People with multiple sclerosis (MS) often experience problems with walking, sensation, and automatic body functions such as bladder control, but these symptoms do not always match the amount of damage seen on conventional magnetic resonance imaging (MRI). This study explored whether diffusion tensor imaging (DTI), an advanced MRI technique that measures the microscopic structure of tissue, could provide additional insights by examining the lumbosacral spinal cord—the lower portion of the spinal cord that helps control the legs and pelvic organs. The researchers compared MRI scans from 29 people with mild relapsing-remitting MS and 27 healthy volunteers. They found that healthy participants showed natural differences in DTI measurements across different spinal cord regions, highlighting the importance of analyzing these areas separately. Overall, people with MS did not show significant differences from healthy participants in normal-appearing spinal cord tissue. However, among participants with MS, higher fractional anisotropy (FA) values in the back (dorsal) portion of the spinal cord were associated with poorer mobility and sensation. These findings suggest that higher FA in the lower spinal cord may not necessarily indicate healthier tissue in people with MS and underscore the need for region-specific analyses and larger studies that include patients with more advanced disease.

Fig 1

Example data for one healthy control (A, 26-year-old female) and one person with multiple sclerosis (B, 42-year-old female). Shown are the anatomical multi-echo, gradient echo (mFFE) images, image with no diffusion-sensitizing gradients (b = 0), mean diffusion-weighted image (DWI), three DWI slices averaged with good gray matter (GM) contrast without (left) and with (right) GM, white matter columns, and lesions overlaid, and the fractional anisotropy (FA) and axial diffusivity (AD) quantitative maps. All images are shown for three slices corresponding to spinal levels throughout the lumbar enlargement. Lesions are highlighted by the yellow area on the anatomical slices.

Multiple sclerosis (MS) affects both the brain and spinal cord, but these areas are often studied separately using magnetic resonance imaging (MRI). This study examined whether tissue damage develops in similar ways across both regions and whether an advanced MRI technique could provide more useful information than conventional imaging methods. The researchers used the same MRI session to image the brain and cervical (neck) spinal cord of 34 people with relapsing-remitting MS and 36 healthy volunteers. They compared traditional diffusion tensor imaging (DTI) with a newer technique called standard model imaging with free water (SMIfw), which provides more detailed information about the brain’s and spinal cord’s microscopic structure. Both methods detected MS-related changes, but their performance differed depending on the region being studied. A measure derived from SMIfw, called neurite density fraction, consistently identified MS-related tissue damage in both the brain and spinal cord, while commonly used DTI measures performed as well only in the brain. The findings also suggest that although damage to nerve fibers is a common feature of MS throughout the central nervous system, the inflammatory environment surrounding lesions differs between the brain and spinal cord. Overall, the results indicate that SMIfw could improve the assessment of MS across the entire central nervous system and may be valuable for future clinical trials and disease monitoring.

FIGURE 1

All diffusion models can be roughly summarized similar concepts, which can be captured either by “sticks” or cylinders, spheres with isometric diffusion, or diffusion tensors with directional diffusion. In this depiction, SMI represents both SMI and SMIfw models, with SMIfw including a f_w measure not otherwise included in the SMI model.

The spinal cord plays an important role in multiple sclerosis, or MS, but it has not been studied as much as the brain. This review summarizes the main takeaways from a 2025 workshop on spinal cord imaging in MS, including recent progress, ongoing problems, and future directions. It explains how damage to the spinal cord, such as lesions and shrinkage, can help doctors diagnose MS, predict how the disease may progress, and monitor how well treatment is working. The review also highlights new markers that may help track disease worsening even when patients are not having relapses. Studies comparing magnetic resonance imaging, or MRI, with tissue samples and patient outcomes support the usefulness of newer spinal cord imaging methods. At the same time, the review notes that spinal cord imaging still faces technical challenges, including the need for better analysis pipelines and more consistent results across studies. Overall, the authors argue that advanced, quantitative spinal cord imaging should be used more widely in clinical trials, research, and, when possible, patient care, because it can help show the full extent of MS and improve outcomes.

