Asymmetry between the right and left hippocampus and amygdala is considered a biomarker of SCD in numerous voxel analysis studies (Yue et al., 2018). In addition to asymmetry between cerebral areas, significant reductions in hippocampal volume in individuals with SCD at baseline have been reported in other studies (Hafkemeijer et al., 2013; Kim et al., 2013; Rogne et al., 2016; Sánchez-Benavides et al., 2018). Identical results have been reported in longitudinal follow-up studies, indicating that the hippocampal volume decreases at a rate of 1.9% per year in people with SCD (Cherbuin et al., 2015; Nunes et al., 2010). Hippocampal atrophy manifests mainly as a significant reduction in the volume of gray matter in the tail of the bilateral hippocampus and enlargement of the bilateral paracentral lobules (Liang et al., 2020). However, other studies have not reported significant hippocampal atrophy (Platero et al., 2019). For example, a study revealed significant cortical atrophy in their bilateral parahippocampus and perirhinal and the left entorhinal cortices (Fan et al., 2018). The other study demonstrated that the volume of the entorhinal cortex in individuals with SCD was smaller than normal cognitive participants (Ryu et al., 2017). These differences across studies may be related to the heterogeneity among individuals included in those studies, which in turn may be related to the sources from which these subjects were recruited (Pini and Wennberg, 2021) and the enrolment criteria (Perrotin et al., 2017) applied in the studies. However, the hippocampal volume is insufficient and lacks specificity as an independent diagnostic basis for SCD. For example, patients with depression may also present with reduced hippocampal volume and CI (Frodl et al., 2002; Sheline, 2003). Furthermore, the hippocampus is highly susceptible to ageing and ageing-related changes (Bettio et al., 2017). In addition, several studies have investigated differences in volume in numerous subcortical areas, including the cholinergic basal forebrain nuclei and hippocampal subregions, between individuals with SCD and normal cognitive controls (NCs). These findings consistently suggest that SCD is associated with a significant reduction in the cholinergic basal forebrain and hippocampal CA1 volumes with respect to NCs (Cantero et al., 2016; Perrotin et al., 2015; Scheef et al., 2019; Zhao et al., 2019).

Other methods for identifying and evaluating SCD have focused mainly on surface-based analyses. In addition to reduced subcortical volume, surface-based analyses have demonstrated that a thinner cortex, particularly in the temporoparietal lobe, is associated with faster memory deterioration and an increased risk of disease progression in patients with SCD than in NCs (Meiberth et al., 2015; Schultz et al., 2015; Verfaillie et al., 2016; Verfaillie et al., 2018). Another study revealed that both SCD and MCI patients presented with a reduced volume in the left inferior parietal lobe (IPL), while SCD patients also presented with morphological changes in the right inferior temporal gyrus (ITG), right insula, and right amygdala (Song et al., 2024). Another study showed that compared with NCs, patients with MCI presented with predominantly left-sided surface morphology changes across various brain regions, including the transverse temporal gyrus, superior temporal gyrus, insula, and pars opercularis. Patients with SCD, meanwhile, exhibited relatively slight surface morphological changes, mainly in the insula and deltoid (Yang J. et al., 2023). A prospective study revealed that larger volumes and lower amyloid loads in the right and left parietal lobes at baseline were associated with a greater probability of cognitive recovery in SCD patients, providing clinicians with key factors associated with cognitive improvement (Na et al., 2024). In another study, the overall amyloid loads of the right hemisphere and specifically the right temporal lobe cortex of amyloid-positive SCD patients were negatively correlated with the gray matter volume, providing a pathological perspective for validating SCD as a preclinical stage of AD (Wang et al., 2022). We summarized the main studies of structural MRI studies in SCD in Table 1.

