Within multiple sclerosis (MS) lesions, T1 weighted hypo-intensity correlates with pathological markers of irreversible damage, including axonal degeneration. Studies have found that T1 “black holes” have a higher correlation with clinical disability than do T2 weighted lesions, and that measurable T1 signal drop out occurs within slowly enlarging lesions or chronic active lesions, as opposed to inactive lesions, suggestive of ongoing axonal degeneration.
To perform this quantitative analysis two pre-processing steps are necessary. The first is correction of low frequency intensity nonuniformity present in image data, also known as bias fields. The second is normalisation of the signal between subjects and timepoints. Both of these steps need to maintain representative contrast between pathological and normal appearing white matter (NAWM) to provide accurate results. Validation of these methods within the MS context is needed to support their use as a biomarker in MS.
To assess the accuracy N34 bias correction on maintaining accurate MS lesion contrast
To assess the accuracy of Freesurfer normalization tool6 in maintaining accurate relative signal intensity
21 relapsing MS and 26 progressive MS subjects had MRI brain scans with 1mm3 3D T1 fast spoiled gradient echo (FSPGR), and 0.6 x 0.46mm2 FLAIR sequences. FLAIR images were resampled and linearly registered to the T1 using FLIRT. MS lesions were segmented from the FLAIR using JIM software.
To calculate an accurate measure of local tissue contrast between MS lesions and surrounding NAWM, the lesion segmentations were dilated and masked for white matter segmentation. The lesion mask was then subtracted from the segmentation to create an edge of NAWM surrounding the lesion. The measure of raw local tissue contrast was taken as the T1 value of
lesion core / lesion edge
N3 bias correction was done with the Freesurfer command mri_nu_correct.mni. Normalization of the T1 sequences was undertaken with Freesurfers mri_normalize which imposes a hard ceiling effect on the white matter peak found, creating an homogenous NAWM intensity of 110 and CSF intensity of ~35.
Correlations of each MS lesion between the bias field corrected and local tissue contrast was explored. Correlations between the bias field corrected and normalized values was explored.
A total of 2401 lesions were analysed. N3 bias corrected T1 images correlated with local tissue contrast of the raw T1 images with an R2 = 0.7556. Normalized T1 images correlated with N3 bias corrected images with an R2 = 0.9415. Average MS lesion normalized T1 intensity was similar between relapsing MS and progressive MS cohorts (82.927 vs 82.793, p = 0.42).
In this analysis T1 intensity was able to be corrected for bias fields and normalized to a set range with reasonable accuracy in maintaining lesion contrast. This supports the use of these methods as a biomarker in quantifying the T1 signal within MS lesions.