Research progress in neuroimaging of sporadic early-onset Alzheimer’s disease

Authors: Xiang Zhang, Chao Yang

Abstract

Early-onset Alzheimer’s disease (EOAD) is defined as Alzheimer’s disease (AD) with an age at onset younger than 65 years, accounting for approximately 5% of all AD cases. More than 90% of EOAD cases do not carry autosomal dominant pathogenic mutations. Although its prevalence is lower than that of late-onset Alzheimer’s disease (LOAD), EOAD follows a more aggressive clinical course. A subset of EOAD patients present with non-amnestic variant phenotypes, including logopenic variant of primary progressive aphasia (lvPPA), frontal variant Alzheimer’s disease (fvAD), posterior cortical atrophy (PCA), and corticobasal syndrome (CBS). However, the neuroimaging characteristics of EOAD and their differences from those of LOAD remain poorly elucidated to date. Therefore, this review systematically summarizes the recent research progress in neuroimaging of EOAD, including structural, functional, and metabolic imaging modalities. We also discuss the potential pathogenesis of EOAD, with the aim to provide evidence-based reference for the development of EOAD-specific imaging assessment systems and the optimization of disease efficacy monitoring protocols in future research.

Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder defined by core neuropathological amyloid-β (Aβ) plaques and neurofibrillary tangles composed of hyperphosphorylated tau protein, and it represents the most common cause of dementia worldwide. AD with symptom onset after the age of 65 years is classified as late-onset Alzheimer’s disease (LOAD), which is the typical form of the disease. In contrast, approximately 5% of AD patients develop clinical symptoms before the age of 65 years, which is the conventional threshold for the definition of early-onset Alzheimer’s disease (EOAD) in the literature (1, 2). Epidemiological data show that the number of global EOAD cases among adults aged 40–64 years more than doubled between 1990 and 2021, with substantial increases in its prevalence, incidence, and mortality (3).

Approximately 5–10% of EOAD patients carry autosomal dominant mutations (in APP, PSEN1, or PSEN2) that drive early cerebral Aβ aggregation, while more than 90% of EOAD cases are sporadic. The diagnosis of sporadic EOAD is delayed by approximately 1.6 years compared with that of elderly AD patients, making this population a major challenge in AD diagnosis and management (4). EOAD patients typically exhibit more prominent impairments in executive function, language, visuospatial processing, and other cognitive domains (5–9), which correlate with neurodegeneration in the posterolateral temporal lobe, posteromedial parietal lobe, frontal lobe, or occipital cortex. Neuropathological studies have also confirmed a “hippocampal-sparing” subtype of AD, which is predominantly observed in younger patients (10). Furthermore, compared with LOAD patients, EOAD patients have a more aggressive disease course, characterized by faster rates of cerebral atrophy, lower Mini-Mental State Examination (MMSE) scores, and more rapid cognitive decline (5, 11, 12). Existing studies suggest that higher tau burden and more severe neuroinflammatory responses in EOAD may underlie its aggressive clinical features (12–20).

Current AD research and clinical management frameworks are centered on typical LOAD. Reliable fluid and imaging biomarkers for LOAD have been fully validated, and artificial intelligence-assisted diagnostic technologies can effectively predict the pathological and clinical progression of individuals with mild cognitive impairment (MCI) or cognitively normal (CN) status (21–24). However, due to the unique age at onset and clinical phenotypic heterogeneity of sporadic EOAD, this population is usually excluded from large-scale observational and therapeutic clinical trials. This exclusion directly leads to delayed diagnosis, calibration bias of existing biomarkers in this population, and inability to extrapolate clinical trial results to sporadic EOAD patients. Ultimately, the pathological characteristics, disease progression patterns, and standardized diagnosis and treatment protocols for sporadic EOAD have not been systematically elucidated (25).

