Brain Scans Reveal Which Sleep Disorder Patients Will Develop Dementia

For neurologists caring for people with a rare and unsettling sleep condition, one of the most consequential questions has always been which disease will eventually emerge. Patients with isolated REM sleep behaviour disorder, in which the normal paralysis that accompanies dreaming is lost and sufferers physically act out their dreams, are known to be at extraordinarily high risk of future neurodegeneration. More than ninety percent of them ultimately convert to a diagnosable synucleinopathy, most commonly Parkinson’s disease or dementia with Lewy bodies. Yet until now, clinicians have had no reliable way to tell, years in advance, which of the two diseases a given patient is destined to develop. A new study published in the Journal of Neurology by Andreas Myhre Baun of Aarhus University Hospital, Alex Iranzo of Hospital Clínic de Barcelona, and an international team of collaborators reports that a carefully constructed combination of brain imaging measures can make exactly that prediction, and it does so with striking statistical force.

The study rested on a deceptively simple observation about how previous research had approached the problem. Most imaging studies of people with isolated REM sleep behaviour disorder have examined one biomarker at a time, measuring, for example, dopamine transporter loss in the striatum, or cortical atrophy on structural MRI, or markers of brain inflammation. Such single-modality measures can indicate that a patient is on track to develop some form of neurodegenerative disease, but they generally cannot distinguish Parkinson’s disease from dementia with Lewy bodies, because both conditions involve overlapping pathology of the alpha-synuclein protein. The Danish-Spanish team reasoned that the key to phenotype-specific prediction might lie not in any single scan but in the relationships between multiple pathological processes measured simultaneously in the same brain.

To test this idea, the researchers recruited a well-characterised cohort of twenty-one patients with isolated REM sleep behaviour disorder and put them through an unusually comprehensive imaging protocol. Each patient received positron emission tomography scans with three different radioactive tracers, each designed to illuminate a distinct facet of brain biology. Fluorine-18 labelled DOPA PET quantified the integrity of the dopaminergic system, particularly the nigrostriatal pathways that degenerate in Parkinson’s disease. Carbon-11 labelled donepezil PET mapped cholinergic function, revealing the density of acetylcholine signalling, which is known to be profoundly disrupted in dementia with Lewy bodies and to underlie many of its cognitive and perceptual symptoms. Carbon-11 labelled PK11195 PET, meanwhile, served as a window onto neuroinflammation by binding to activated microglia, the brain’s resident immune cells, which become mobilised in response to ongoing neurodegenerative injury.

Alongside the PET examinations, each participant underwent structural magnetic resonance imaging to quantify grey matter volume across the cortex, as well as dynamic susceptibility contrast MRI, a technique that tracks a bolus of contrast agent moving through the cerebral vasculature. From these perfusion measurements the team extracted indices of microcirculatory dysregulation, capturing how well blood was being delivered and distributed through the brain’s smallest vessels. The inclusion of this microvascular dimension reflects a growing appreciation, championed in earlier work by co-author Leif Østergaard’s group in Aarhus, that capillary dysfunction may be an underappreciated contributor to neurodegenerative disease rather than a mere bystander.

The analytical centrepiece of the study was a multimodal adaptation of the scaled sub-profile model, a statistical framework originally developed to identify disease-related patterns of covariance in functional imaging data from patients with Parkinson’s disease. Rather than asking whether any single region or single tracer differed between patients, the method searches for spatially distributed networks across which the multiple imaging modalities co-vary together. In other words, it looks for brain regions where, across the cohort, dopaminergic loss, cholinergic decline, inflammation, atrophy and microvascular impairment rise and fall in a coordinated fashion. This approach transforms a collection of separate scans into a single integrated portrait of each patient’s brain, and it is precisely this integration that gave the study its predictive power.

What emerged was a coherent multimodal network converging on the medial occipito-parietal cortex, the posterior region of the brain encompassing the cuneus and adjacent parietal areas. Within this network, the same patients who showed high neuroinflammation also tended to show cholinergic dysfunction, grey matter atrophy and microcirculatory problems, all concentrated in the same posterior cortical territory. This convergence is biologically meaningful. The occipito-parietal cortex has long been implicated in the visual hallucinations, fluctuating cognition and visuospatial deficits that distinguish dementia with Lewy bodies from other dementias, and earlier studies by several of the same authors had already flagged cuneus atrophy as a harbinger of phenoconversion. The new findings suggest that this region is not merely one affected area among many but a genuine crossroads where multiple pathological processes meet.