Notably, a growing number of studies have measured gray matter volumes among brain regions by employing a structural covariance network approach (de Schipper et al., 2017; Drenthen et al., 2022; Watanabe et al., 2020), in which the gray matter covariance is examined by mapping gray matter correlations across the brain to seed regions. The structural covariance approach is essentially a correlation analysis of cross-sectional morphometric imaging data in which gray matter shrinkage is measured in common processes between brain regions (Alexander-Bloch et al., 2013). Gray matter covariances may provide more neural information than the volumes of individual brain regions. For example, structural covariances have been reported to reflect the effects of brain development and structural plasticity (Duan and Wen, 2023). Studies on structural covariance patterns have revealed reduced structural covariance (Tsai et al., 2023) and weakened connectivity strength in SCD patients (Fu et al., 2021). Furthermore, reduced gray matter volumes in the bilateral ventral lateral prefrontal cortex and right insula in SCD patients relative to NCs have also been identified. Take the above-mentioned area as the ROIs, there was an abnormal structural association between the left ventrolateral prefrontal cortex (vlPFC) and regions, including the left anterior cingulate cortex (ACC), right vlPFC, left insula, and right dorsolateral prefrontal cortex (dlPFC) in the patients with SCD compared to NCs. The left ACC, left middle occipital cortex (MOC), bilateral fusiform and bilateral precuneus (PCUN) showed decreased structural covariance with the right vlPFC in patients with SCD compared to NCs. For the right insula ROI, decreased structural covariance was observed between the right insula and regions including the right median cingulate cortex (MCC), right PCUN, right MOC, left hippocampus, left thalamus, and right ACC (Xu et al., 2022). Since the pathology of AD (such as Aβ and Tau deposition) may affect the distal brain regions through neural connections, leading to the disruption of the covariance network, the decreased covariance association may predict the early damage of the memory-related network. In addition to traditional methods of analysis, an increasing number of studies are employing novel analytical modalities to process preclinical AD imaging findings, opening new avenues for exploring pathophysiological alterations in AD.

VBM-based studies have revealed that an abnormal reduction in hippocampal volume can be observed 8 years prior to a clinical diagnosis of AD (Jia et al., 2024). Furthermore, the hippocampus continues to atrophy as the disease progresses, making its volume one of the most well-established imaging biomarkers of the AD disease spectrum (Bayram et al., 2018). Similarly, atrophy of the hippocampus has been observed in patients with MCI, in addition to substantial atrophy of the internal olfactory cortex (Pennanen et al., 2004; Teipel et al., 2006), which may be related to the pathological mechanisms underlying cognitive dysfunction. In addition, atrophy of the nucleus accumbens and the right caudate nucleus has been reported to mark the transition from MCI to AD, whereas volume reductions in the right thalamus and bilateral caudate nucleus have been associated with CI in AD patients (Tuokkola et al., 2019). In conclusion, we can use neuroimaging techniques to assess the volume changes of voxels that are significantly associated with MCI. Among them, structural MRI has demonstrated the greatest benefits, as it can achieve excellent soft tissue contrast and high spatial resolution, which are important for detecting AD. Although structural changes revealed on sMRI cannot be used as a specific feature of MCI, they provide a biased basis for the evaluation of this condition. The combination of sMRI and automated diagnostic methods provides a new perspective for predicting the transformation of MCI to AD (Garg et al., 2023).

SBM-based analyses, meanwhile, have suggested that cortical thickness is associated with AD severity and CI (van Oostveen and de Lange, 2021). Gray matter atrophy, mainly in the medial temporal lobe and subsequently extending along the temporo-parietal-frontal lobes to the remaining cortex, has been reported in patients with early AD (Pini et al., 2016). Researchers have suggested that temporal occipital cortex atrophy could be a sensitive and specific marker of pathological changes in AD (Mao et al., 2022). Medical researchers have compared the temporal lobe volumes of NCs and MCI and AD groups and reported that compared with those of the NCs, the hippocampal volumes of the MCI and AD groups were significantly reduced, whereas temporal lobe neocortical atrophy was identified only in the AD group (Convit et al., 1997). Follow-up studies have revealed that MCI patients who progress to AD are more likely to gradually develop temporal neocortical atrophy than patients with stable MCI, suggesting temporal neocortical atrophy may be a biomarker of disease progression (Convit et al., 2000). Fellow Li reported that reductions in cortical thickness in individuals with MCI starts in the temporal lobe, but the range of thickness changes gradually expands as the disease progresses. In addition, reductions in cortical thickness have been positively associated with reduced scores on the Mini-Mental State Exam (MMSE). Thus, changes in temporal cortex thickness may reflect alterations in cognitive functioning in patients with AD (Li et al., 2022). A longitudinal study revealed that the bilateral temporal cortex was significantly thinner in AD patients than in NCs, whereas this difference was less pronounced between MCI patients and NCs (Bachmann et al., 2023). In addition, Yuanlin revealed that the gray matter volumes in the left superior and right middle temporal gyri were lower in AD patients than in MCI patients, suggesting that gray matter atrophy in the lateral temporal lobe may help differentiate MCI and AD (Yuanlin et al., 2021). A large body of evidence suggests that atrophy of the medial temporal lobe and changes in the cortical thickness of the lateral temporal lobe are early biomarkers of AD-associated CI (Table 2).