Neuroimaging techniques play an irreplaceable core role in EOAD research, for four key First, they enable in vivo visualization of abnormal alterations in cerebral structure, function, and molecular pathology, compensating for the scarcity of autopsy neuropathological data on sporadic EOAD. Second, they provide quantifiable and reproducible biomarkers of disease progression, overcoming the subjectivity and lag of clinical cognitive rating scales. Third, they can serve as in vivo visualization tools for targeted therapeutic targets and efficacy monitoring. Based on the above background, this paper systematically reviews the recent research progress in structural MRI, functional MRI, and PET imaging of sporadic EOAD. We aim to (i) focus on the neuroimaging characteristics of EOAD and their differences from those of LOAD; (ii) explore the potential pathophysiological mechanisms of EOAD in combination with existing imaging findings; and (iii) critically evaluate the limitations of current research and future research directions.

Methodology

This article is a narrative review. Literature retrieval was performed primarily based on the PubMed database, covering the publication period from 2010 to 2025. The search keywords included “early-onset Alzheimer’s disease,” “sporadic early-onset Alzheimer’s disease,” “logopenic variant of primary progressive aphasia “, “posterior cortical atrophy “, “MRI,” “BOLD,” “DTI,” “ASL,” “amyloid PET,” “tau PET,” “FDG PET,” “TSPO PET,” and “SV2A PET”.

The inclusion criteria were as (i) study subjects were patients with sporadic EOAD, including comparative analyses between EOAD and cognitively normal (CN) controls, early-onset non-AD dementia (EOnonAD), and LOAD; (ii) the core content of the study focused on neuroimaging (MRI/PET) biomarkers and their associations with fluid biomarkers; (iii) full-text original research (cross-sectional/longitudinal studies), high-quality systematic reviews, and meta-analyses published in English. Studies with larger sample sizes and standardized image acquisition and analysis methods were preferentially included, and studies on EOAD with different clinical phenotypes were also included to reflect the heterogeneity of the disease. Due to the overall limited sample size of EOAD studies, strict methodological quality scoring of the literature was not performed, but the methodological limitations of each study were addressed in the discussion section.

Magnetic resonance imaging (MRI)

MRI techniques enable non-invasive assessment of neurodegenerative changes in EOAD across multiple dimensions, including cortical thickness, white matter microstructure, brain functional networks, and cerebral perfusion. It is currently the most widely used imaging modality in clinical EOAD research.

Structural MRI

Structural MRI studies have consistently demonstrated cortical atrophy in EOAD patients in the inferior parietal lobule, superior parietal lobule, precuneus, middle temporal gyrus, inferior temporal gyrus, posterior cingulate gyrus, middle frontal gyrus, superior frontal gyrus, and fusiform gyrus. Notably, this characteristic cortical atrophy pattern across these 9 brain regions was not observed in EOnonAD group, indicating its specificity for differentiating EOAD from other early-onset cognitive impairment disorders (26, 27). Network-based characteristic gray matter atrophy patterns have also been observed in lvPPA (28–30) and PCA (28, 31, 32) (the language network in lvPPA and the visual network in PCA). Furthermore, compared with LOAD, EOAD patients exhibit significantly more severe atrophy in the inferior parietal lobule and precuneus. Based on findings from the Chinese Aging and Neurodegenerative Disease Initiative (CANDI) cohort, precuneus volume in the EOAD group was significantly negatively correlated with serum glial fibrillary acidic protein (GFAP) levels, while no such association was observed in the LOAD group (33).

However, there remains controversy regarding the laterality of the atrophy pattern in EOAD. Previous studies have reported lateralized atrophic changes in EOAD, particularly in the left temporal and parietal lobes (34), as well as interhemispheric differences in functional network connectivity (35). This discrepancy in laterality findings may be attributed to differences in the distribution of clinical phenotypes in the enrolled samples or variations in statistical power caused by different sample sizes across studies.

Most of the above studies are cross-sectional designs with limited sample sizes and single ethnic populations. Future studies need to perform longitudinal follow-up of EOAD patients stratified by clinical phenotype in larger cohorts, and analyze the associations between sensitive fluid biomarkers and regional brain alterations. Such research may improve the clinical applicability of the characteristic atrophy patterns of EOAD, to establish a more accurate diagnostic and subtyping system for the disease.

