Automated Quality Assessment of Structural Magnetic Resonance Brain Images Based on a Supervised Machine Learning Algorithm

Abstract

High-resolution three-dimensional magnetic resonance imaging (3D-MRI) is being increasingly used to delineate morphological changes underlying neuropsychiatric disorders. Unfortunately, artifacts frequently compromise the utility of 3D-MRI yielding irreproducible results, from both type I and type II errors. It is therefore critical to screen 3D-MRIs for artifacts before use. Currently, quality assessment involves slice-wise visual inspection of 3D-MRI volumes, a procedure that is both subjective and time consuming. Automating the quality rating of 3D-MRI could improve the efficiency and reproducibility of the procedure. The present study is one of the first efforts to apply a support vector machine (SVM) algorithm in the quality assessment of structural brain images, using global and region of interest (ROI) automated image quality features developed in-house. SVM is a supervised machine-learning algorithm that can predict the category of test datasets based on the knowledge acquired from a learning dataset. The performance (accuracy) of the automated SVM approach was assessed, by comparing the SVM-predicted quality labels to investigator-determined quality labels. The accuracy for classifying 1457 3D-MRI volumes from our database using the SVM approach is around 80%. These results are promising and illustrate the possibility of using SVM as an automated quality assessment tool for 3D-MRI.

Adolescent Tuning of Association Cortex in Human Structural Brain Networks

“…While we have carried out stringent quality control of our structural scans and FreeSurfer reconstructions of cortical thickness (details in Supplementary Information), we cannot completely rule out potential artefactual effects of motion on our results. Thus, further analysis of structural correlation development in datasets including estimates of head motion from volumetric tracking (Tisdall et al 2012, 2016) or novel automated estimates of data quality (Shehzad et al 2015; Pizarro et al 2016; Rosen et al 2017) will be important in the future.…”

Automated quality assessment of structural magnetic resonance images in children: Comparison with visual inspection and surface‐based reconstruction

“…Furthermore, for large population‐based studies where multiple scans need to be rated, visual inspection of data can be prone to rater‐differences (inter‐rater reliability) and rater‐drift (intra‐rater reliability), which is not a problem with automated approaches. There have been several algorithms developed to automatically assess the quality of structural images (Mortamet et al, , Pizarro et al, ). Mortamet et al () measured voxel intensities outside of the head with the hypothesis that artifacts enlarge the noise intensity and causes a right‐skew (greater intensity) in the distribution.…”

“…There are several different approaches in the literature to automatically derive quality assessment metrics from structural MRI scans (Atkinson, Hill, Stoyle, Summers, & Keevil, , Mortamet et al, , Pizarro et al, ), and available software to generate quality assessment metrics (http://preprocessed-connectomes-project.org/quality-assessment-protocol/index.html). Mortamet et al () measured voxel intensities in the background noise with the hypothesis that artifacts cause a right‐skew in the distribution of voxel intensities.…”

“…While artifacts in some sequences have the advantage of becoming a contrast medium in other sequences (i.e., diffusion or flow), other artifacts, such as motion artifacts, remain problematic in image analyses. In fact, motion related artifacts are one of the major challenges associated with imaging pediatric populations (Blumenthal, Zijdenbos, Molloy, & Giedd, 2002, Backhausen et al, 2016 and has received considerable attention due to the influence of motion on connectivity-based analyses of resting-state fMRI (Power, Barnes, Snyder, Schlaggar, & Petersen, 2012, Satterthwaite et al, 2012, Van Dijk, Sabuncu, & Buckner, 2012. However, motion-related issues also impact structural MRI with evidence that movement related artifacts can influence measures such as volumes of cortical and deeper structures (Blumenthal et al, 2002, Alexander-Bloch et al, 2016 and cortical thickness and surface area (Reuter et al, 2015, Backhausen et al, 2016, Ducharme et al, 2016.…”

Automated image quality evaluation of structural brain MRI using an ensemble of deep learning networks

“…However, the sensitivity (0.28) was rather low. Pizarro et al achieved good accuracy (0.80) in evaluating the image quality in a single‐center study using three volumetric and three artifact‐specific features and a support vector machine classifier. All these methods assessed image quality using handcrafted image quality metrics, whose selection is subjective, and their computation can be very time‐consuming.…”