sMRI analysis based on the calculation of surface area is also frequently employed as a strategy in AD-related CI studies. According to Rechberger and colleagues, the surface areas of the temporal lobe, superior frontal gyrus, and anterior cingulate cortex in the AD group are significantly lower than those in the NC group, which suggests that a decrease in the surface area of the cerebral cortex may be related to the reduction in cognitive function observed in AD patients (Rechberger et al., 2022). The cerebral cortex is well characterized by substantial gyrification; consequently, the cortical surface is mostly hidden in the sulci and lateral fossa. Recent studies have shown that the fractal dimension (FD), gyration index (GI), and sulcus depth (SD) can be used as valuable measurements of the complexity of cortical folding (Qin et al., 2022). Other studies have indicated that the degree of cortical folding is related to the cognitive functioning of older adults (Serra et al., 2022). A longitudinal study revealed that reductions in the FD, GI, and SD were associated with cognitive decline in AD patients. Furthermore, overall changes in the FD, GI, and SD were more pronounced in patients with AD than in those with amnestic MCI (aMCI), whereas changes in the FD were associated with cognitive ability only in AD patients but not in aMCI patients (Bachmann et al., 2023). As shown in another study, a greater insular cortical GI is strongly associated with better memory function and semantic fluency in AD patients, whereas this association was much less pronounced in MCI patients and was not present in NCs. This could suggest that as the disease progresses, the increasingly atrophic medial temporal lobe is unable to maintain memory function, requiring the insular cortex to gradually take it over (Núñez et al., 2020).

In conclusion, morphology-based analyses suggest that hippocampal atrophy, as well as medial temporal lobe atrophy, is a common pathological change in PPAD and that the degree of atrophy increases as SCD progresses to MCI. Moreover, as the patient’s cognitive deficits worsen, cortical atrophy advances from the temporal lobe to the parietal lobe, then frontal lobe, and finally spreads to the entire cerebral cortex. In addition, asymmetry of the hippocampus and amygdala, cholinergic basal ganglia, nucleus accumbens, caudate nucleus, and thalamus are potential biomarkers of preclinical AD. Despite its relative novelty, structural covariance analysis has been gradually applied to preclinical imaging analysis of AD (Alexander-Bloch et al., 2013; Montembeault et al., 2016; Seeley et al., 2009). Compared with other analytical methods, it can better analyze the correlations between the morphologies of different regions of the brain, providing richer evidence of early pathological changes in AD. However, there are still some obstacles to its application. For instance, there are gaps in the comparability between different sample datasets and the reliability of the results for the same subject at different periods and image resolutions. Additional studies are needed to explore and verify the reproducibility of methods based on structured correlation analysis (Table 2).

Diffusion tensor imaging (DTI) is commonly used to study the microstructural changes in WM in patients with neurodegenerative diseases, including axonal loss, damage, or demyelination (Alexander et al., 2007). DTI images are usually characterized by parameters including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD). FA is the most commonly used metric in evaluating DTI data and reflects the directionality and integrity of WM fibers; a reduction in the FA value indicates that the integrity of the WM has been compromised (Chung et al., 2011). The AD represents the diffusion rate of water molecules in the direction of the main axis and can reflect the growth of axons; this value decreases when axons are damaged. The RD represents the diffusion rate of water molecules in the direction perpendicular to the main axis of diffusion and can reflect the formation of myelin sheaths; the value increases as the amount of myelin in the sheaths decreases. Finally, the MD represents the overall diffusion of molecules regardless of direction; an increase in this value suggests an increase in the content of free water molecules in the tissue. DTI analytical techniques can be divided into traditional voxel-based analysis (VBA) and tract-based spatial statistics (TBSS).