T2-weighted imaging

White matter hyperintensities (WMHs) appear as regions of relatively high signal intensity on T2-weighted images.

A study based on the Longitudinal Early-onset Alzheimer’s Disease Study (LEADS) cohort found that the spatial distribution of WMHs across regions in the left and right cerebral hemispheres was similar in the EOAD group, with the highest WMH volume in the frontal and parietal regions and the lowest WMH volume in the temporal lobe. Compared with the CN and EOnonAD groups, the EOAD group had a significantly higher mean WMH volume, with the most prominent differences observed in the frontal, parietal, and occipital regions. In addition, total WMH volume (sum across all regions) was significantly correlated with tau burden in the EOAD group (36). The regional distribution of subcortical WMHs differed between lvPPA and PCA, with these regional variations largely aligning with the regional patterns of neurodegeneration associated with these two syndromes (37, 38). Both lvPPA and PCA had greater subcortical and periventricular WMHs in the occipital lobe compared with the age-matched amnestic AD group.

Luo et al. investigated the differences in CSVD injury patterns between EOAD and LOAD patients and their impact on cognitive impairment using the peak width of skeletonized mean diffusivity (PSMD) (39). Compared with traditional CSVD markers such as WMHs, PSMD values showed the strongest correlation with most cognitive domains. However, the specificity of PSMD values in EOAD requires further validation.

It remains unclear whether increased WMH volume in AD is a manifestation of comorbid cerebral small vessel disease (CSVD) or a consequence of AD-related pathological processes. The mechanisms driving WMH are likely complex and involve an interplay between age-related vulnerability and neurodegeneration. Further studies are needed to investigate underlying mechanisms.

Diffusion tensor imaging (DTI)

DTI can visualize the anisotropy and structural integrity of white matter fiber tracts, and is highly sensitive to early microstructural damage in the cerebral white matter.

The comparability of results from existing DTI studies is limited by differences in regions of interest, voxel-based analysis, or tract-based spatial statistics methods used across studies. However, most studies consistently found that, compared with healthy controls, EOAD patients showed reduced fractional anisotropy (FA) in extensive white matter regions (posterior thalamus, genu of the corpus callosum) and increased mean diffusivity (MD) in posterior white matter regions (parietal and occipital lobes) on DTI (40). DTI analyses consistently reveal that lvPPA is characterized by bilateral, yet predominantly left-sided, microstructural alterations within frontally originating white matter pathways (including the superior and inferior longitudinal fasciculi and the uncinate fasciculus) as well as the parietotemporal junction (41–43), whereas PCA demonstrates predominantly right-sided white matter microstructural changes involving the superior and inferior longitudinal fasciculi, inferior fronto-occipital fasciculus, and right fronto-parietal pathways (44, 45). These impairments involve both long-range deep fiber tracts and short-range superficial fibers, disrupting the central hubs of information transmission between brain regions (46, 47). Meanwhile, compared with LOAD, white matter damage in EOAD patients is predominantly located in the posterior cerebral white matter (posterior cingulate gyrus and parietal lobe) and major fronto-parietal white matter pathways, with relative sparing of the medial temporal lobe (47).

Notably, regardless of the clinical phenotype of EOAD, patients exhibit more extensive white matter damage than would be predicted by the degree of gray matter atrophy, suggesting that white matter damage may be an early pathological change in EOAD that precedes cortical atrophy (48). However, this conclusion requires further validation in longitudinal studies.

Resting-state fMRI (rs-fMRI)

rs-fMRI assesses the strength of functional connectivity between brain regions by detecting low-frequency fluctuations in the blood oxygen level-dependent (BOLD) signal, providing an important perspective for investigating the early pathological mechanisms of the disease.

