Hippocampal subfield-specific imaging in depression: the translational power of ultra-high field MRI

Abstract

Major depressive disorder is a highly prevalent psychiatric condition characterized by diverse symptom profiles and variable treatment responses. The hippocampus has long been implicated in its pathophysiology, yet most human neuroimaging studies have treated it as a unitary structure, overlooking the distinct cytoarchitecture and connectivity of its subfields. Preclinical research demonstrates that subfields such as cornu ammonis 1 (CA1) and the dentate gyrus (DG) are differentially involved in stress susceptibility and antidepressant response. However, limited spatial resolution of conventional human neuroimaging has constrained in vivo characterization of these subfields and testing of translational hypotheses in humans. While advances in high resolution 3 tesla (T) imaging have improved subfield delineation, ultra-high field (UHF) MRI (7 T and above) extends these capabilities by combining whole-brain coverage with greater sensitivity to structural, functional and microstructural contrasts. This review synthesizes emerging applications of UHF MRI in depression, highlighting its advantages over conventional imaging approaches and early evidence for subfield-specific alterations. While most studies to date have focused on volumetric analyses, preliminary literature suggests structural and microstructural abnormalities in CA1 and DG, paralleling preclinical findings and implicating these subfields in network-level dysfunction and plasticity-related mechanisms. Future research should move beyond volumetric analyses to adopt more diverse MRI protocols capable of probing how subfield-specific architecture contributes to whole-brain networks and relates to clinical heterogeneity. UHF MRI offers a powerful platform to test mechanistic hypotheses derived from animal models, identify subfield-specific biomarkers, and ultimately guide personalized interventions targeting hippocampal circuits most relevant to individual symptom profiles.

Introduction

Major depressive disorder (MDD) is a leading cause of global disability [1], marked by heterogeneous symptom profiles and variable treatment responses [2]. A major challenge in the field is identifying biomarkers that map onto mechanistically distinct dimensions of MDD to support more targeted and effective treatment strategies [3]. The hippocampus–a medial temporal lobe structure studied extensively in memory, emotion regulation and stress sensitivity [4]–has consistently been implicated in MDD through structural and functional abnormalities [5–7]. However, most human magnetic resonance imaging (MRI) studies treat the hippocampus as a unitary structure, overlooking its subfields, each characterized by distinct cytoarchitecture, connectivity, and functional roles [8].

In neuroimaging contexts, the term hippocampus typically refers to the hippocampal formation, consisting of the cornu ammonis (CA) fields, the dentate gyrus (DG), and the subicular complex (Fig. 1). Traditionally, CA1-4 are considered subfields of the hippocampus proper. However, considerable variation exists in how CA4 is defined. Some anatomical and imaging studies merge CA4 with either CA3 or the DG, reflecting its position between the two subregions [9–12]. Notably, a distinct CA4 subfield is not typically recognized in rodent neuroanatomy, where this region is recognized as a layer of the DG [13].

Fig. 1. Schematic illustration of human hippocampal subfields.

Coronal view of the hippocampal formation showing the dentate gyrus (DG), cornu ammonis fields (CA1-4), and the subicular complex, consisting of the presubiculum, parasubiculum, and the subiculum proper. Created with BioRender.com [160].

Preclinical research has demonstrated that subfields such as CA1, CA3 and the DG are differentially involved in stress reactivity and mechanisms of antidepressant medications [14–16]. These findings suggest that subfield resolution is critical to probe hippocampal involvement in depression. However, testing subfield-specific hypotheses in humans remains challenging, as conventional MRI typically lacks the spatial resolution and tissue contrasts needed to reliably delineate internal hippocampal boundaries in vivo [17–19]. Recent advances in ultra-high resolution (UHR) 3 tesla (T) imaging enable subfield-level structural detail, but remain limited to slab acquisitions that preclude whole-brain coverage [20].

Ultra-high field (UHF) MRI, defined as 7 T and above, advances subfield-specific imaging of the hippocampus by providing improved tissue contrast and sensitivity to diverse physiological and microstructural properties, while enabling whole-brain coverage across sequences [17, 21, 22]. These advances open new avenues for testing mechanistic hypotheses derived from preclinical models and for probing the role of hippocampal subfields within large-scale functional and structural brain networks rather than in isolation.

This narrative review examines how UHF MRI has been applied to study hippocampal subfields in depression (supplementary material 1.1). We focus on CA1, CA2, CA3, CA4, the DG and the subicular complex of the hippocampal formation. We begin by summarizing preclinical evidence for subfield-specific roles in stress response and antidepressant mechanisms. We then summarize findings from conventional MRI studies and highlight the limitations that motivate a shift toward UHF imaging. Next, we synthesize current UHF studies of hippocampal subfields in MDD and discuss avenues for future work. We conclude with a discussion on how subfield-specific insights may inform future directions in biomarker discoveries, personalized interventions, and cross-species translational insights.

Preclinical evidence for hippocampal subfield-specific roles in stress and antidepressant treatment response

A substantial body of preclinical research has highlighted the distinct roles of hippocampal subfields in stress-related behavior and antidepressant-like behavioral responses [14]. Although animal models cannot fully capture the complexity of human depression, chronic stress paradigms remain a widely used and well-validated approach for studying neural mechanisms relevant to affective behaviors [23].

