dDLR statistically higher than DnCNN and SCNN (P < 0.001); dDLR-NAQ2 significantly better than NAQ2 for SSIM; superior image quality per two neuroradiologists.
Impact
Peer-reviewed study with 242 citations showing dDLR outperformed existing denoising methods on stated metrics.
Record
| Venue | Magnetic Resonance in Medical Sciences |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Masafumi Kidoh, Kensuke Shinoda, Mika Kitajima, Kenzo Isogawa, Masahito Nambu, Hiroyuki Uetani, Kosuke Morita, Takeshi Nakaura et al. |
| Citations | 242 (OpenAlex, at import time) |
| References | 30 |
| Model or method | deep learning-based reconstruction (dDLR), compared to DnCNN and SCNN |
| Organization | not stated in source |
| Primary source | https://doi.org/10.2463/mrms.mp.2019-0018 |
| Open access | diamond |
| Keywords | deep learning, MRI denoising, PSNR, SSIM, brain imaging |
Importance rating
real but narrower advance
MRI denoising could enable faster scans and better diagnostics, a real but narrow advance; abstract only demonstrates feasibility on phantoms and 15 healthy volunteers, no scale stated.