Introduction

Cognitive impairment is one of the primary public health challenges in our era (Hampel & Lista, 2016). Vascular-related cognitive decline is the second most prevalent cause, and in East Asia, it may take precedence as the primary cause (Iadecola et al., 2019). Subcortical vascular mild cognitive impairment (svMCI) is a type of cognitive impairment caused by cerebral small vessel disorder (Zotin et al., 2021). svMCI is characterized by a gradual onset and is mainly associated with attention and executive function impairment (Hamilton et al., 2021; Zotin et al., 2021). As the prodromal stage of subcortical vascular dementia, svMCI is the most responsive subtype to treatment through modification of vascular risk factors (Iadecola et al., 2019; Van Der Flier et al., 2018), signifying the early phase of the disease. As a result, identifying svMCI at an early stage is of great importance in clinical practice. The mechanisms underlying cognitive impairment associated with subcortical vascular lesions need to be largely explored.

Hippocampal atrophy is a common structural and pathophysiological change in various cognitive dysfunctions, such as Alzheimer’s disease (AD) (Devanand et al., 2007), post stroke dementia, and subjective cognitive decline (Das et al., 2019; Devanand et al., 2007; Huang et al., 2022). Moreover, dynamic hippocampal atrophy has been observed during healthy human aging (Oschwald et al., 2019). Neurofibrillary tangle deposition and neuronal death are linked to hippocampal shrinkage, which is a significant cause of cognitive impairment in vascular illness (Yamamoto et al., 2021) and is linked to hippocampal atrophy. Even minor vascular disorders can result in hippocampus atrophy and cognitive deterioration (Han et al., 2020). However, the hippocampus was considered a single, homogeneous structure in most previous studies, and potentially useful information about its sub-regions has been discarded. Several distinct sub-regions with various structural and functional profiles have been identified in the bilateral hippocampus (Fraser et al., 2015). Among these, a statistical atlas of the hippocampal using ultra-high resolution ex vivo magnetic resonance imaging (MRI) combined with in vivo data in the FreeSurfer 6 and above version (https://surfer.nmr.mgh.harvard.edu/fswiki/rel7downloads) has been widely adopted to investigate structural alterations of the hippocampal at the sub-regional level in many brain disorders (Bai et al., 2019; Christidi et al., 2019; Xu et al., 2022). Previous studies have demonstrated that selective atrophy of hippocampal sub-regions is associated with particular brain diseases. For example, the cornu ammonis 2/3 subfield and the hippocampus-amygdala transition area are the most affected regions in amyotrophic lateral sclerosis in contrast to AD, where the presubiculum and subiculum are the most vulnerable regions (Christidi et al., 2019). Moreover, some studies reported that patients with mild cognitive impairment showed smaller gray matter volumes, primarily in cornu ammonis CA1-2 volumes, compared to healthy controls (Mueller et al., 2010). Considering the various etiologies of svMCI, whether there is a selective vulnerability of hippocampal sub-fields is extremely worth to be further explored.

Consequently, we hypothesized that patients with svMCI would exhibit distinctive atrophy patterns of volumetric changes in hippocampal sub-regions. To test this hypothesis, we obtained T1 images from 34 svMCI patients and 34 normal controls, and performed hippocampal segmentation of these T1 images using FreeSurfer 7.1. The gray matter volume (GMV) of the hippocampal sub-regions was measured and compared between groups after adjusting for age, sex and educational level. Finally, Pearson’s correlation was performed between changed GMV and MoCA in the svMCI.

Methods and materials

Participants

This research was conducted in the Shenzhen Traditional Chinese Medicine Hospital and authorized by their Institutional Review Board (K2021-041). The diagnosis of MCI was made by two neurological physicians with more than five years of experience (XDQ and JJW) based on Peterson’s criteria as follows (Jia et al., 2016; Petersen et al., 1999). The inclusion criteria required cognitive complains in memory and/or other cognitive domains for at least three months, objective cognitive impairments not meeting the DSM-IV criteria for dementia, an abnormal clinical dementia rating (CDR) of ≥ 0.5 on at least one domain and a global score ≤ 0.5 (Saygin et al., 2017), a Montreal cognitive assessment (MoCA) score of < 26 (Nasreddine et al., 2005; Sun et al., 2023), and normal or near-normal performance of daily life activities. The following cases were excluded from the study: individuals with a history of psychiatric disorder in any two lines of first- to third-degree biological family; those with a history of strokes or transient ischemic attack within three months; those with a history of seizures, schizophrenia, or major depressive disorder; those with an inherited or inflammatory small vessel illness; those who previously suffered from a head injury with unconsciousness; those who used nootropics, such as Donepezil and Kabalatine; those with significant clinical or surgical conditions; those with physical disabilities, including aphasia, blindness, or hemiplegia, which can prevent the fulfillment of neuropsychological tests; and those with contraindications for MRI.