Rodent studies demonstrate that stress exposure induces subfield-specific alterations in hippocampal structure and function. Chronic stress leads to atrophy and reduced synaptic activity in CA1 and CA3 pyramidal cells [14, 15, 24], and decreased granule cell activity and impaired plasticity in the DG [16]. DG plasticity has also been shown to mediate behavioral response to antidepressant treatment [25–27], including fluoxetine [28], although this effect appears strain-dependent, as some high-anxiety mouse strains show behavioral recovery in the absence of DG plasticity [29]. Robust increases in DG plasticity are also consistently reported in rodent models of electroconvulsive therapy (ECT), after both single and repeated sessions [30–34]. While CA1, CA3 and the DG have been extensively studied in preclinical stress models, the subiculum has been comparatively overlooked. Emerging evidence suggests that chronic stress may bias pyramidal neurons in the subiculum towards firing patterns associated with enhanced output [35], though this effect requires further validation.

Inter-subfield dynamics may also have functional relevance. For example, the ratio of activity between the DG and CA1 has been correlated with depressive-like behaviors and antidepressant-like behavioral responses [36], suggesting that measures of relative activity between subfields of the hippocampus may provide an additional level of sensitivity to affective states.

Beyond local microcircuitry, distinct output pathways from hippocampal subfields play a critical role in regulating affective behavior. CA1 and subiculum project to regions implicated in emotional regulation, including the medial prefrontal cortex (mPFC) [37], amygdala [38], and nucleus accumbens (NAc) [39]. A growing body of circuit-level research shows that modulation of these projections influences affective behavior. For example, increased glutamatergic activity in subiculum to NAc projections heightens chronic stress susceptibility [40], while synaptic plasticity of CA1 to NAc projections is necessary for ketamine’s antidepressant-like response [41].

Despite species differences in scale and cortical organization, hippocampal subfield architecture and circuit connectivity are highly conserved across mammalian species [42], reinforcing the translational relevance of rodent models for understanding human hippocampal function. However, whether such subfield-specific mechanisms are present in humans remains unclear, in part due to the resolution and contrast limitations of most conventional clinical neuroimaging.

Human neuroimaging evidence of hippocampal involvement in depression using conventional MRI

Human neuroimaging studies have consistently implicated the hippocampus in the pathophysiology of MDD, with alterations observed across multiple imaging approaches – including volume, connectivity, task-based activation and microstructure – often linked to symptom severity and treatment response [5, 43–45]. Most of this work has treated the hippocampus as a unitary structure, limiting insight into the distinct contributions of its subfields. The following sections synthesize findings from studies conducted at conventional field strengths (i.e., 1.5 and 3 T MRI), highlighting how the hippocampus has been studied in MDD and identifying gaps in subfield-level characterization that motivate the use of UHF MRI.

Volumetric abnormalities

Smaller hippocampal volumes in MDD have been consistently reported across meta-analyses [5, 7], with large-scale efforts like the ENIGMA Group attributing these findings mainly to recurrent MDD [46]. Antidepressant treatments, including ECT and pharmacological interventions, may increase total hippocampal volume; however, the relationship between volumetric change and symptom improvement remains inconsistent [47–49], likely reflecting differences in treatment timepoints, illness durations or methodological variability.

Hippocampal volume is estimated to be up to 80% heritable [50–52]. However, preliminary investigations into shared genetic architecture between hippocampal volume and MDD have not yielded significant associations [53]. It therefore remains unclear whether volume differences in MDD reflect existing genetic factors, disease-related changes, or both.

Conventional MRI studies have increasingly examined subfield-specific volume differences and treatment effects in MDD. These studies vary markedly in spatial resolution, with recent advances in 3 T imaging now enabling ultra-high resolution (UHR) acquisitions that improve in vivo segmentation strategies and volume precision (see Overcoming limitations of conventional imaging section). A meta-analysis of 28 studies reported smaller CA3 and CA4 volumes in MDD, although approximately half of the included studies reported no significant differences [54]. Among the subset of UHR studies, two studies converged on smaller CA1, CA2/3 (combined due to segmentation strategy) and CA4 volumes [55, 56], while one observed no group differences [57].

Subfield-specific treatment effects have also been explored. Volume increases in CA1, CA2/3, the DG and subiculum have been reported within 6 days of ECT treatment in both UHR 3T [57] and conventional imaging studies [58, 59]. Notably, in one longitudinal study, subfield volumes returned to baseline after 6 months and were unrelated to sustained antidepressant response [59], suggesting that ECT-related volume increases may reflect acute, transient treatment effects [60]. In parallel, six serial ketamine infusions have been associated with increases in CA4 and DG volume [61]; however, other studies, including single and four serial infusion protocols, have reported no subfield volume changes at any timepoint, including 5 weeks after the final ketamine infusion [62, 63]. To date, no longer-term longitudinal studies have examined ketamine-related effects on hippocampal subfield volumes.