Both VBA and TBSS allow the spatial localization of WM fiber bundles for subsequent quantitative analysis. In VBA, the individual diffusion information is first registered to a standard spatial template, which is then segmented on the basis of spatial location via a WM atlas to obtain quantitative information on the corresponding fiber bundles. The VBA method is intuitive, convenient and is not greatly influenced by human subjective, but the results are highly dependent on the accuracy of the registration; if it is incorrect, false positives may occur, and individual differences may be underestimated (Shu et al., 2009). To overcome the shortcomings related to registration in VBA and smooth kernels without uniform values (Jones et al., 2005), several scholars (Smith et al., 2006) have proposed the TBSS analysis method. First, a skeleton is formed in standard space on the basis of the FA chart of each subject, and then the individual dispersion index is projected onto the skeleton for point-by-point statistical analysis. In addition, in TBSS, a WM segmentation template can be applied to obtain the dispersion index of each fiber bundle. TBSS can be performed automatically by some software, resulting in a rapid and convenient analysis. However, this method is not sensitive to the direction of the fibers. Owing to individual differences in the positions of WM fibers and the existence of cross-fibers, mutual interference from fiber bundles is not uncommon in the results.

Although DTI is currently the most widely used technique worldwide for studying the microscopic structural characteristics of WM in the central nervous system (Palacios et al., 2020), it has limitations such as the inability to effectively display crossed or branched fibers, and is susceptible to interference from CSF and free water (Andica et al., 2020). Therefore, a series of DTI derivative techniques have emerged based on the differences in diffusion direction and speed of water molecules in different structures, including Diffusion Kurtosis Imaging (DKI), Free Water DTI (FW-DTI), and Neurite Direction Dispersion and Density Imaging (NODDI).

A growing number of studies have combined multimodal neuroimaging techniques with PET and MRI to observe AD-induced pathological changes in SCD patients (Chételat, 2018). Abnormal amyloid deposition has been identified as a key factor in triggering downstream neurodegenerative cascade responses (Chen et al., 2024; Song et al., 2022; Yan et al., 2024). Between-group analyses revealed that SCD subjects with higher levels of amyloid deposition presented with significantly greater gray matter atrophy than those with lower levels of amyloid deposition (Chételat et al., 2010). A disease severity index has been developed from multivariate analyses involving both amyloid PET and structural MRI; this index may identify individuals with AD-like patterns of SCD as an appropriate risk population (Ferreira et al., 2017). In addition, researchers have combined amyloid PET, FDG-PET, and structural MRI data to identify pathological patterns corresponding to different points along the AD continuum. The results revealed three different imaging biomarker patterns present in patients with different stages of AD (Wirth et al., 2018). A longitudinal study indicated that reductions in cognitive performance over time were associated with cerebral hypometabolism in the precuneus at baseline without gray matter atrophy according to the joint use of FDG-PET and MRI (Scheef et al., 2012). Patients with clinical SCD have been shown to have greater progression of gray matter atrophy over time than patients with community SCD, suggesting that clinical SCD may represent a greater risk for dementia due to AD (Kuhn et al., 2019).

In addition to observing cerebral WM degeneration in SCD patients, DTI can be used to study structural and functional changes in the brain of SCD patients in depth from multiple perspectives when combined with other techniques. The accumulation of β-amyloid Aβ and tau neurofibrillary tangles in the brain is a typical pathological feature of AD (Scheltens et al., 2021). Thus, studies of MCI and SCD patients with DTI combined with tau PET have confirmed a link between tau protein abnormalities and WM degeneration (Wen et al., 2021). Interactions between gray and WM neurodegeneration can also be identified with sMRI combined with DTI, confirming that more subjective cognitive complaints are associated with smaller volumes of the hippocampus, frontal lobes, temporal lobes and insula and an increased WM MD (Cedres et al., 2021). Combined fMRI and DTI analyses revealed that brain connectivity and excitability and WM integrity were significantly lower in patients with SCD than in NCs (Fogel et al., 2021). In another study, the accuracy of DTI in discriminating β-amyloid-positive patients with MCI from β-amyloid-negative controls was 80%. However, patients with SCD only presented with spatially restricted WM alterations in the anisotropic modality, and DTI lacks the ability to identify β-amyloid-positive and β-amyloid-negative patients with SCD (Teipel et al., 2019). The studies of multimodal MRI in SCD also been summarized in Table 6.