Figure 1. Spinal cord pathological features and MRI-pathology correlations. (a) Lesion frequency heatmaps of total demyelinated lesions in the cervical (top), thoracic (middle), and lumbar (bottom) spinal cord. Lesion predilection sites include the dorsal columns, lateral columns, and gray matter as a whole, with relative sparing of the subpial surface. (b) Myelin (proteolipid protein) and (c) fibrin(ogen) immunostaining in adjacent spinal cord sections from an MS case. Fibrin(ogen) deposition is consistently found in the central part of the cord, including gray matter, and mesial aspects of the lateral columns and central part of the dorsal column in areas outside demyelinated lesions. (d) Hematoxylin and eosin–stained section with magnified inset (e), demonstrating thickened vasculature in the MS spinal cord, a finding consistently found in younger cases. Perivascular space dilatation is also a common feature (not shown). (f) and (g) Palmgren silver-stained sections showing reduced axonal density in an MS case (g) compared with control (f), with predilection for loss of small diameter axons. (h) Luxol Fast Blue stain for myelin and (i) Bielschowsky stain for axons in the secondary progressive MS section demonstrate reduced staining in a focal lesion (arrow), which is visualized on 7 Tesla ex vivo MRI as (j) T2-weighted hyperintensity and (k) myelin water fraction imaging hypointensity. Comparison between Luxol Fast Blue myelin staining optical density and quantitative MRI, (l) radial diffusivity (RD), (m) inhomogeneous magnetization transfer (ihMT) and (n) myelin water fraction (MWF) show strong quantitative correlations between histology and MRI markers for myelin in the spinal cord. (a) Adapted from Waldman et al.,��Acta Neuropathol��2024 ��һ��ݶ�; (f) and (g) adapted from DeLuca et al.,��Brain��2004.

]]> /valiant/2026/05/27/generalizable-spinal-cord-multiple-sclerosis-lesion-segmentation-across-mri-contrasts-protocols-and-centers/ Wed, 27 May 2026 02:05:25 +0000 /valiant/?p=6821 Benveniste, Pierre-Louis.; Létourneau-Guillon, Laurent.; Araujo, David.; Chougar, Lydia.; Fetco, Dumitru.; Hori, Masaaki.; Kamiya, Kouhei.; Messina, Steven.; Tsagkas, Charidimos.; Audoin, Bertrand.; Bakshi, Rohit.; Bannier, Elise.; Blezek, Daniel.; Brisset, Jean-Christophe.; Callot, Virginie.; Charlson, Erik.; Chen, Michelle.; Ciccarelli, Olga.; Demortière, Sarah.; Edan, Gilles.; Filippi, Massimo.; Granberg, Tobias.; Granziera, Cristina.; Hemond, Christopher C.; Keegan, B. Mark.; Kerbrat, Anne.; Kirschke, Jan.; Kolind, Shannon.; Labauge, Pierre.; Lee, Lisa Eunyoung.; Liu, Yaou.; Mainero, Caterina.; McGinnis, Julian.; Laines Medina, Nilser.; Mühlau, Mark.; Nair, Govind.; O’Grady, Kristin P.; Oh, Jiwon.; Ouellette, Russell.; Prat, Alexandre.; Reich, Daniel S.; Rocca, Maria A.; Shepherd, Timothy M.; Smith, Seth A.; Stawiarz, Leszek.; Talbott, Jason.; Tam, Roger.; Tauhid, Shahamat.; Traboulsee, Anthony.; Treaba, Constantina Andrada.; Valsasina, Paola.; Vavasour, Zachary.; Yiannakas, Marios.; Lombaert, Hervé.; Cohen-Adad, Julien. (2026).��.��Multiple Sclerosis Journal.��

Magnetic resonance imaging, or MRI, is an important tool for finding and tracking spinal cord lesions in people with multiple sclerosis (MS), which are areas of damage caused by the disease. But automatic computer methods for detecting and outlining these lesions often work well only for one MRI type or one hospital’s scanning setup, which makes them less useful in real clinics where scan methods vary a lot. To address this, the researchers developed a more robust segmentation system, meaning a model that can automatically identify lesion boundaries, across many MRI contrasts and imaging sites. They trained and tested it on a large dataset of 4,428 annotated images from 1,849 people with MS across 23 imaging centers, using six different MRI contrast types and scans taken at 1.5, 3, and 7 tesla, which refers to the strength of the MRI scanner. Compared with existing methods that are designed for only one contrast type, the new model generalized better across different scan settings, according to neuroradiologist ratings. It also remained strong when tested across different spinal cord levels, image resolutions, threshold settings, and external datasets. Overall, the study shows that this approach can detect spinal cord MS lesions accurately and reliably across diverse MRI data, which is an important step toward making automated lesion analysis more useful in everyday clinical care.

Figure 1. Sankey diagram of annotated MRI scans across clinical sites. Line thickness is associated with the number of scans.

MRI scan distribution is clustered per acquisition type (3D, 2D sagittal, or 2D axial) and per MRI contrast, for each site.

Multiple sclerosis (MS) causes lesions, or areas of damage, in the brain that can be seen on multicontrast magnetic resonance (MR) images. Automatically segmenting, or outlining, these lesions using deep learning (DL) can improve speed and consistency compared to manual tracing by experts. Although many DL methods perform well on data similar to what they were trained on, they often struggle when tested on new datasets from different hospitals or scanners, a problem known as poor out-of-domain generalization.

To address this issue, the researchers developed a new method called UNISELF. The goal of UNISELF is to achieve high segmentation accuracy within the original training domain while also performing well on data from different sources. UNISELF introduces a test-time self-ensembled lesion fusion strategy, which combines multiple predictions at test time to improve accuracy. It also uses test-time instance normalization (TTIN) of latent features, meaning it adjusts internal feature representations during testing to better handle domain shifts and missing input contrasts, such as when certain MR image types are unavailable.

The model was trained using data from the ISBI 2015 longitudinal MS segmentation challenge. On the official test dataset, UNISELF ranked among the top-performing methods. Importantly, when evaluated on out-of-domain datasets with different scanners, imaging protocols, and missing contrasts—including the MICCAI 2016 dataset, the UMCL dataset, and a private multisite dataset—UNISELF outperformed other benchmark models trained on the same ISBI data. These results suggest that UNISELF is both accurate and robust to real-world variations in MR imaging, making it a promising tool for automated MS lesion segmentation across diverse clinical settings.

Fig. 1.��An illustration of the spatial augmentation, network input, and network output during training in UNISELF.

Multiple sclerosis (MS) is an autoimmune disease that damages the central nervous system, particularly by destroying the protective myelin around nerves. Studying changes in the spinal cord could help us understand why MS causes neurological problems and clinical symptoms. However, conventional MRI does not detect subtle molecular changes in tissue.��Chemical exchange saturation transfer (CEST) is an MRI technique that can measure biochemical changes in tissue with high sensitivity and without the need for contrast agents. In practice, CEST signals are influenced by other effects—such as semi-solid magnetization transfer (MT), direct water saturation, and T1 relaxation—that can be altered in MS, so these confounding factors must be removed to accurately quantify changes.

In this study, 53 people with relapsing-remitting MS (pwRRMS) and 45 healthy controls were scanned at 3 T to measure amide�����Ի���nuclear Overhauser enhancement (NOE) CEST effects in the cervical spinal cord. Using a method called Lorentzian fitting, we removed confounding effects and calculated the apparent exchange-dependent relaxation (AREX) contrast. Comparing uncorrected and corrected AREX contrasts across tissue types and groups revealed that AREX NOE differed significantly in lesions compared to normal-appearing white matter in pwRRMS. People with MS also showed greater variability in both CEST contrasts than healthy controls. A subgroup analysis based on neurological disability showed that AREX amide differed significantly between pwRRMS patients with and without disability.

These findings highlight the importance of correcting for confounding effects in CEST imaging to isolate true biochemical changes in the cervical spinal cord. Doing so provides a more specific characterization of tissue pathology and helps link molecular changes to disease severity in MS.