Taken together, these findings suggest that while hippocampal volume differences in MDD are robust at the whole-structure level, subfield-specific effects are less consistent. This variability likely reflects differences in imaging resolution and, critically, the timing of scans relative to illness stage or treatment exposure. Further investigation is needed to assess whether UHF imaging can clarify these inconsistencies, or if volume reductions are inherently limited in their clinical relevance.

Functional connectivity alterations

Resting-state functional connectivity involving the hippocampus is consistently altered in MDD, with disruptions observed across multiple functional networks. Increased connectivity within the default mode network (DMN), specifically involving the hippocampus, has been reported, potentially contributing to rumination and self-referential thinking [43, 64]. Conversely, reduced DMN connectivity has also been reported in recurrent MDD, particularly in association with antidepressant usage, suggesting that connectivity alterations may differ by illness duration and with treatment [65]. Other studies have found increased functional connectivity with the subgenual anterior cingulate cortex, alongside reduced functional connectivity with regions involved in salience processing, such as the insular cortex [6]. Dynamic functional connectivity analyses further support aberrant hippocampal function in MDD, with multiple studies reporting decreased temporal variability in the hippocampus [66]. Preliminary evidence suggests that ketamine decreases hippocampal hyperconnectivity with limbic structures such as the thalamus, supporting a potential role for circuit-level plasticity in its antidepressant effects [67].

A small number of subfield-specific functional connectivity studies at conventional field strengths have begun to reveal spatial and temporal heterogeneity in hippocampal network organization. Across investigations, both decreased and increased functional connectivity have been reported between hippocampal subfields, particularly CA1–3, DG and subiculum, and cortical areas involved in sensory, associative and salience-related processing [68, 69]. Reduced functional connectivity between CA1 and fusiform gyrus has been associated with increased depression severity [68], while decreased connectivity between CA1 and prefrontal regions has been linked to a history of childhood trauma in MDD patients [69]. Preliminary evidence from dynamic connectivity analyses suggests that temporal variability in CA1-prefrontal cortex connectivity is altered in MDD and may relate to symptom expression or clinical state [69]. Both studies acquired fMRI data with voxel sizes of approximately 3.5 mm isotropic, substantially limiting spatial precision for resolving hippocampal subfield-level functional dynamics.

Together, these findings highlight the importance of the hippocampus in large-scale functional network alterations in MDD. Preliminary subfield-level analyses suggest that CA1 may play an especially prominent role in hippocampal-cortical connectivity related to affective states. Replication using individualized subfield delineation approaches and higher resolution fMRI sequences, enabled by UHF MRI, will be critical for evaluating the specificity and clinical relevance of hippocampal subfield alterations.

Abnormal task-based activity

Task-evoked changes in hippocampal activity have been examined across emotional, cognitive, and reward-based paradigms in MDD, but meta-analytic findings are mixed. Some studies report hippocampal hyperactivation during emotional processing tasks, such as passive viewing or face-matching paradigms [70, 71]. However, valence-specific effects (i.e. responses to sad versus happy faces) and associations with symptom severity are inconsistent [45]. Other meta-analyses have failed to identify reliable hippocampal engagement during emotional tasks, instead reporting convergent effects in adjacent regions such as the parahippocampal gyrus [72–75]. Analyses pooling across studies using cognitive, emotional and reward paradigms have similarly produced conflicting results. Some studies report hippocampal hyperactivation [76], while others report no significant hippocampal effects [77].

Treatment-related changes in hippocampal activation have also been observed. Cognitive behavioral therapy has been associated with reduced hippocampal activity during emotional and reward tasks, although some studies have reported increased activation and such changes have not been clearly linked to symptom improvement [78]. Preliminary evidence suggests that pharmacological treatments may differentially modulate hippocampal activity; venlafaxine has been shown to increase hippocampal activity during explicit emotion tasks, while mirtazapine has not [79].

Mixed task-based findings may result from averaging signals across functionally diverse hippocampal subfields. To date, no task-based fMRI studies in MDD have characterized hippocampal activation patterns at the subfield-level. Such analyses could identify whether distinct hippocampal subfields differentially contribute to affective, cognitive, and reward-related dysfunction–key symptom domains that may benefit from targeted therapeutic approaches in MDD.

Structural connectivity and neurite microstructure

Diffusion MRI studies have consistently identified structural abnormalities in hippocampal circuitry in MDD, with converging evidence for decreased fractional anisotropy (FA) in white matter tracts connecting the prefrontal cortex to limbic structures including the hippocampus [44]. A large multi-site diffusion MRI study reported significantly lower FA in 16 of 25 major white matter regions, with the most pronounced reductions observed in tracts implicated in hippocampal-prefrontal connectivity [80].