The predictive result was the study’s most dramatic finding. When the team combined the multimodal occipito-parietal pattern with striatal fluorine-18 DOPA uptake, a measure of dopaminergic integrity in the basal ganglia, the composite measure specifically predicted conversion to dementia with Lewy bodies rather than to Parkinson’s disease. The strength of the association was expressed as a sub-distribution hazard ratio of 38.68, with a ninety-five percent confidence interval running from 7.545 to 198.3, and the effect survived statistical correction for age and disease duration. In practical terms, patients whose brains displayed the posterior cortical pathological signature together with striatal dopaminergic impairment were dramatically more likely to develop the dementing form of synucleinopathy, while those without it were more likely to convert to the motor-predominant Parkinsonian phenotype.

The implications for clinical practice and for drug development are considerable. Neuroprotective trials for Parkinson’s disease and for dementia with Lewy bodies have repeatedly been hampered by the inclusion of heterogeneous patient populations, in which participants destined for different clinical outcomes are lumped together, diluting any apparent treatment effect. A biomarker that can stratify prodromal patients by their likely conversion phenotype would allow trials to enrol more homogeneous cohorts, match experimental therapies to the disease process they are actually designed to slow, and interpret outcomes more cleanly. For patients themselves, knowing years in advance whether the likely future holds a movement disorder or a dementia could transform planning, monitoring and, eventually, the timing of interventions aimed at the earliest stages of disease.

The study also adds weight to a broader conceptual shift in how prodromal synucleinopathies are understood. Recent biological staging frameworks, including the SynNeurGe criteria for Parkinson’s disease and the integrated staging system for neuronal alpha-synuclein disease, have argued that diagnosis and prognosis should rest on biological markers rather than on clinical syndromes that appear only late in the disease course. The multimodal imaging profile described by Baun and colleagues fits squarely within this vision, offering a way to characterise the internal biology of prodromal disease rather than simply waiting for symptoms to declare themselves. It complements other emerging biomarkers, such as the detection of misfolded alpha-synuclein in cerebrospinal fluid by seed amplification assays, by adding spatial and mechanistic information that fluid biomarkers cannot provide.

Caveats remain, and the authors are careful about them. The cohort of twenty-one patients is small, and multimodal PET studies of this depth are expensive and technically demanding, which limits how readily the approach can be scaled to large populations. The supporting data are available only from the corresponding authors upon reasonable request, reflecting privacy constraints on the participants. Nevertheless, the consistency of the posterior cortical findings with a growing independent literature, including studies of cortical thickness, perfusion, metabolism and atrophy progression in the same patient population, lends credibility to the central conclusion. If future studies replicate the result in larger cohorts, the dream of telling a sleeping patient which disease awaits them, and of intervening before it arrives, will have moved a decisive step closer to the clinic.

Subject of Research: Multimodal PET-MRI biomarkers for predicting phenoconversion to dementia with Lewy bodies in isolated REM sleep behaviour disorder

Article Title: Multimodal PET–MRI profiling predicts dementia with Lewy bodies in isolated REM sleep behaviour disorder

Article References: Baun, A. M., Iranzo, A., Terkelsen, M. H., Hinz, R., Stokholm, M. G., Serradell, M., Svendsen, K. B., Garrido, A., Vilas, D., Møller, A., Gaig, C., Tolosa, E., Brooks, D. J., Borghammer, P., Eskildsen, S. F., & Pavese, N. (2026). Multimodal PET–MRI profiling predicts dementia with Lewy bodies in isolated REM sleep behaviour disorder. Journal of Neurology, 273(10), Article 635. https://doi.org/10.1007/s00415-026-14176-3

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14176-3

Keywords: REM sleep behaviour disorder, dementia with Lewy bodies, Parkinson's disease, PET imaging, MRI, neuroinflammation, cholinergic dysfunction, striatal dopamine, occipito-parietal cortex, biomarkers, neurodegeneration, alpha-synuclein