A study combining 18F-flortaucipir PET and sMRI revealed that the whole-brain 18F-flortaucipir standardized uptake ratio (SUVr) was significantly correlated with the levels of cerebrospinal fluid t-tau and p-tau; the strongest correlations were found in temporal regions (as confirmed with voxel analyses), suggesting that PET is a useful tool for visualizing pathological changes in AD. The metrics derived from 18F-flortaucipir PET were also associated with CSF β-amyloid levels, situational memory, visuospatial working memory, and cortical thickness, regardless of hippocampal volume (Okafor et al., 2020). Another study, involving the use of 18F-FDG PET/MRI (specifically T1-weighted MRI), revealed that the posterior cingulate cortex and precuneus demonstrated reduced metabolic activity in patients with MCI and AD, whereas the atrophy in these regions was not significant. These results suggest that the uptake of 18F-FDG in these two regions may be a marker for early AD (Bailly et al., 2015). A study combining automated MRI-based brain volume measurements and THK-5351 PET revealed that the isthmus of cingulate gyrus and IPL are important regions for distinguishing AD patients from NCs and MCI patients, suggesting the potential discriminatory value of this combined method (Kim et al., 2021). Other researchers have combined MRI and PET/CT for head-to-head multimodal imaging with 18F-FDG and 18F-AV45 to visualize and quantify brain morphology, glucose metabolism, and amyloid levels, identifying the AV45/FDG/NVol index as a novel quantitative molecular imaging biomarker associated with the clinical neurocognitive status of the patient (Thientunyakit et al., 2020).

Another study based on sMRI and fMRI revealed a significant reduction in gray matter volume in frontotemporal parietal structures as well as impaired resting functional connectivity in the default mode network (DMN) and executive network in patients with aMCI. Moreover, the abovementioned methods were reported to distinguish aMCI patients from healthy ageing individuals, and the results were considered to constitute a neuroimaging marker of early dementia (Bharath et al., 2017). The use of 18F-PI-2620 PET and T1-weighted MRI for MCI patients with positive β-amyloid-PET visual findings and follow-up for 1 year revealed that increased tau deposition in the fusiform gyrus was associated with high levels of p-tau and t-tau in the CSF. However, a low hippocampal volume was associated with increased tau loading at baseline, with a significant increase in tau loading observed in the temporal cortex, fusiform gyrus, and inferotemporal cortex after 1 year; however, the change in amyloid load was not significant (Bullich et al., 2022).

Diffusion tensor imaging (DTI) and resting-state fMRI were combined to assess brain function in AD and MCI patients and NCs. Patterns of interhemispheric functional connectivity were measured with voxel-mirror-homotopic connectivity (VMHC), and a reduction in VMHC was observed in forebrain regions, including the prefrontal cortex and subcortical regions, of AD and MCI subjects. DTI analysis further revealed that the most significant difference among the three cohorts was in the FA in the genu of the corpus callosum, which was positively correlated with VMHC in the prefrontal and subcortical regions. In addition, the diffusion parameters in the aforementioned brain regions were significantly correlated with metrics of cognitive performance. These results suggest a specific pattern of interhemispheric functional connectivity changes in AD and MCI patients that is significantly associated with changes in the structural integrity of the WM along the midline (Wang Z. et al., 2015). We have summarized the studies of multimodal MRI in MCI in Table 7.

Risk genes also have an impact on the structure of the brain. A study examined cross-sectional AD biomarkers for participant with MCI (n = 930), which found number of ApoEε4 alleles was associated with smaller hippocampal and amygdala volumes and poorer cognition (Hobel et al., 2019). BIN1, a key AD susceptibility gene after APOE, has higher expression in AD. After adjusting for confounding factors, the association between the BIN1 rs10200967 genotype and left hippocampal-amygdaloid transition area atrophy significant in the MCI (Luo et al., 2025). The follow-up study found that CTCF, UQCR11 and WDR5B gene had the most significant correlations with left paracentral lobule and sulcus and right subparietal sulcus thickness (Wang et al., 2021). In addition to exploring morphological, functional and pathological changes in the brain, MRI can compare the impact of genetic factors on the brain among different populations.

With the emergence of β-amyloid-PET and tau-PET, imaging observations of AD have shifted to the molecular pathological level. Furthermore, the combination of PET and MRI has provided a new way to spatially visualize pathological changes in AD and more data on the pathogenesis associated with PPAD. Thus, multimodal neuroimaging techniques have shown great advantages in exploring the relationships among different AD biomarkers, but further multimodal neuroimaging studies of SCD are needed.