Patients with svMCI needed to meet at least one of the following criteria: three or more supratentorial subcortical minor infarcts with or without white matter lesions of any severity; moderate to severe white matter lesion (score ≥ 2 on the Fazekas rating scale) with or without small infarcts; or one or more subcortical minor infarcts strategically positioned in the caudate nucleus, globus pallidus, or thalamus. The evaluation of MRI images was done by two experienced radiologists (JXC and HQL), who were unaware of the participants’ identities. In cases of disagreement, they would re-evaluate the images after discussion.

The NCs were recruited through advertisements from the community of Shenzhen and had no prior history of neurological or psychiatric disorders, no cognitive issues, and their conventional brain MRI images were normal. Herein, 34 svMCI and 34 NCs were involved. The demographics and clinical characteristics of all participants are presented in Table 1.

Table 1 Demographic data and clinical measures
Full size table

MRI acquisition

MRI scanning was performed on a 3T scanner (GE medical system, MR750) with a 32-channel head coil. The scanning criteria of T1 images were as follows: repetition time = 8.656 ms, echo time = 3.22 ms, inversion time = 450 ms, flip angle = 12, matrix size = 256 × 256, slice thickness = 1 mm, voxel size = 1 × 1 × 1 mm3, and sections = 152.

Image processing

Utilizing the usual “recon-all” process in FreeSurfer 7.1 (http://surfer.nmr.mgh.harvard.edu/), the cortical surface was reconstructed. Overall, the pictures were preprocessed by movement correction, brain retrieval, Talairach transformation, intensity correction, and segmentation of gray matter, white matter, and cerebrospinal fluid from the brain tissue. Additionally, the barrier between gray matter and white matter was tessellated. The segmentation of subcortical structures was further explored using a nonlinear warping atlas.

Hippocampal segmentation

The hippocampal segmentation was performed using the automated hippocampal module in FreeSurfer 7.1. Specifically, it was carried out using a Bayesian inference technique that employed a probabilistic atlas of the hippocampal formation that was trained on a hybrid data set involving in-vivo ultra-high-resolution (0.1 mm) MRI and ex-vivo autopsied brain MRI from multiple subjects. This probabilistic atlas was registered to T1 images in the individual space for each subject. The published works of Saygin et al. (Saygin et al., 2017), Iglesias et al. (Iglesias et al., 2015), and Van Leemput et al. (Puonti et al., 2016) provided extra details on the segmentation technique. To evaluate the quality of these segmentations, all segmented volumes of participants were visually inspected by a trained operator (JXC). Moreover, hippocampal subfield volumes were plotted and outliers were flagged for a secondary inspection.

Gray matter volume alterations of hippocampal sub-regions

Two-sample t-tests were employed to identify volume shifts in hippocampal sub-regions adjusting for age, sex, education, and estimated total intracranial volume (eTIV) (Jack et al., 2015; Perrotin et al., 2015) utilizing SPSS Statistics (Armonk, NY: IBM Corp., v.19.0). p < 0.05 was judged statistically significant, not corrected.

Correlation analyses

Pearson’s partial correlation analyses between the volume and MoCA scores of all hippocampal sub-regions that exhibited gray matter atrophy were performed with age, sex, eTIV, and educational level as covariates. The p < 0.05 was judged statistically significant, not corrected. These analyses were performed using SPSS 25.

Results

Demographics and clinical data

Demographics and clinical data (Table 1) were compared between the two groups using a two-sample t-test. Sex was analyzed using χ2. Consequently, we found no significant variation in age (p = 0.137), sex (p = 0.807), education (p = 0.072), or eTIV (p = 0.296), but there was a significant difference in MoCA between svMCI and NCs (p < 0.001).