To date, few studies have investigated subfield-specific alterations using diffusion techniques. Preliminary work has reported no significant FA differences in hippocampal subfields [68]. However, recent findings using neurite orientation dispersion and density imaging (NODDI)–a diffusion technique sensitive to microstructural complexity–identified ECT-associated increases in neurite density and dispersion and decreases in isotropic fraction within the DG [81]. These changes likely reflect alterations in dendritic architecture; however, they were not associated with clinical response and the limited spatial resolution (2 mm isotropic) prevented the analysis of other hippocampal subfields. While limited to a single study, these findings highlight the potential of diffusion imaging to probe DG plasticity mechanisms hypothesized by preclinical models. Additional studies, particularly at UHF, are needed to clarify clinical relationships and investigate subfield-specific microstructure beyond the DG.

Despite robust evidence for white matter tract disruptions involving the hippocampus, no conventional MRI studies have investigated subfield-specific structural connectivity. Regional FA measures offer insight into local microstructural integrity, but do not probe whether structural abnormalities exist in pathways between specific hippocampal subfields and key cortical and subcortical targets. UHF may help bridge this gap by enabling subfield-specific tractography to test circuit-level hypotheses of hippocampal dysconnectivity in MDD–an approach now beginning to be explored in recent UHF studies.

Conventional MRI summary

While hippocampal involvement in MDD is well established, the specific roles of individual subfields–and whether they are selectively modulated by antidepressant treatments–remain largely unresolved. Preliminary human studies suggest spatially and temporally distinct patterns across hippocampal subregions–most notably in CA1 and the DG, echoing preclinical evidence for their roles in affective behavior and treatment response. However, most subfield-level findings from conventional MRI remain sparse, methodologically constrained, and unreplicated. This likely contributes to the heterogeneity observed across MRI contrasts, especially within task paradigms, where whole-structure analyses have yielded inconsistent results and limited mechanistic insight. Addressing these gaps will require tools capable of linking subfield-level organization to broader network dysfunction and clinical heterogeneity. Clarifying whether specific subfields, such as CA1 or subiculum, exhibit altered connectivity with prefrontal regions may help bridge observed circuit abnormalities with cognitive and affective symptoms in MDD. UHF MRI presents a path forward, offering the anatomical precision needed to test subfield-specific hypotheses and the flexibility to incorporate more MRI contrasts to ask more mechanistically grounded questions about hippocampal function in depression.

Ultra-high field MRI as a tool for advancing hippocampal subfield imaging in depression

Overcoming limitations of conventional MRI for subfield imaging

Advancing segmentation strategies

Manual delineation of hippocampal subfields in vivo requires clear visualization of key internal structures, such as the boundary between the DG and CA fields, which is typically less than 1 mm thick [9, 82, 83]. On T2-weighted (T2-w) images acquired perpendicular to the hippocampal axis, manual segmentation typically relies on in-plane resolutions of 0.2-0.6mm [18, 84], a range that is routinely available at 7 T. The increased field strength provides greater intrinsic signal-to-noise (SNR) and contrast-to-noise (CNR) ratios which can be leveraged to enhance spatial resolution, sharpen tissue boundaries and improve sensitivity to subtle anatomical variation [21, 22, 85, 86]. Moreover, 7 T imaging reduces partial volume effects which further improves visualization of key anatomical features necessary for delineation of subfields [17, 86].

At conventional field strengths, restricted spatial resolution and tissue contrast limit visibility of internal boundaries, restricting reliable manual segmentation at 3 T and 1.5T [17–19]. Recent advances in UHR 3 T imaging have significantly improved the ability to delineate hippocampal subfields in vivo, with T2-w images achieving sub-millimeter in-plane resolution comparable to that of 7 T protocols [20, 87–89]. While tissue contrast afforded from higher CNR at 7 T remains superior, the spatial resolution gains at 3 T provide substantially improved anatomical precision [90].

Manual segmentation was previously considered the gold standard; however, its susceptibility to human error and concerns about reliability have led many studies to adopt automated segmentation pipelines. Early automated tools depended heavily on probabilistic atlases, as the image quality on 1.5 T and 3 T scanners is generally insufficient to reliably delineate subfield boundaries directly [19]. This reliance on prior assumptions of average anatomy could be masking individual variability and reducing sensitivity to disease-related changes [19]. Due to this limited spatial resolution, segmentation strategies often combine adjacent subfields to improve reliability, reducing anatomical specificity and complicating preclinical translation.

UHR T2-w imaging at 7 T and 3 T supports automated tools that incorporate subject-specific tissue contrasts, rather than relying exclusively on probabilistic atlases [91–95] (Fig. 2) (UHF-specific segmentation reviewed here [18]; 3 T vs 7 T tools compared here [10]). These approaches mark a significant improvement in the anatomical fidelity of hippocampal subfield segmentation by capturing variability in individual anatomy, with several tools demonstrating reliability comparable to manual segmentations [84, 92–94].

Fig. 2. Hippocampal subfield segmentations on 7 T MRI images from one healthy volunteer.

Coronal views of 7 T T2-w TSE (0.43 × 0.43 × 2.0 mm), T1-w MP2RAGE (0.7 mm isotropic), and fMRI images (1.1 mm isotropic) were acquired on a Siemens Magnetom 7 T scanner at the Icahn School of Medicine at Mount Sinai. Hippocampal subfield segmentations were generated using ASHS using both T2w and T1w inputs [95] then co-registered into fMRI space. 3 T 3 tesla, 7 T 7 tesla, fMRI functional magnetic resonance imaging, CA cornu ammonis, DG dentate gyrus.