Existing cross-sectional studies have shown that EOAD is mainly characterized by reduced connectivity in the fronto-parietal networks (including the executive control network, salience network, language network, and high-order visual network) (35, 42, 49–53), rather than the reduced medial temporal lobe-hippocampal connectivity typical of LOAD (54–57). Longitudinal studies have demonstrated that brain regions with strong functional connectivity in EOAD have higher baseline tau-PET uptake and more rapid tau-PET accumulation over time. Functional connectivity of the core tau deposition regions can effectively predict the spreading pathway of tau protein. This association between connectivity and tau progression is highly consistent across EOAD patients with different clinical phenotypes (58).

In addition, although EOAD with different clinical phenotypes may involve distinct neural functional networks, tau deposition is consistently observed in the posterior nodes of the DMN (59, 60), and baseline tau burden in the DMN has important predictive value for cognitive decline in EOAD patients (61). However, differences in functional network parcellation methods, seed point selection, and connectivity definitions across studies mean that the reproducibility of these results remains to be verified.

Based on the network-based neurodegeneration theory, pathological proteins spread through the brain via network connections of the cerebral cortex in typical AD (62–65). Accordingly, the study hypothesized that these differences in network connectivity are likely related to the distinct distribution patterns of pathological proteins between EOAD and LOAD.

Taken together, further longitudinal studies tracking the dynamic trajectory of functional network changes and their association with tau distribution in EOAD patients with different clinical phenotypes across the preclinical to dementia stages of the disease will facilitate the development of accurate predictive models for EOAD.

Diffusion tensor image analysis along the perivascular space (DTI-ALPS)

Glymphatic system dysfunction, which leads to the accumulation of pathological proteins in the brain, is considered one of the potential pathogenic mechanisms of AD. The DTI-ALPS index is a novel quantitative marker for non-invasive assessment of the clearance function of the cerebral glymphatic system, calculated by the ratio of diffusivity perpendicular to the main fiber direction in specific white matter regions (projection fibers and association fibers) (66).

Existing studies have shown that the ALPS index is already abnormal before CSF Aβ42 reaches the positive threshold, and accelerates to decline after Aβ deposition (67). A retrospective cohort study by Yan et al. showed no significant difference in the ALPS index between the EOAD and LOAD groups. However, a significant correlation between the ALPS index and cognitive impairment (assessed by MMSE) was observed in the EOAD group, but not in the LOAD group. Furthermore, there was no significant correlation between the ALPS index and Aβ burden in the EOAD group (68), suggesting that glymphatic dysfunction in EOAD may involve factors other than Aβ plaque deposition, which remains to be further investigated.

Notably, a recent study demonstrated that the conventional ALPS index is significantly affected by the microstructural asymmetry of white matter fibers itself, which is associated with Aβ burden and age. After correction for this microstructural bias, the correlation between the ALPS index and Aβ burden disappeared, and its correlation with cognitive decline was also reduced (69). This finding indicates that the influence of white matter microstructure itself needs to be considered when interpreting the DTI-ALPS index.

Arterial spin labeling (ASL)

ASL is a non-invasive MRI technique that does not require contrast agent injection, and can quantify absolute cerebral blood flow (CBF). Verclytte et al. systematically compared cortical perfusion differences between EOAD patients with different clinical phenotypes, as well as between EOAD and LOAD, using ASL (70). The study showed that both amnestic and non-amnestic EOAD patients exhibited extensive hypoperfusion regions compared with healthy controls. Compared with amnestic EOAD, non-amnestic EOAD patients had lower perfusion in the bilateral temporoparietal lobes, precuneus, and anterior cingulate gyrus. Compared with the LOAD group, the EOAD group showed more severe hypoperfusion in the bilateral superior temporal gyri, bilateral frontal lobes, right anterior cingulate gyrus, and right precuneus (Figures 1, 2). However, the sample size of each group in this study was small, and the perfusion patterns of different EOAD clinical subtypes need to be verified in larger samples in the future.

Neurovascular coupling (NVC) reflects the interaction between local cerebral perfusion and neuronal activity within specific brain regions or functional networks. Combined BOLD and ASL techniques have been applied to NVC research, by calculating the correlation and ratio to reflect the synergy between cerebral perfusion and neuronal activity in each brain voxel (71). Longitudinal studies investigating the association between BOLD-ASL coupling, pathological proteins, and cognitive decline in the early stages of EOAD may help evaluate the diagnostic efficacy of BOLD-ASL coupling in the disease.