Hippocampal segmentation

Hippocampal segmentation resulted in 19 sub-regions for each hemisphere (Fig. 1): the hippocampal tail, subiculum body, subiculum head, cornu ammonis (CA) 1 body, CA1 head, CA3 body, CA3 head, CA4 head, CA4 body, hippocampal fissure, presubiculum head, presubiculum body, parasubiculum, molecular layer hippocampal (ML-HP) head, ML-HP body, granule cell molecular layer of dentate gyrus (GC-ML-DG) head, GC-ML-DG body, fimbria, and hippocampus-amygdala-transition-area (HATA).

Fig. 1
figure 1

Segmentation results of hippocampus on T1 MRI scans. The images were shown with FSLeyes (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLeyes). Abbreviations: HATA, hippocampus-amygdala-transition-area; CA, cornu ammonis; GC-ML-DG, granule cell molecular layer of the dentate gyrus; ML-HP, molecular layer hippocampus

Full size image

GMV shifts of hippocampal sub-regions

In contrast to NCs, patients with svMCI demonstrated significantly reduced GMV in several left hippocampal sub-regions, including the left hippocampal tail(puncorrected = 0.004), left hippocampal fissure (puncorrected = 0.040), left CA1 head(puncorrected = 0.011), left ML-HP head(puncorrected = 0.011), left CA4 head(puncorrected = 0.033), and left CA3 head(puncorrected = 0.011), but only in the hippocampal tail in the right hippocampal sub-regions(puncorrected = 0.016) (Tables 2and Fig. 2).

Table 2 Gray matter volume alterations in hippocampal sub-regions in the svMCI compared to NC
Full size table
Fig. 2
figure 2

The gray matter volume changes of hippocampal sub-regions. Two-sample t-tests were employed to identify volume changes in hippocampal sub-regions adjusting for age, sex, education, and estimated total intracranial volume (eTIV) between patients with svMCI and NC.* represents significant results with puncorrected< 0.05

Full size image

Results of correlation analyses

The GMV in the left CA3 head was negatively correlated with MoCA scores in the svMCI (r = -0.468, p = 0.005, Fig. 3). However, no significant correlations were observed between GMV of other altered hippocampal sub-regions and MoCA scores in the svMCI (Table 3).

Fig. 3
figure 3

Relationship between gray matter volume and cognitive impairments in patients with svMCI. The gray matter volume of left CA3 head was adjusted by regressing out covariates of age, sex, eTIV, and educational level. Then, Pearson’s correlation analyses between the volume of left CA3 head and MoCA scores in the patients with svMCI were performed. Abbreviations: MoCA, Montreal cognitive assessment

Full size image
Table 3 Results of correlation analyses
Full size table

Discussion

Herein, we identified that the svMCI group presented hippocampal atrophy primarily in the right hippocampal tail and several left hippocampal sub-regions, such as the hippocampal tail, hippocampal fissure, CA1 head, ML-HP head, CA4 head, and CA3 head. Moreover, the GMV in the left CA3 head was negatively associated to the MoCA scores in patients with svMCI. Our results confirm a selective vulnerability of hippocampal sub-fields in patients with svMCI, which adds to our understanding of the pathophysiological mechanisms of cognitive impairment with subcortical vascular origins.

Notably, the findings of this study have demonstrated a discernible susceptibility of bilateral hippocampal sub-regions, with a notable emphasis on the left. Consistent with this finding, a previous study reported that hippocampal atrophy in the svMCI group was more lateralized towards the left hemisphere, particularly in the left subiculum and presubiculum (Li et al., 2016). Furthermore, the shape of the left anterior hippocampus was reduced in patients with svMCI, while those with moderate-to-severe clinical severities had a reduction in the shape of both anterior hippocampi (Thong et al., 2014). Previous research with more severe cognitive impairments reported more remarkable morphological alterations in the right hippocampal sub-regions (He et al., 2022). In healthy elderly adults, the right hippocampus tends to exhibit a larger volume than the left hippocampus, with the left hippocampus being more susceptible to vascular pathology than the right (Hou et al., 2013; Sarica et al., 2018; Shi et al., 2009; Woolard & Heckers, 2012). Furthermore, the asymmetry analysis in specific brain regions was observed to be sensitive for predicting cognitive decline (Long et al., 2018). It is possible to speculate that the left hippocampus might be more sensitive than the right one in the disease progression of svMCI.