Enabling whole-brain coverage for network-level analysis

Despite major advances in UHR 3 T imaging, whole-brain diffusion and functional imaging at 3 T remain constrained by spatial resolution necessary for subfield-resolved measurements. Postmortem data suggests that voxel sizes of at least 1-1.5 mm should be used to reliably capture subfield-specific signal to mitigate partial volume effects [82]. UHR diffusion sequences at 3 T have achieved 1 mm isotropic resolution by restricting brain coverage (i.e. slab acquisitions) [96–98], but this precludes whole-brain tractography or network-level analyses. Similarly, UHR 3 T fMRI has improved characterization of hippocampal functional dynamics [87], yet to our knowledge sub-1.5 mm resolution has not been achieved. Ultimately, UHR 3 T can support localized subfield diffusion and volumetric analyses, but it remains insufficient in capturing subfield-specific functional signal or large-scale network interactions.

By contrast, whole-brain fMRI and diffusion sequences at 7 T have achieved sub-1.5 mm isotropic resolution [86, 99], allowing hippocampal subfields to be studied within large-scale brain networks rather than in isolation (Fig. 3). Even in studies without UHR structural sequences for subfield delineation (e.g. 0.4 × 0.4 mm in-plane T2-w scans), acquisitions at UHF support voxel-wise, whole-brain analyses in standard Cartesian space that can resolve effects at the level of individual subfields. This enables detection of subfield-specific functional or structural alterations even in protocols that do not include individualized segmentations.

Fig. 3. Comparison of whole-brain fMRI images at 7 T and 3 T with overlaid hippocampal subfield segmentation.

Coronal views of resting-state fMRI data from healthy volunteers obtained at ultra-high field (Magnetom Siemens 7 T, data collected at Icahn School of Medicine at Mount Sinai; voxel size: 1.1 mm isotropic) and conventional field (Magnetom Siemens Skyra 3 T, data obtained through Human Connectome Project [161], voxel size: 2 mm isotropic). Hippocampal subfield segmentations were generated using FreeSurfer v7.4.1 [156] and co-registered to fMRI images. At 3 T, lower spatial resolution, reduced functional contrast, and pronounced partial volume effects obscure internal hippocampal boundaries, limiting accurate functional localization to subfields. In contrast, 7 T fMRI provides improved spatial resolution, signal-to-noise ratio, and functional contrast, supporting more reliable alignment of subfield segmentations and enabling subfield-resolved analyses of hippocampal function. fMRI functional magnetic resonance imaging, 3 T 3 tesla, 7 T, 7 tesla, CA cornu ammonis, DG dentate gyrus.

Improving sensitivity to broad tissue properties

Building on advances in spatial resolution and coverage, 7 T MRI further improves sensitivity to the microstructural, vascular, and metabolic properties of hippocampal tissue–enabling more biologically specific markers of subfield integrity. At 7 T, functional MRI gains not only spatial precision but also greater sensitivity to localized neural activity and reduced contaminations from large blood vessels [100–103], both critical for resolving blood-oxygen-level-dependent (BOLD) signals from small subfields. Similarly, diffusion imaging at 7 T provides more reliable estimates of microstructure–even when voxel sizes match UHR 3T– due to advances in acquisition and reconstruction methods at 7 T supporting improved signal fidelity in historically difficult temporal lobe regions [104, 105].

Among diffusion techniques, NODDI holds particular promise in reconciling subfield-specific microstructural complexity changes in depression [81, 106, 107]. Although not yet used to investigate hippocampal subfields at 7 T, NODDI has demonstrated preliminary feasibility for subfield-specific resolution, with a previous 7 T study using 1.25 mm isotropic resolution to explore locus coeruleus microstructure [108]. NODDI offers neurite density and orientation dispersion indexes that provide insights into both gray and white matter microstructural complexities. Additional emerging gray-matter sensitive diffusion models, such as diffusion kurtosis imaging [109], further expand the potential of 7 T MRI to characterize hippocampal microstructure in vivo. Diffusion kurtosis quantifies non-Gaussian diffusion, providing sensitivity to microstructural heterogeneity and complexity particularly relevant to gray matter structures. Subfield-specific kurtosis quantification is now feasible in whole-brain acquisitions with 1.5 mm isotropic resolution [110].

Susceptibility-based techniques such as quantitative T2* mapping, quantitative susceptibility mapping (QSM), and susceptibility weighted imaging (SWI) also benefit from increased field strength at 7 T, offering complementary sensitivity to tissue properties like microvascular integrity and microstructural differences across hippocampal subfields, while achieving submillimeter resolutions [111, 112]. In addition, UHF expands the scope of perfusion imaging such as arterial spin labeling (ASL). Beyond enabling adequate spatial resolution for subfield-resolved ASL acquisitions (e.g. 1 mm isotropic), UHF increases labeling efficiency and SNR, allowing subfield-level quantification of cerebral blood flow and coupling between perfusion and neural activity [113, 114]. Such measures could complement BOLD and diffusion-based measures in assessing vascular and metabolic components of hippocampal dysfunction in depression.