Positron emission tomography (PET)

PET techniques enable in vivo visualization of molecular pathological alterations, neuronal metabolism, neuroinflammation, synaptic density, and other core biological processes in the brain using specific tracers. They are critical tools for deciphering the pathological mechanisms of EOAD and exploring disease-specific biomarkers.

Aβ-PET and tau-PET

Increasing evidence indicates a spatiotemporal dissociation between Aβ plaques and tau protein during the course of AD. These two pathological proteins act synergistically in certain brain regions, while independently exerting effects in other regions to induce downstream neurodegenerative changes (72–74).

Some Aβ-PET studies have shown no significant difference in global Aβ burden between EOAD and LOAD (75, 76). However, other studies have reported higher Aβ burden in EOAD compared with LOAD (77–79). This inconsistency may be related to clinical differences in the study samples (APOE genotype, clinical phenotype), methodological differences (tracer type, partial volume correction, spatial normalization), and statistical power across studies.

Existing tau-PET studies have found that LOAD patients have higher tau burden in the basal forebrain and medial temporal lobe (Figure 3), while EOAD patients show higher tau uptake in neocortical regions, including the precuneus, inferior parietal lobule, and dorsolateral prefrontal cortex (13, 18, 19, 80). Patients with PCA have high tau-PET signal in the occipital and parietal cortex, while patients with lvPPA have a higher burden in the left temporo-parietal areas (20, 81–83). In addition, the association between tau pathology in the lateral temporal and occipitoparietal lobes and cognitive impairment is stronger in EOAD than in LOAD (84).

In correlation analyses between Aβ-PET and tau-PET, LOAD patients showed a significant positive correlation between tau standardized uptake value ratio (SUVR) in the parietal and temporal lobes and Aβ SUVR across the entire neocortex. In contrast, EOAD patients only showed weak local correlations between tau SUVR and Aβ SUVR in 7 brain regions, with no correlation observed in distant brain regions (84). A mediation effect analysis based on the LEADS cohort by Cho et al. showed that the effect of Aβ on cognitive scores (MMSE, Montreal Cognitive Assessment [MoCA], Clinical Dementia Rating Scale Sum of Boxes [CDR-SB]) was fully mediated by tau protein (85).

The differences in pathological protein deposition patterns between EOAD and LOAD are critical for the development of treatment strategies for these patient populations. Multiple clinical trials of anti-Aβ monoclonal antibodies have shown that significant reduction of Aβ plaques (detected by Aβ-PET) is associated with reduced concentrations of phosphorylated tau protein in plasma and CSF, slowed accumulation of neurofibrillary tangles (detected by tau-PET), and delayed progression of cognitive decline (18, 86). Currently, there are limited published data on the efficacy of anti-Aβ drugs in EOAD, but a recent exploratory analysis of the Lecanemab trial suggested that younger patients may have a poorer treatment response than older patients (87). This observational finding needs to be further explored in clinical trials specifically designed for EOAD and LOAD cohorts.

The spatial distribution patterns and interaction between cerebral Aβ and tau protein in EOAD remain unclear. The affinity and selectivity of different tracers for Aβ and tau vary, and most existing studies are cross-sectional designs, which limit the direct comparison of results across studies. Future studies need to perform longitudinal follow-up with standardized image acquisition and processing methods to clarify the spatiotemporal relationship between Aβ and tau in EOAD.

[^18^F]-FDG PET

[^18^F]-fluorodeoxyglucose ([^18^F]-FDG) PET indirectly reflects synaptic activity of neurons by quantifying the glucose uptake rate in brain regions.

Existing studies have shown that EOAD patients have lower cerebral glucose metabolism than LOAD patients, particularly with significant hypometabolism in the parietal cortex, which is more prominent in the left cerebral hemisphere (75, 77, 88–92). Even in patients with amnestic EOAD, reduced metabolism in the left precuneus and left supramarginal gyrus was observed compared with LOAD (93).