We identified alterations in CA1, ML, CA4, and CA3 in the individuals with svMCI. These results were similar to previous studies on amnestic mild cognitive impairment (Kang et al., 2018), another sub-type of mild cognitive impairment, which shared similar symptoms especially memory impairments. Hippocampal CA1 pyramidal neurons play a crucial role in the memory circuit of the medial temporal lobe and are particularly susceptible to damage during the course of Alzheimer’s disease. Meanwhile, research has depicted that the CA1 subfield of the hippocampus in gerbils appears to be vulnerable to anoxic-ischemic insults (Huang et al., 2022; Kirino & Sano, 1984). Further, autopsy studies of patients with microvascular pathology have demonstrated notable atrophy and neuronal loss in the CA1 region (Kril et al., 2002). Consequently, alterations identified in neurodegenerative and vascular cognitive impairments may imply a common pathological mechanism for cognitive decline. Meanwhile, the current investigation revealed that the hippocampal tail, hippocampal fissure, and ML-HP head displayed discernible features that could differentiate svMCI from control subjects. This suggests that the subcortical vascular origin of this condition may result in a distinctive pattern of hippocampal damage. However, different whole hippocampal volumes derived from automatically or manually segmented MRI scans have been published (Pluta et al., 2012; Yushkevich et al., 2010). It is important to note that it is unsuitable to directly compare our results to previous ones if different hippocampal subfields were used. In this regard, we used the latest one, which is built with ultra-high resolution ex vivo MRI data (∼ 0.1 mm isotropic) and ex-vivo autopsied brain MRI from multiple subjects.

Unexpected, we identified a negative correlation between decreased GMV in the left CA3 head and MoCA scores in individuals with svMCI. The hippocampus CA3 is primarily responsible for memory and cognitive control and is composed of pyramidal neurons (Kesner, 2013; Nakashiba et al., 2009; Suthana et al., 2015). The arterial vascularization of the hippocampus depends on the collateral branches of the posterior cerebral artery and the anterior choroidal artery, whereas the CA3 region may only be supplied by the large dorsal hippocampal artery (Tatu & Vuillier, 2014). As a consequence, brain ischemia has a particularly detrimental impact on CA3, one of the most vulnerable subregions of the hippocampus (Vatrinet et al., 2017). Hypoxia in the CA3 area of the hippocampus may impede energy production, lead to hippocampal degeneration, and ultimately result in neural malfunction by triggering inflammation and disrupting the integrity of the blood-brain barrier (Liu et al., 2022). Since all subcortical infarcts were very small (< 20 mm) and white matter lesions (WMLs) were around the ventricle or randomly distributed in the whole brain, it is extremely difficult to obtain all lacunas and microvascular lesions and calculate WMH volume for each subject. Thus, we did not explore the relationship between WMH volume and adjusted CA3 volume of the subjects in this study to exclude the possibility that CA3 might be susceptible to ischemic lesions. Additional attention should be paid to explain the negative relationship between CA3 volume and MoCA in the svMCI.

Limitations

This research has some drawbacks. First, the relatively small sample size with heterogeneous vascular factors by nature could reduce the external validity of current findings. A larger sample size with multi-center is necessary to replicate our results. Second, the cross-sectional research was unable to establish a cause-and-effect connection between neuroimaging alterations and cognitive deficits. Third, this study focused on the structural modifications of hippocampus sub-regions. Further studies considering the relationship between the hippocampus and the rest of the whole brain are required. Fourth, in the analysis of the relationship between changes in imaging and clinical symptoms, specific focus on regions such as executive function and their correlation with imaging changes was unexplored. Additionally, future studies should consider including information on amyloid-beta deposition and APOE genotyping, as these were important biomarkers of Alzheimer’s disease. Finally, the significant p value was not adjusted for multiple comparisons. More caution is needed for the explanation of our results.

Conclusions

In our investigation, we observed several modifications in GMV that were primarily localized in the left hippocampal sub-regions of individuals with svMCI. Notably, a reduction in GMV within the left CA3 head sub-region was significantly associated with disease severity. Our findings reveal distinct patterns of atrophy within the hippocampal subfields that underlie the pathophysiological mechanisms of cognitive impairment in patients with svMCI. These results may serve as a valuable resource for future investigations aimed at elucidating svMCI pathogenesis.