QC-Automator: Deep Learning-Based Automated Quality Control for Diffusion MR Images

摘要: Quality assessment of diffusion MRI (dMRI) data is essential prior to any analysis, so that appropriate pre-processing can be used improve quality and ensure the presence artifacts do not affect results subsequent image analysis. Manual subjective, possibly error-prone, infeasible, especially considering growing number consortium-like studies, underlining need for automation process. In this paper, we have developed a deep-learning-based automated control (QC) tool, QC-Automator, dMRI data, handle variety such as motion, multiband interleaving, ghosting, susceptibility, herringbone, chemical shifts. QC-Automator uses convolutional neural networks along with transfer learning train artifact detection on labeled dataset ∼332,000 slices from 155 unique subjects 5 scanners different acquisitions, achieving 98% accuracy in detecting artifacts. The method fast paves way efficient effective large datasets. It also demonstrated replicable other datasets acquisition parameters.