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

Abstract

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

Keywords: MRI; artifacts; convolutional neural networks; diffusion MRI; quality control.

FIGURE 1

Representative slices of the different artifacts that the QC-Automator was trained to detect.

FIGURE 2

A typical architecture of a CNN: A set of convolution and pooling layers with successive fully connected and softmax layer.

FIGURE 3

Pipeline of the proposed approach for the QC-Automator: (Top) CNN pre-trained on ImageNet to obtain parameters used for transfer learning, where the last layer of the network was re-trained with our dataset of manually labeled artifactual and artifact-free data. The process was replicated to create the axial (Middle) and the sagittal detector (Bottom). The blue box represents the QC-Automator. Given an input image (Left), both the axial and sagittal detectors are applied to it and the status of each slice as artifact-free or artifactual is predicted.

FIGURE 4

Results of Axial Detector: Representative slices of correctly and incorrectly classified slices are presented.

FIGURE 5

Results of Sagittal Detector Representative slices of correctly and incorrectly classified artifactual slices.

FIGURE 6

A sample of false positive slice for Dataset 4: The slice contains aliasing artifact. Our expert labeled it as artifact-free one. But our QC-Automator caught it as it contained a similar pattern to ghosting artifact.