Effect of in-painting on cortical thickness measurements in multiple sclerosis: A large cohort study

Abstract

A comprehensive analysis of the effect of lesion in-painting on the estimation of cortical thickness using magnetic resonance imaging was performed on a large cohort of 918 relapsing-remitting multiple sclerosis patients who participated in a phase III multicenter clinical trial. An automatic lesion in-painting algorithm was developed and implemented. Cortical thickness was measured using the FreeSurfer pipeline with and without in-painting. The effect of in-painting was evaluated using FreeSurfer's paired analysis pipeline. Multivariate regression analysis was also performed with field strength and lesion load as additional factors. Overall, the estimated cortical thickness was different with in-painting than without. The effect of in-painting was observed to be region dependent, more significant in the left hemisphere compared to the right, was more prominent at 1.5 T relative to 3 T, and was greater at higher lesion volumes. Our results show that even for data acquired at 1.5 T in patients with high lesion load, the mean cortical thickness difference with and without in-painting is ∼2%. Based on these results, it appears that in-painting has only a small effect on the estimated regional and global cortical thickness. Hum Brain Mapp 36:3749-3760, 2015. © 2015 Wiley Periodicals, Inc.

Keywords: FreeSurfer; cortical thickness; lesion in-painting; multiple sclerosis; multiple sclerosis lesions.

Figure 1

Age distribution of RRMS patients by gender and field strength.

Figure 2

An axial slice of original (left), in‐painted (middle), and segmented (right) image. In the color‐coded image WM, GM, CSF, T2‐lesions, and T1‐hypointense lesions are indicated by white, gray, blue, pink, and red, respectively. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 3

Lesion volume distribution in the RRMS patients (a) whole group, (b) low lesion volume, and (c) high lesion volume. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 4

Images showing segmentation of juxtacortical lesions on two patients (top and bottom rows). (a) Short echo FSE, (b) long echo FSE, (c) FLAIR, (d) 3DT1, and (e) segmented images. In the color‐coded segmented image WM, GM, CSF, and lesions are indicated by white, gray, blue, and red, respectively. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 5

Lateral views of inflated left (L) and right (R) hemispheres depicting group difference maps highlighting significantly thinner regions in nonpainted images compared to lesion in‐painted images. The different lobes are denoted in different colors: frontal (blue), temporal (yellow), parietal (green), occipital (red), and insula (brown). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 6

Lesion distribution maps in RRMS patients in the whole group and at each of the two field strengths. The temperature bar indicates the lesion probability. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 7

Lateral views of inflated left (L) hemisphere grouped by lesion load for 1.5 T and 3 T data. Images depict group difference maps highlighting significantly thinner regions in nonpainted images compared to lesion in‐painted images. Rows 1 and 2 show significant differences at P = 0.01, the right hemisphere showed a similar behavior. The different lobes are depicted in different colors: frontal (blue), temporal (yellow), parietal (green), occipital (red), and insula (brown). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]