Fig. 1. Example data from one multiple sclerosis (MS) participant (32 years, female, 0 Expanded Disability Status Scale (EDSS) score, 1.5 years disease duration).��A. Representative anatomical multi-echo, gradient echo (mFFE) image corresponding to one slice in the CEST volume with teal arrow pointing to the lesion (top), with average gray matter (GM), white matter (WM), and lesion segmentations overlaid (bottom).��B. A four-pool Lorentzian model (magnetization transfer (MT), direct saturation (DS), nuclear Overhauser effect (NOE), and amide) was fit to the measured Z-spectrum (black). Average measured Z-spectra and fitted pools are shown for GM (left), WM (middle), and lesioned (right) tissue.��C. Uncorrected average raw Z-spectra and inverted Z-spectra for the three tissue types (left). After subtracting the MT and NOE fitted pools, differences between GM and WM on the Z-spectra and inverted spectra upfield from water are less apparent for the amide contrast Z-spectra (middle). After subtracting the MT and amide fitted pools, differences between GM and WM on the Z-spectra and inverted spectra downfield from water are less apparent for the NOE contrast Z-spectra (right).

Multiple sclerosis (MS) is a chronic disease of the nervous system that causes inflammation, damage to the protective covering of nerve fibers (demyelination), and degeneration of axons. These changes can be studied using diffusion tensor imaging (DTI), which measures microstructural damage in the brain and spinal cord. In the brain, researchers often use white matter (WM) tractography to examine changes along specific pathways. In the spinal cord (SC), however, anatomy is naturally divided into cervical levels, which provides a different way to study regional changes.

In this study, we used an along-level approach to measure both microstructural features (such as fractional anisotropy, a DTI measure of tissue integrity) and macrostructural features (such as cross-sectional area) of the SC in people with relapsing-remitting MS (pwRRMS) compared to healthy controls (HCs).

The results showed that analyzing the SC level by level was more sensitive to detecting group differences than averaging across the whole cord. Segmenting the cord into WM tracts and gray matter (GM) subregions revealed specific, localized changes along the cord and within its cross-sections. Importantly, GM atrophy was linked with greater clinical disability, whereas microstructural changes did not show significant associations with disability measures.

These findings highlight the value of level-specific analysis for identifying localized spinal cord pathology and suggest a more refined framework for studying SC changes in MS.

Figure 1. Depiction of healthy and MS cord processing, including delineation of the masks relative to healthy or lesioned tissue. The contrasts are included in the right column. CSA and diffusion-derived indices were calculated for HCs, and CSA, diffusion-derived indices, and lesion load were calculated for pwRRMS.

In multiple sclerosis (MS), a disease where the protective covering of nerve fibers is damaged, the symptoms people experience often do not match what standard MRI scans show. Functional MRI (fMRI) can help us understand how the brain and spinal cord’s networks adapt to this structural damage. While fMRI studies in the brain are common, studying the spinal cord is more difficult due to its small size and interference from normal body movements.

In this study, we used resting-state fMRI at 3T to examine the spinal cord of healthy people and those with relapsing-remitting MS. We looked at functional connectivity, which measures how different regions of the spinal cord communicate, and related these findings to clinical measures of disability.

We found that the strongest connectivity occurs between the ventral gray matter regions in both healthy participants and people with MS. Reduced connectivity was linked to poorer mobility. Using a data-driven analysis, we also observed a possible compensatory increase in connectivity in earlier stages of MS compared with later stages.

These results suggest that MS affects how the spinal cord functions and that the nervous system may try to compensate for early damage. Further research is needed, but our findings support the idea that functional changes in the spinal cord are an important part of MS.

Fig.1. Anatomic and functional data processing pipelines. For the anatomic image, vertebral levels were identified on the sagittal T2w image before co-registration of the T2w and multi-echo fast field echo (mFFE) image. For the functional image (fMRI), motion correction was followed by physiologic noise regression using AFNI-RETROICOR and band-pass filtering via a Chebyshev Type II filter. The resulting denoised fMRI and mFFE images were co-registered to one another, and then to the PAM50 template between spinal levels C3 and C5. The gray matter (GM) horns applied on top of the final functional image were extracted from the mFFE image in functional space. ROI correlations were identified between each horn, per slice.