Although the sequences discussed are not exclusive to 7 T, UHF MRI uniquely enables whole-brain acquisitions at 1.5 mm (and below) isotropic resolution with superior signal fidelity, contrast, and physiological sensitivity–allowing network-level rather than purely regional analyses and subfield-resolved investigations of a diversity of tissue properties.

Technical considerations and trade-offs at UHF

At all field strengths, MRI acquisition involves inherent trade-offs between spatial and temporal resolution. As the main field strength (B₀) increases, intrinsic SNR and CNR scale approximately linearly [90]. At 7 T, this gain can be leveraged to achieve higher spatial resolution; however, this typically requires longer acquisitions or comes at the expense of temporal resolution. For fMRI, to achieve sub-1.5 mm isotropic resolution at 7 T requires protocols to be optimized to maintain usable temporal resolution (typically under 2 seconds [115]), while mitigating increased geometric distortion and susceptibility artifacts inherent to UHF EPI sequences [86]. Similarly, for high-resolution whole-brain diffusion and structural sequences, the decreased voxel size increases overall acquisition times. This challenge is primarily mitigated by the use of multiband acceleration techniques, making the scan duration more feasible for whole-brain sub-1.5 mm isotropic sequences [99].

In addition, the heightened specific absorption rate (SAR) at 7 T can constrain sequence design and increase power deposition, necessitating careful protocol optimization to ensure patient safety [116]. Early applications of 7 T MRI also raised concerns about B₀ and B₁ inhomogeneities, artifact susceptibility, and SAR constraints; however, the past decade has seen rapid progress in addressing these technical limitations [90, 117]. Advances in coil design, pulse sequences and image correction strategies have significantly improved signal fidelity, particularly in historically challenging regions like the temporal lobe [90, 117, 118]. As a result, many of the initial barriers to implementation of UHF MRI are being actively mitigated, enabling more reliable subfield-specific analyses across contrasts.

Current applications of UHF MRI of hippocampal subfields in depression

Volumetric analyses

Early UHF MRI studies in MDD have primarily focused on hippocampal subfield volumetrics (Table 1). Across eight studies comparing individuals with MDD and healthy controls, no consistent volumetrics differences across hippocampal subfields have been reported [62, 119–125]. These studies varied in sample characteristics, analytic approaches, and segmentation strategies, with only four incorporating UHR T2-w scans (in-plane resolution ~0.4 × 0.4 mm) for subfield delineation [120, 122, 124, 125]. None of the UHR studies identified significant group differences [120, 122, 124, 125], although one study reported uncorrected findings of smaller volume in CA1 and CA3/4 [124].

Table 1.

Summary of published 7 T MRI studies investigating hippocampal subfields in major depressive disorder or individuals with depressive symptoms.

