Spiral MRSI and tissue segmentation of normal-appearing white matter and white matter lesions in relapsing remitting multiple sclerosis patients☆

CEST 2022 - Differences in APT-weighted signal in T1 weighted isointense lesions, black holes and normal-appearing white matter in people with relapsing-remitting multiple sclerosis

PURPOSE

To evaluate amide proton transfer weighted (APTw) signal differences between multiple sclerosis (MS) lesions and contralateral normal-appearing white matter (cNAWM). Cellular changes during the demyelination process were also assessed by comparing APTw signal intensity in T1weighted isointense (ISO) and hypointense (black hole -BH) MS lesions in relation to cNAWM.

METHODS

Twenty-four people with relapsing-remitting MS (pw-RRMS) on stable therapy were recruited. MRI/APTw acquisitions were undertaken on a 3 T MRI scanner. The pre and post-processing, analysis, co-registration with structural MRI maps, and identification of regions of interest (ROIs) were all performed with Olea Sphere 3.0 software. Generalized linear model (GLM) univariate ANOVA was undertaken to test the hypotheses that differences in mean APTw were entered as dependent variables. ROIs were entered as random effect variables, which allowed all data to be included. Regions (lesions and cNAWM) and/or structure (ISO and BH) were the main factor variables. The models also included age, sex, disease duration, EDSS, and ROI volumes as covariates. Receiver operating characteristic (ROC) curve analyses were performed to evaluate the diagnostic performance of these comparisons.

RESULTS

A total of 502 MS lesions manually identified on T2-FLAIR from twenty-four pw-RRMS were subcategorized as 359 ISO and 143 BH with reference to the T1-MPRAGE cerebral cortex signal. Also, 490 ROIs of cNAWM were manually delineated to match the MS lesion positions. A two-tailed t-test showed that mean APTw values were higher in females than in males (t = 3.52, p < 0.001). Additionally, the mean APTw values of MS lesions were higher than those of cNAWM after accounting for covariates (mean lesion = 0.44, mean cNAWM = 0.13, F = 44.12, p < 0.001).The mean APTw values of ISO lesions were higher than those of cNAWM after accounting for covariates (mean ISO lesions = 0.42, mean cNAWM = 0.21, F = 12.12, p < 0.001). The mean APTw values of BH were also higher than those of cNAWM (mean BH lesions = 0.47, mean cNAWM = 0.033, F = 40.3, p < 0.001). The effect size (i.e., difference between lesion and cNAWM) for BH was found to be higher than for ISO (14 vs. 2). Diagnostic performance showed that APT was able to discriminate between all lesions and cNAWM with an accuracy of >75% (AUC = 0.79, SE = 0.014). Discrimination between ISO lesions and cNAWM was accomplished with an accuracy of >69% (AUC = 0.74, SE = 0.018), while discrimination between BH lesions and cNAWM was achieved at an accuracy of >80% (AUC = 0.87, SE = 0.021).

CONCLUSIONS

Our results highlight the potential of APTw imaging for use as a non-invasive technique that is able to provide essential molecular information to clinicians and researchers so that the stages of inflammation and degeneration in MS lesions can be better characterized.

Medullary vein architecture modulates the white matter BOLD cerebrovascular reactivity signal response to CO2: Observations from high-resolution T2* weighted imaging at 7T

Brain stress testing using blood oxygenation level-dependent (BOLD) MRI to evaluate changes in cerebrovascular reactivity (CVR) is of growing interest for evaluating white matter integrity. However, even under healthy conditions, the white matter BOLD-CVR response differs notably from that observed in the gray matter. In addition to actual arterial vascular control, the venous draining topology may influence the WM-CVR response leading to signal delays and dispersions. These types of alterations in hemodynamic parameters are sometimes linked with pathology, but may also arise from differences in normal venous architecture. In this work, high-resolution T2*weighted anatomical images combined with BOLD imaging during a hypercapnic breathing protocol were acquired using a 7 tesla MRI system. Hemodynamic parameters including base CVR, hemodynamic lag, lag-corrected CVR, response onset and signal dispersion, and finally ΔCVR (corrected CVR minus base CVR) were calculated in 8 subjects. Parameter maps were spatially normalized and correlated against an MNI-registered white matter medullary vein atlas. Moderate correlations (Pearson's rho) were observed between medullary vessel frequency (MVF) and ΔCVR (0.52; 0.58 for total WM), MVF and hemodynamic lag (0.42; 0.54 for total WM), MVF and signal dispersion (0.44; 0.53 for total WM), and finally MVF and signal onset (0.43; 0.52 for total WM). Results indicate that, when assessed in the context of the WM venous architecture, changes in the response shape may only be partially reflective of the actual vascular reactivity response occurring further upstream by control vessels. This finding may have implications when attributing diseases mechanisms and/or progression to presumed impaired WM BOLD-CVR.

