Evaluation of prospective motion correction of high-resolution 3D-T2-FLAIR acquisitions in epilepsy patients

Abstract

T2-FLAIR is the single most sensitive MRI contrast to detect lesions underlying focal epilepsies but 3D sequences used to obtain isotropic high-resolution images are susceptible to motion artefacts. Prospective motion correction (PMC) - demonstrated to improve 3D-T1 image quality in a pediatric population - was applied to high-resolution 3D-T2-FLAIR scans in adult epilepsy patients to evaluate its clinical benefit. Coronal 3D-T2-FLAIR scans were acquired with a 1mm isotropic resolution on a 3T MRI scanner. Two expert neuroradiologists reviewed 40 scans without PMC and 40 with navigator-based PMC. Visual assessment addressed six criteria of image quality (resolution, SNR, WM-GM contrast, intensity homogeneity, lesion conspicuity, diagnostic confidence) on a seven-point Likert scale (from non-diagnostic to outstanding). SNR was also objectively quantified within the white matter. PMC scans had near-identical scores on the criteria of image quality to non-PMC scans, with the notable exception that intensity homogeneity was generally worse. Using PMC, the percentage of scans with bad image quality was substantially lower than without PMC (3.25% vs. 12.5%) on the other five criteria. Quantitative SNR estimates revealed that PMC and non-PMC had no significant difference in SNR (P=0.07). Application of prospective motion correction to 3D-T2-FLAIR sequences decreased the percentage of low-quality scans, reducing the number of scans that need to be repeated to obtain clinically useful data.

Keywords: Epilepsy; FLAIR; Image quality; Prospective motion correction.

Fig. 1

Coronal (top), axial (middle), and sagittal (bottom row) slices of examples of a non-PMC scans with good image quality (left-most column), non-PMC scan with acceptable image quality without obvious motion-corruption (second column) non-PMC scan obviously corrupted by motion (third column), and a PMC scan (right-most column). The two raters score and average score of above-average (5) and standard (4) for the left scan, poor (2) for the second scan, poor (2) and non-diagnostic (1) for the third scan, and standard (4) and average (3) for the right-most scan, respectively.

Fig. 2

Cumulative histograms of image quality scores for all criteria for the two groups of scans combining the scores from both raters. Color legend is displayed in the bottom of the figure. The use of PMC reduces the number of scans rated non-diagnostic and poor, for all criteria except intensity homogeneity. This comes at the apparent expense of having fewer high-ranking scores (above average and better).

Fig. 3

Correlations between visual and quantitative SNR ratings for the two scan groups. The blue dots and black crosses represent the scores from the first and second human rater, respectively.