#17
Nick Fox 

New roles and new questions for MRI in the era of anti-amyloid therapies for AD

MRI has taken on new, and clinically important, roles with the advent of recently licensed anti-amyloid immunotherapies (lecanemab and donanemab). MRI is an essential requirement for eligibility for these therapies – supporting the diagnosis and identifying exclusionary features such as the presence of another pathology (e.g. extensive vascular burden) or a safety concern (e.g. >4 microhaemorrhages for an anti-amyloid immunotherapy). These therapies have only been licensed for early-stage AD (MCI or mild AD) and there is evidence that starting earlier has greater benefit. This will add pressure to reduce delays in diagnosis, so individuals do not miss a limited window of opportunity. In addition, (serial) MRI is needed for those who are started on one of these therapies. Monitoring for (and of) amyloid related imaging abnormalities (ARIA) is critical for safety.

All this will be challenging for health care systems – to provide a timely diagnosis and to assess treatment eligibility and then safe use. This in turn means greater demand for MRI. Methods to accelerate sequences and reduce scan times may help with these pressures. Moreover, more scans means greater demand for MRI reporting and so there is increasing interest in methods that provide decision support tools (e.g. AI-based) – for diagnosis and ARIA.

There will also be important roles of MRI (and other imaging) in trials in the presymptomatic stage of AD – a growing focus – and in the non-AD dementias.

The trials of anti-amyloid immunotherapies also showed (initially) unexpected changes on MRI: greater brain volume loss and greater ventricular enlargement relative to those on placebo. This has been a source of controversy. I will present results looking across trials and at subject-level data to argue that amyloid plaque removal (along with associated inflammatory response) is a plausible explanation for the brain volume changes – i.e. amyloid-removal related pseudoatrophy (ARPA). I will also discuss the disproportionate ventricular enlargement that appears to have a relationship to ARIA-E.

#18
Antonio Valero-Cabré

MRI Neuronavigated and E-Field Model based Transcranial Direct Current stimulation (tDCS) to modulate language impairments in early-onset neurodegenerative conditions: Comparative outcomes & upcoming challenges

Non-Invasive Brain Stimulation has shown promise improving cognitive deficits in several neurological conditions, including stroke and dementia. Whether such approaches can consistently improve language dysfunction across distinct neurodegenerative syndromes remains incompletely understood. We here aimed to compare the immediate effects of specific neuromodulation protocols on language performance in patients with behavioral-variant Frontotemporal Dementia (bv-FTD), Progressive Supranuclear Palsy (PSP) and the semantic (sv-PPA) and logopenic (lv-PPA) variants of Primary Progressive Aphasia. These four cohorts underwent in a cross-over design, three tDCS sessions (20 min, intensity: 1.59 mA, charge density: 0.06 mA/m2) delivered to relevant left hemisphere atrophic regions with anodal-tDCS the dorsolateral prefrontal cortex (DLPFC) in bv-FTD and PSP, the anterior temporal lobe (ATL) in sv-PPA, and the temporo-parietal junction (TPJ) in lv-PPA, compared to the stimulation of their right homotopic location with cathodal-tDCS and to sham stimulation. Letter fluency, categorical decision, semantic association and picture naming were assessed neuromodulation potential immediately following the discontinuation of stimulation, compared to baseline outcomes. MRI-based computational electric-field modeling was conducted to confirm target engagement, evaluate its extent while assessing interindividual variability. All groups exhibited measurable impairments across language domains compared to healthy controls. MRI-based modeling analyses demonstrated that E-fields encompassed intended

cortical targets supporting the anatomical rationale of each protocol. Improvements following active intervention were most evident in PSP and sv-PPA. PSP participants showed gains in semantic association and letter fluency, whereas sv-PPA participants exhibited enhanced semantic association performance. These benefits were observed after one or both active conditions and were absent following sham intervention. In contrast, no significant language changes were detected for bv-FTD nor lv-PPA. We conclude that neuromodulation approaches may transiently improve selected language functions in PSP and svPPA, supporting further investigation of interventions for communication deficits in neurodegenerative disorders and syndromes. Pending challenges and potential solutions driving more efficient interventions based on the manipulation of oscillations, use of MRI E-field based anatomically personalized interventions, close-loop state-dependent approaches and intensive home-delivered treatments, will be discussed.