Vanhoutte et al. were the first to systematically compare the whole-brain metabolic patterns of different clinical subtypes of sporadic EOAD, after controlling for the effects of cortical atrophy (94). The language impairment subtype showed hypometabolism in Wernicke’s area, Broca’s area, insula, and pulvinar of the thalamus in the left cerebral hemisphere; the visuospatial impairment subtype exhibited hypometabolism in the bilateral parieto-occipital lobes, fusiform gyrus, and cuneus; the executive dysfunction subtype showed hypometabolism in the bilateral orbitofrontal cortex, dorsolateral and ventrolateral prefrontal cortex; and the typical amnestic subtype was characterized by hypometabolism mainly in the limbic system (Figure 4). In addition, the metabolic pattern of each subtype was linearly correlated with neuropsychological test scores of the corresponding cognitive domain, suggesting that the clinical heterogeneity of EOAD may be related to the differential susceptibility of distinct functional brain regions to the disease.

A study by Alice et al. found that EOAD patients with different CSF biomarker profiles exhibited distinct cerebral metabolic patterns. Elevated CSF tau levels were associated with hypometabolism in the bilateral frontal lobes and anterior cingulate gyrus, and the most prominent frontal hypometabolism was observed in the group with extremely high CSF tau levels (95).

Regional cerebral glucose metabolism is an early and progressive characteristic of AD. However, the relationship between cerebral metabolic changes and other pathological processes during the course of EOAD (such as Aβ plaque deposition, neurofibrillary tangle formation, neuroinflammation, and synaptic dysfunction) still needs to be further explored through integrated multimodal imaging.

TSPO-PET

Neuroinflammation mediated by microglial activation is recognized as one of the early pathological changes in AD (96–99). Existing studies have shown that microglial activity is higher in EOAD patients than in LOAD patients (100). The 18 kDa translocator protein (TSPO) is the most common target for inflammatory PET imaging, as it reflects the density of microglia in brain regions (101, 102).

A study by Tondo et al. combining ^18^F-FDG PET and [^11^C]-(R)-PK11195 PET (TSPO-PET) found that regions of reduced glucose metabolism in the inferior temporal gyrus, precuneus, angular gyrus, and inferior parietal lobule showed spatial overlap and a significant positive correlation with regions of dense microglial activation. Network analysis revealed a loss of long-range connectivity between this overlapping region and the frontal lobe, which was preserved in healthy controls (103). In addition, existing studies consistently found that neuroinflammation in sporadic EOAD is preferentially distributed along brain regions with strong functional connectivity, showing a certain spatial overlap with tau-PET imaging. However, the mechanism by which tau protein and neuroinflammation contribute to disease progression remains unclear (104).

It is important to note that the binding affinity of TSPO tracers is affected by the rs6971 polymorphism of the TSPO gene, which results in three binding high-affinity binder (HAB), medium-affinity binder (MAB), and low-affinity binder (LAB) (105). LAB individuals have low sensitivity to conventional TSPO tracers, which may introduce bias into studies. Appleton et al. used the next-generation tracer [^11^C]ER176 to achieve neuroinflammation imaging across all TSPO genotypes in EOAD for the first time, and found that the distribution of inflammation was more strongly correlated with tau deposition than with Aβ in EOAD patients at the mild cognitive impairment stage (106). Next-generation TSPO tracers are expected to provide more reliable tools for EOAD research.

TSPO-PET provides a unique perspective for in vivo studies of neuroinflammation in EOAD. However, it should be noted that upregulated TSPO expression is not a completely specific marker of microglial activation, and the results should be interpreted with caution. In addition, most current studies are cross-sectional, small-sample designs, and the conclusions need to be replicated and validated.

SV2A-PET

Synaptic loss is also one of the early pathological changes in AD (107, 108), and its correlation with cognitive decline is stronger than that of Aβ or tau burden (109–112). Synaptic vesicle glycoprotein 2A (SV2A) is a membrane protein expressed in synaptic vesicles of presynaptic axon terminals. [^11^C]-UCB-J is the most widely used SV2A-PET tracer in current research, for in vivo assessment of synaptic density.