Authors	Imaging analysis type	Subfield delineation method	Subfields included	Scanner and sequence acquisition	Study design	Sample details	Main findings
Liu et al. [119]	Volumetric	FreeSurfer 7.2.0 [156]	CA1*,CA3*,CA4*, GC-ML-DG*, subiculum*, presubiculum*, ML*, HATA, fimbria, tail, fissure *head and body separated	Philips Achieva 7 T T1-w MP2RAGME (0.7 mm iso)	Case-control	56 MDD† (14 atypical MDD, 42 typical MDD) 14 HC	No significant differences in subfield volumes in MDD vs HC. Typical MDD had lower volume in HATA, CA4 body, CA1 head, GC-ML-DG body, CA4 head, GC-ML-DG head vs. atypical MDD.
Alper et al. [122]	Volumetric	ASHS [95] with manual atlas using 15 MDD subjects	CA1, CA2/3, CA4/DG, subiculum, presubiculum	Siemens Magnetom 7 T T1-w MP2RAGE (0.7 mm iso) T2-TSE (0.45 × 0.45 × 2mm, coronal oblique slab)	Case-control	35 MDD (13 TRD) 28 HC	Smaller CA2/3 volume in MDD vs. HC.# Smaller CA2/3 in TRD vs. HC. CA1 and CA4/DG volume negatively associated with life stressors.
Tannous et al. [121]	Volumetric	FreeSurfer 6.0 [156]	CA1, CA2/3, CA4, GC-ML-DG, ML, Presubiculum, subiculum, tail	Siemens Magnetom 7 T T1-w MPRAGE (1 mm iso)	Case-control	71 MDD† 46 HC	No significant differences in subfield volumes in MDD vs. HC. No significant relationships between subfield volumes and illness duration, depression clinical scales, biological stress assays, or childhood trauma in MDD.
Kraus et al. [123]	Volumetric	FreeSurfer 6.0 [156]	CA1, CA3, CA4, fimbria, fissure, GC-ML-DG, HATA, ML, parasubiculum, presubiculum, subiculum, and tail	Siemens Magnetom 7 T T1-w MP2RAGE, (0.74 × 0.68 × 0.68 mm)	12-week longitudinal; MRI at baseline & week 12; aMDD treated with SSRI or transitioned to SNRI if non-response	28 rMDD 20 aMDD 22 HC	No change in hippocampal subfield volumes after antidepressant treatment in aMDD. Larger volume in fissure and HATA in rMDD vs. HC at both MRIs. Larger subiculum volume in aMDD and rMDD vs. HC at MRI-2 only.
Van der A et al. [129]	Volumetric	ASHS [95]	CA1, CA2, CA3, collateral sulcus, DG, subiculum	Philips 7 T T1-w TFE (1 mm iso) T2-TSE (0.286 × 0.286 × 2mm, coronal oblique slab)	5 week longitudinal; MRI pre and post 2x/week ECT treatments for 5 weeks	23 MDD†	Correlation between DG volume increase and lower immediate recall, delayed recall and recognition performance.# Correlation between CA1 volume increase and decreased immediate recall.#
Wisse et al. [127]	Volumetric	Manual delineation [157]	CA1, CA2, CA3, DG/CA4, subiculum, tail	Philips 7 T T2-TSE (0.7 mm iso)	General practice cohort	34 no MDE† 13 ever MDE†	No significant differences in subfield volumes in ever-MDE vs. no-MDE. Smaller subiculum volume associated with increased number of MDEs.#
Nuninga et al. [128]	Volumetric	ASHS [95]	CA1, CA2, CA3, collateral sulcus, DG, subiculum	Philips 7 T T1-w TFE(1 mm iso) T2-TSE (0.286 × 0.286 × 2mm, coronal oblique slab)	5 week longitudinal; MRI pre and post; 2x/week ECT in MDD	23 MDD† 8 HC	Significant increase in bilateral DG volume in MDD but not HC.# DG volume change associated with depression score reduction in MDD.#
Twait et al. [126]	Volumetric	ASHS [95]	CA1, CA2, CA3, CA4, DG, subiculum, tail	Philips 7 T T1-w 3D (1 mm iso) T2 3D (0.7 mm iso)	Mixed cohort study [158]	333 adults with no CI‡ (social support:177 high, 76 moderate, 80 low)	No significant association between subfields and depressive and anxiety symptoms. Smaller CA3 volume in low vs. high social support.
Evans et al. [62]	Volumetric	FreeSurfer 6.0 longitudinal pipeline [156, 159]	CA1, CA3, CA4, fimbria, fissure, GC-ML-DG, HATA, ML, parasubiculum, presubiculum, subiculum, and tail	Siemens Magnetom 7 T T1-w MPRAGE (0.7 mm iso)	Randomized, double-blind, placebo-controlled, crossover; MRI at baseline, 1-2 days (acute), 9-10 days(interim) post ketamine/saline infusion	30 TRD 17 HC	No baseline subfield differences in TRD vs. HC. No significant subfield changes pre/post ketamine or placebo at any timepoints.
Brown et al. [124]	Volumetric	FreeSurfer 6.0 [156]	CA1, CA3/4, DG, subicular complex	Siemens Magnetom 7 T T1-w MPRAGE (0.7 mm iso) T2-TSE (0.43 × 0.43 × 2.0 mm, coronal oblique slab)	Case-control	24 MDD 20 HC	No subfield volumetric differences in MDD vs HC. No association between subfield volumetrics and illness duration in MDD. Increased depression severity associated with reduced CA1 and CA3/4 volume.#
Jacob et al. [125]	Volumetric, diffusion tractography	FreeSurfer 6.0 [156]	CA1, CA3/4, DG, subicular complex	Siemens Magnetom 7 T T1-w MPRAGE (0.7 mm iso) T2-TSE (0.43 × 0.43 × 2.0 mm, coronal oblique slab) HARDI (1.05 mm iso, directions=64)	Case-control	38 MDD 40 HC	No subfield volumetric differences in MDD vs. HC. Decreased node strength of right CA3/4 in MDD related to greater depression severity.
Rutland et al. [120]	Volumetric, diffusion tractography	FreeSurfer 6.0 [156]	CA1, CA3/4, DG, subicular complex	Siemens Magnetom 7 T T1-w MPRAGE (0.7 mm iso) T2-TSE (0.4 × 0.4 × 2.0 mm, coronal oblique slab) HARDI (1.05 mm iso, directions=64)	Case-control	18 MDD 18 HC	No subfield volumetric differences in MDD vs. HC. Reduced SC in DG in MDD vs HC, positively correlated with age of onset. Increased MD and RD of fimbria in MDD vs. HC. No FA or AD subfield differences in MDD vs. HC.
Cho et al. 2011 [131]	Relaxometry	Manual delineation	CA1, CA2, CA3, CA4, subiculum	Siemens Magnetom 7 T T2* dual echo FLASH (0.347 × 0.347 × 1.5mm, coronal oblique slab, TE = 13.6/28.8ms)	Case-control	16 MDD† (9 rMDD) 16 HC	Longer T2*-RT in CA1 and subiculum in MDD vs. HC.# T2*-RT in CA3 correlated with illness duration in MDD.# Longer T2*-RT in subiculum in rMDD vs. non-rMDD.# Longer T2*-RT in CA1, CA3 and subiculum in rMDD vs. HC.#