#19
Emma Gummesson

Detecting Learning-induced Synaptic Changes In Mild Cognitive Impairment Using Sv2a Pet

Background: Alzheimer’s disease (AD) is characterised by amyloid and tau pathology, which associate with synaptic loss and impaired cognition and learning abilities. While synaptic dysfunction is central to AD, the temporal relationship between AD core pathology and synaptic loss remains unclear. Synaptic vesicle glycoprotein 2A (SV2A) is a transmembrane protein located in synaptic vesicles, and recent PET tracers enable its use as an in vivo marker of synaptic density. Here, we describe an ongoing study examining whether learning-based interventions are associated with measurable changes in SV2A-binding using in vivo [¹⁸F]SynVesT-1 PET in individuals with AD-associated early cognitive impairment.

Methods: Participants aged 65-75 years who have been diagnosed with mild cognitive impairment are randomised into either a fine motor skill training or spatial navigation training group. They initially undergo a baseline PET/MR scan using the [¹⁸F]SynVesT-1 tracer followed by a training intervention approximately a week later. Spatial navigation learning is designed to target synaptic changes in the hippocampus, whereas fine motor learning targets changes in the motor cortex. Follow-up PET/MR imaging is performed after completion of the intervention. Clinical development will be assessed after two and four years, enabling evaluation of longitudinal change and individual trajectories.

Results: We will evaluate whether learning-based interventions are associated with measurable changes in SV2A-binding in targeted brain regions by investigating regionally specific changes in synaptic density, with navigation training expected to primarily affect hippocampal SV2A binding and motor learning primarily affecting motor cortical regions. Data collection is ongoing, and 10 participants meeting the inclusion criteria currently have complete baseline (pre- and post-intervention) datasets. Recruitment has now become more consistent, with further data expected by September 2026.

Conclusion: This pilot study will be the first to test whether learning-based interventions can induce detectable short-term changes in synaptic density in individuals with cognitive impairment, as measured by SV2A PET. Our findings will provide insight about the feasibility, regional specificity, and sensitivity of SV2A PET as an outcome measure in interventional studies targeting synaptic mechanisms across the AD spectrum.

#20
Konstantinos Chiotis 

Clinical and PET outcomes following Aβ-targeting therapies in sporadic Early Onset Alzheimer’s disease; trial emulation using the LEADS cohort

Objective: Trial data on A-targeting therapy in early-onset Alzheimer’s disease (EOAD) remain limited. Using Longitudinal Early-Onset Alzheimer’s Study (LEADS) data, we examined whether treatment initiation was associated with inflection points in biomarker and clinical trajectories.

Method: We identified 31 participants with sporadic EOAD who had received A-targeting therapies, and had both pre- and post-treatment LEADS visits (Table 1). In a target-trial emulation design, each treated participant was matched 1:2 to untreated participants by age, sex, APOE, clinical stage, amyloid and tau PET, CDR global, CDR-SB, and MoCA. For each untreated match, the treated participant’s treatment start defined pseudo–time zero. Piecewise mixed-effects models estimated pre- and post-treatment slopes for amyloid and tau PET, and CDR-SB (Fig.1A).

Result: Before treatment, amyloid-PET averaged 96±3 Centiloids with a flat, non-significant slope (Fig.1B). In untreated participants, slopes remained unchanged after time zero. After treatment initiation, treated participants showed a significant decline of Δ-from-Δ=−34.0±2.9 Centiloids/year (p<0.001) versus untreated, with reductions predominantly in frontotemporal-parietal areas and the basal ganglia (Fig.2A). Pre-treatment, tau-PET was elevated (2.41±0.09 SUVR), mainly in temporoparietal areas, and increased significantly in widespread neocortical areas (+0.09±0.02 SUVR/year; p<0.001). No post-treatment slope differences in tau PET were observed between groups (Fig.1C,2B). Before treatment, CDR-SB averaged 3.4 (SE 0.3) and worsened by +0.5±0.1 units/year (p<0.001); after time zero, deterioration accelerated in untreated participants (Δ=+1.5±0.3 units/year; p<0.001), but was significantly slower in treated relative to untreated participants (Δ-from-Δ=−0.9±0.3 units/year; p<0.001) (Fig.1D).

Conclusion: In this EOAD cohort, our emulated-trial analysis captured—in treated patients—amyloid clearance rates comparable to late-onset Alzheimer’s disease trials, with no change in tau trajectories and without the acceleration of CDR-SB decline that was seen among untreated participants. The amyloid clearance displayed regional variation requiring further investigation. These findings support the feasibility of real-world, biomarker-based observational studies to assess treatment effects.