One study showed that EOAD patients and frontotemporal dementia patients with GRN mutations (FTLD-GRN) had more significant reductions in synaptic density compared with LOAD patients (113).

Current SV2A-PET research on EOAD is still in its infancy. Whether the binding of SV2A tracers to SV2A reflects neuronal loss, reduced synaptic vesicles, or off-target binding to other pathological proteins still lacks neuropathological validation, which limits the interpretation of SV2A-PET imaging results. In addition, it remains unclear whether the quantitative relationship between SV2A density and synaptic number is consistent across different brain regions and pathological stages in EOAD patients. Future studies with standardized methods and multicenter longitudinal designs will help us understand the pathological mechanisms of EOAD.

Summary and outlook

This paper systematically reviews the research progress in multimodal neuroimaging of sporadic EOAD. Based on the existing evidence, the following robust findings have been identified regarding the neuroimaging characteristics of EOAD: (1) On structural imaging, compared with LOAD, EOAD shows more significant atrophy in the posterior cortex (precuneus, inferior parietal lobule) and more severe damage in the posterior cerebral white matter; (2) On functional imaging, EOAD is mainly characterized by reduced functional connectivity in the fronto-parietal networks; (3) On molecular and metabolic imaging, EOAD exhibits more severe and extensive tau burden and reduced cerebral glucose metabolism compared with LOAD. In addition, EOAD with different clinical phenotypes shows relatively characteristic differences in atrophy, metabolism, and network connectivity patterns, with a consistent spatial correspondence to clinical manifestations.

Although existing studies have initially outlined the framework of neuroimaging features of sporadic EOAD, there are still many key unresolved issues and limitations in current research. For example, what are the differences in pathological spreading pathways between sporadic EOAD and LOAD? Does tau-PET show alterations in the preclinical stage of sporadic EOAD? With the advancement of multimodal MRI and PET imaging, what role do fluid biomarkers play in disease diagnosis? How can they be integrated with neuroimaging to improve the sensitivity and specificity of disease diagnosis?

Furthermore, the vast majority of existing studies are cross-sectional designs, which can only identify correlations between imaging abnormalities, pathological changes, and cognitive impairment, but cannot clarify the causal relationship and temporal sequence of these alterations, lacking validation from long-term longitudinal studies. The neuroimaging characteristics and laterality differences of different clinical phenotypes still need to be further validated in large-sample studies with strict stratification by phenotype. Meanwhile, there is substantial heterogeneity in image acquisition parameters, post-processing methods, and quantitative standards across different studies, which directly leads to poor comparability and reproducibility of results across studies. Many imaging features identified in existing studies have not been systematically evaluated for their practical efficacy in early disease identification and differential diagnosis at the individual level, nor has their specificity for differentiating EOAD from other early-onset dementias such as frontotemporal lobar degeneration and dementia with Lewy bodies been clarified. As no universally accepted standardized acquisition and interpretation protocols for PET imaging have been established to date, we herein summarize the methodological characteristics of the core published PET studies in this field (Table 1).

Future neuroimaging research on sporadic EOAD should focus on two key directions. On the one hand, large-sample, multicenter, long-term follow-up longitudinal cohort studies (such as LEADS) are needed, with strict stratification of patients by clinical phenotype, to systematically track the dynamic trajectory of multimodal imaging changes from the preclinical stage to the dementia stage of the disease, and clarify the temporal relationship between imaging abnormalities, pathological progression, and cognitive decline. On the other hand, it is necessary to promote the deep integration of multimodal imaging with genomics, fluid biomarkers, and cognitive phenotypes, to construct a multi-dimensional feature system of the disease, and deeply decipher its underlying pathophysiological mechanisms. These efforts will provide reliable imaging tools for the early identification, precise subtyping, individualized treatment, and efficacy monitoring of sporadic EOAD.