Unsupervised dMRI Artifact Detection via Angular Resolution Enhancement and Cycle Consistency Learning

TL;DR

A novel unsupervised deep learning framework called UdAD-AC is proposed, which leverages dMRI angular resolution enhancement and cycle consistency learning to capture the effective representation of artifact-free dMRI data during training, and it identifies data containing artifacts using designed confidence score during inference.

Abstract

Diffusion magnetic resonance imaging (dMRI) is a crucial technique in neuroimaging studies, allowing for the non-invasive probing of the underlying structures of brain tissues. Clinical dMRI data is susceptible to various artifacts during acquisition, which can lead to unreliable subsequent analyses. Therefore, dMRI preprocessing is essential for improving image quality, and manual inspection is often required to ensure that the preprocessed data is sufficiently corrected. However, manual inspection requires expertise and is time-consuming, especially with large-scale dMRI datasets. Given these challenges, an automated dMRI artifact detection tool is necessary to increase the productivity and reliability of dMRI data analysis. To this end, we propose a novel unsupervised deep learning framework called $\textbf{U}$nsupervised $\textbf{d}$MRI $\textbf{A}$rtifact $\textbf{D}$etection via $\textbf{A}$ngular Resolution Enhancement and $\textbf{C}$ycle Consistency Learning (UdAD-AC). UdAD-AC leverages dMRI angular resolution enhancement and cycle consistency learning to capture the effective representation of artifact-free dMRI data during training, and it identifies data containing artifacts using designed confidence score during inference. To assess the capability of UdAD-AC, several commonly reported dMRI artifacts, including bias field, susceptibility distortion, and corrupted volume, were added to the testing data. Experimental results demonstrate that UdAD-AC achieves the best performance compared to competitive methods in unsupervised dMRI artifact detection.

Figures (5)

• Figure 1: The detailed framework of the proposed UdAD-AC.
• Figure 2: An illustration of relationships between the derived FA and the dMRI volumes used for estimation.
• Figure 3: Qualitative comparisons in axial view of two sets of dMRI data and corresponding predicted FA maps by UdAD-AC. $\widehat{FA}_{X_1}$ is generated from artifact-free $X_1$, $FA^*_{X_1}$ is derived from all artifact-free dMRI volumes (including $X_1$) of the subject, $\widehat{FA}_{X_2}$ is generated from $X_2$, and $FA^*_{X_2}$ is derived from all available dMRI volumes (including $X_2$) of the subject.
• Figure 4: Visual comparison of FA maps generated from dMRI data with corrupted volumes. (a) $X_{corr}$: dMRI data with corrupted volumes. (b) $\widehat{FA}_{corr}$: predicted FA map generated from the corrupted dMRI data, showing a blurry appearance. (c) $FA^*_{corr}$: FA map derived using all available dMRI volumes, showing clear anatomical details.
• Figure 5: Qualitative examples of a subject with susceptibility distortion in axial view . a) $X_{dist}$ is the input dMRI data to the UdAD-AC, b) $\widehat{FA}_{dist}$ is the predicted FA generated from $X_{dist}$, and c) $FA^*_{dist}$ is the FA derived by using all dMRI volumes of the subject.