Improved particle swarm optimized deep convolutional neural network with super-pixel clustering for multiple sclerosis lesion segmentation in brain MRI imaging

A central nervous system (CNS) disease affecting the insulating myelin sheaths around the brain axons is called multiple sclerosis (MS). In today's world, MS is extensively diagnosed and monitored using the MRI, because of the structural MRI sensitivity in dissemination of white matter lesions with respect to space and time. The main aim of this study is to propose Multiple Sclerosis Lesion Segmentation in Brain MRI imaging using Optimized Deep Convolutional Neural Network and Super-pixel Clustering. Three stages included in the proposed methodology are: (a) preprocessing, (b) segmentation of super-pixel, and (c) classification of super-pixel. In the first stage, image enhancement and skull stripping is done through performing a preprocessing step. In the second stage, the MS lesion and Non-MS lesion regions are segmented through applying SLICO algorithm over each slice of the volume. In the fourth stage, a CNN training and classification is performed using this segmented lesion and non-lesion regions. To handle this complex task, a newly developed Improved Particle Swarm Optimization (IPSO) based optimized convolutional neural network classifier is applied. On clinical MS data, the approach exhibits a significant increase in the accuracy segmenting of WM lesions when compared with the rest of evaluated methods.

3D magnetic resonance spectroscopic imaging reveals links between brain metabolites and multidimensional pain features in fibromyalgia

BACKGROUND

Fibromyalgia is a centralized multidimensional chronic pain syndrome, but its pathophysiology is not fully understood.

METHODS

We applied 3D magnetic resonance spectroscopic imaging (MRSI), covering multiple cortical and subcortical brain regions, to investigate the association between neuro-metabolite (e.g. combined glutamate and glutamine, Glx; myo-inositol, mIno; and combined (total) N-acetylaspartate and N-acetylaspartylglutamate, tNAA) levels and multidimensional clinical/behavioural variables (e.g. pain catastrophizing, clinical pain severity and evoked pain sensitivity) in women with fibromyalgia (N = 87).

RESULTS

Pain catastrophizing scores were positively correlated with Glx and tNAA levels in insular cortex, and negatively correlated with mIno levels in posterior cingulate cortex (PCC). Clinical pain severity was positively correlated with Glx levels in insula and PCC, and with tNAA levels in anterior midcingulate cortex (aMCC), but negatively correlated with mIno levels in aMCC and thalamus. Evoked pain sensitivity was negatively correlated with levels of tNAA in insular cortex, MCC, PCC and thalamus.

CONCLUSIONS

These findings support single voxel placement targeting nociceptive processing areas in prior ¹ H-MRS studies, but also highlight other areas not as commonly targeted, such as PCC, as important for chronic pain pathophysiology. Identifying target brain regions linked to multidimensional symptoms of fibromyalgia (e.g. negative cognitive/affective response to pain, clinical pain, evoked pain sensitivity) may aid the development of neuromodulatory and individualized therapies. Furthermore, efficient multi-region sampling with 3D MRSI could reduce the burden of lengthy scan time for clinical research applications of molecular brain-based mechanisms supporting multidimensional aspects of fibromyalgia.

SIGNIFICANCE

This large N study linked brain metabolites and pain features in fibromyalgia patients, with a better spatial resolution and brain coverage, to understand a molecular mechanism underlying pain catastrophizing and other aspects of pain transmission. Metabolite levels in self-referential cognitive processing area as well as pain-processing regions were associated with pain outcomes. These results could help the understanding of its pathophysiology and treatment strategies for clinicians.