The scale of the problem is striking. Multiple sclerosis affects more than 2.9 million people worldwide, and spinal cord lesions are present on up to 85 percent of cord MRI scans in newly diagnosed patients. As many as 76 percent of people with relapsing-remitting MS and 90 percent of those with progressive disease carry at least one cord lesion when the entire cord is imaged. Yet a 2026 systematic review cited in the paper found that of 291 MS clinical trials, only eight, roughly 2.7 percent, planned to include spinal cord MRI metrics as an outcome, and just three ever published those results. None showed a significant effect on cord atrophy or lesion number compared with placebo.

Why has the cord been so hard to study? The review describes a perfect storm of technical obstacles. The cord is small, demanding thin image slices of one to three millimetres to capture lesions reliably. It is encased in bone and bathed in cerebrospinal fluid that pulses rhythmically with every heartbeat, creating motion artefact. Susceptibility differences between cord, fluid, bone and air distort the magnetic field, and minor inconsistencies in slice positioning between scans can introduce spurious apparent change over time. Krieger famously likened standard clinical cord MRI to cave paintings, a vivid indictment of images whose resolution and signal-to-noise ratio lag far behind what is routinely achieved in the brain.

Despite these limitations, the prognostic evidence is unambiguous. Spinal cord lesions predict conversion to MS in people with radiologically isolated syndrome, and patients whose first clinical event involves the cord convert to definite MS faster than those without cord involvement. Total cord lesion volume at diagnosis correlates with later disability measured by the Expanded Disability Status Scale, the Nine-Hole Peg Test and the Timed 25-Foot Walk. Crucially, both lesion volume and lesion number at diagnosis are associated with progression independent of relapse activity, the insidious worsening known as PIRA that unfolds even when inflammatory attacks are suppressed. This suggests that standard monitoring protocols, which focus heavily on the brain, miss much of the biology driving long-term disability.

Atrophy tells an even sharper story. Healthy adults lose spinal cord cross-sectional area at a physiological rate of about 0.06 percent per year; people with MS lose it at roughly 1.78 percent per year. In a retrospective study of more than 360 people with relapsing-remitting MS, those who later converted to secondary progressive disease showed cord atrophy of 2.19 percent per year over four years, versus 0.88 percent in those who did not. Cord cross-sectional area correlates with disability more closely than brain atrophy does, and grey matter loss within the cord, which is greater in progressive disease, shows even stronger associations with clinical impairment than whole-cord measures.

The diagnostic stakes have risen with the 2024 revisions to the McDonald criteria. The cord now counts as one of five anatomical topographies for demonstrating dissemination in space, and in suspected primary progressive MS, just two cord lesions are sufficient to establish that criterion without a second topography. The criteria also caution that in patients over fifty or with significant vascular risk factors, cord lesions should be sought before attributing white matter changes to small-vessel ischaemia. Lesion morphology matters too: MS cord lesions are typically short, peripheral, cigar-shaped on sagittal views and wedge-shaped axially, features that help separate MS from neuromyelitis optica spectrum disorder and MOG antibody disease, which favour longitudinally extensive, central or grey-matter-predominant lesions, including the so-called H sign in MOGAD.

The differential diagnosis extends well beyond other inflammatory diseases. Spinal cord infarction produces the classic owl-eyes or pencil-like patterns on MRI, while spinal dural arteriovenous fistulas can mimic progressive MS with slowly worsening myelopathy, and missing one can be devastating, particularly since steroids may worsen the congested cord. Compressive spondylotic myelopathy, vitamin B12 and copper deficiency with its inverted dorsal V sign, infectious myelitides, neurosarcoidosis with its trident sign, and even rare pegivirus-associated cord tractopathy all crowd the differential. The review underscores that accurate pattern recognition on cord MRI is a diagnostic safety net as much as a research tool.

On the acquisition side, consensus is crystallising. The 2024 MAGNIMS-CMSC-NAIMS recommendations call for whole-cord imaging with two complementary sagittal sequences, gapless T2 fast spin echo, STIR, PSIR or MP(2)RAGE at slice thickness under three millimetres, plus one axial sequence, and explicitly discourage FLAIR for cord lesion detection. Notably, moving from 1.5 to 3 Tesla confers no overall advantage for cord lesion detection, because field inhomogeneity, motion sensitivity and energy deposition offset the signal gains. Newer three-dimensional sequences such as phase-sensitive inversion recovery have demonstrated greater sensitivity for cervical cord lesions than conventional two-dimensional STIR, and ultra-high-field 7 Tesla imaging has boosted lesion detection by a reported 52 percent while revealing cord analogues of the brain’s central vein sign and paramagnetic rim lesions, though these remain research tools for now.

The most forward-looking section of the review surveys quantitative techniques edging toward the clinic. Diffusion tensor imaging reveals reduced fractional anisotropy and elevated diffusivity extending into normal-appearing cord tissue, correlating with disability even in patients with low lesion burdens. Magnetisation transfer ratio, sensitive to myelin integrity, is reduced across the cervical cord of recently diagnosed patients even when visible lesions are excluded, hinting at a biomarker for PIRA. Myelin water imaging fell by more than ten percent over two years in primary progressive MS, and proton spectroscopy shows N-acetyl-aspartate reductions signalling axonal compromise before atrophy becomes measurable. Emerging chemical exchange saturation transfer and sodium MRI point to diffuse biochemical disruption beyond focal lesions. Artificial intelligence is beginning to automate cord and lesion segmentation through tools such as the open-source Spinal Cord Toolbox, though no model yet supports reliable longitudinal lesion tracking, and none has achieved regulatory approval as a clinical device.

The authors’ conclusion is pragmatic. Every patient should receive high-quality MRI of the entire spinal cord at diagnosis, and because a meaningful fraction of cord lesions are asymptomatic, they propose routine whole-cord imaging every two to three years, an expert opinion offered in the absence of formal consensus, to catch subclinical activity before irreversible disability accumulates. They also urge trial designers to incorporate cord lesion burden and atrophy as endpoints, particularly in progressive disease cohorts where cord pathology dominates the clinical picture. The evidence base still suffers from small, single-centre, cross-sectional studies and heterogeneous acquisition protocols, and most advanced biomarkers remain confined to research settings. But the direction of travel is clear: the spinal cord, long relegated to cave paintings, is being repainted in high resolution, and what it reveals may reshape how MS is diagnosed, monitored and treated.

Subject of Research: Spinal cord MRI biomarkers for diagnosis and progression monitoring in multiple sclerosis

Article Title: Spinal cord imaging in multiple sclerosis: from diagnosis to disease progression

Article References: Spinal cord imaging in multiple sclerosis: from diagnosis to disease progression. (n.d.). https://doi.org/10.1007/s00415-026-14122-3

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14122-3

Keywords: multiple sclerosis, spinal cord, MRI, spinal cord atrophy, McDonald criteria, progression independent of relapse activity, demyelination, magnetisation transfer, diffusion MRI, myelin water imaging, artificial intelligence, disease-modifying therapy