GC-ML-DG granule cell-molecular layer-dentate gyrus, ML molecular layer, HATA hippocampus-amygdala transition area, 7T 7 tesla, T1-w T1-weighted, iso isotropic, MDD major depressive disorder, HC healthy control, ASHS Automatic Segmentation of Hippocampal Subfields, T2- T2-weighted, TRD treatment-resistant MDD, SSRI selective serotonin reuptake inhibitor, SNRI serotonin-norepinephrine reuptake inhibitor, rMDD recurrent MDD, aMDD acute MDD, DG dentate gyrus, ECT electroconvulsive therapy, MDE major depressive episode, CI cognitive impairment, HARDI high-angular-resolved diffusion-weighted imaging, SC streamline count, MD mean diffusivity, RD radial diffusivity, FA fractional anisotropy, AD axial diffusivity, T2*-RT T2* relaxation time.

Symbols: ^†Antidepressant medication allowed, ^‡Antidepressant medication status not reported, ^#Findings not corrected for multiple comparisons.

Additional work has explored the relationship between subfield volumes and other dimensions of depression, including subthreshold depression symptomatology [126], MDD subtypes [119], and history of major depressive episodes [127]. No volumetric differences were associated with individuals with subthreshold depression symptoms or a history of major depressive episodes in comparison to controls [126, 127]. One study reported that specific clinical features of MDD, such as emotional reactivity, were associated with volumetric differences across multiple subfields [119]; however others found no relationship between clinical features or illness severity and subfield volumes [121, 124]. Increased life stress has also been associated with reduced volume in CA1 and CA3/4 [122], but further follow-up studies are needed.

UHF MRI studies investigating subfield-specific volume changes with antidepressant treatment have yielded mixed results. No significant changes in total hippocampal or subfield volumes were detected 24 or 216 hours after a single ketamine infusion or after 12 weeks of selective serotonin reuptake inhibitor (SSRI) or serotonin-norepinephrine reuptake inhibitor (SNRI) treatment [62, 123]. In contrast, 10 serial ECT sessions were found to increase DG volume in patients with MDD, with baseline DG volume predicting clinical response [128]. Notably, total hippocampal volume did not predict treatment outcome, underscoring the importance of subfield-level resolution in capturing treatment effects. A later follow-up in the same participant cohort revealed that volume increases in DG and CA1 were selectively associated with distinct domains of cognitive performance decline after ECT: DG volume was associated with immediate recall, delayed recall and recognition memory declines, while CA1 volume was associated with immediate recall decline [129]. Critically, while the ketamine and SSRI/SNRI studies relied solely on T1-weighted (T1-w) imaging for subfield delineation, the ECT studies utilized UHR T2-w imaging, which likely enhanced sensitivity to subfield-specific effects.

Taken together, volumetric findings at UHF remain inconclusive, with few consistent group-level differences in subfield volumes. Although subfield-level volumetric analyses may be more sensitive to clinical heterogeneity–particularly CA1 and DG in relation to treatment response and symptom dimensions–findings to date remain variable and largely unreplicated. Methodological differences across studies, including variation in acquisition and segmentation protocols, did not appear to alter the overall pattern of group-level findings in MDD. While further validation is needed, especially in treatment studies and with UHR T2-w sequences, volume alone is unlikely to fully capture the neurobiological mechanisms underlying depression and antidepressant response. Postmortem studies suggest that hippocampal volume loss may reflect reductions in neuropil or dendritic complexity rather than neuronal death [130]. These microstructural changes are spatially heterogenous and are unlikely to be detectable reliably through gross volumetric measures alone. Moreover, uncertainty about whether hippocampal volume differences reflect heritable, developmental, or disease-related factors further underscores the interpretive limitations of volumetric approaches. Addressing these challenges will require genetically informed and longitudinal designs that prospectively track individuals from preadolescence into adulthood to distinguish inherited or early-life influences from disease-related trajectories. Moving forward, volumetric analyses may be most informative when embedded within MRI studies that assess microstructure, connectivity, and circuit-level dynamics–approaches that offer greater mechanistic interpretability and stronger correspondence with preclinical models of hippocampal circuitry.

Beyond volumetric analyses

Very few studies have utilized UHF MRI to investigate hippocampal subfields in MDD beyond volumetrics. To our knowledge, only two studies–both conducted at our center–have investigated structural connectivity of hippocampal subfields in MDD with 7 T diffusion MRI [120, 125]. Although these studies used different approaches–one applying graph theory to model whole-brain structural connectivity [125], and the other relying on standard diffusion metrics like streamline count [120]–they both provide evidence of reduced structural connectivity within DG. Additionally, reduced connectivity of CA3/4 was associated with greater depression severity [125]. Complementing diffusion-based studies, a quantitative T2* mapping study at UHF found prolonged relaxation time in CA1 and subiculum of individuals with MDD, suggesting selective microstructural tissue injury across hippocampal subfields [131]. However, replication is needed as these results were not corrected for multiple comparisons.

Although limited, existing UHF studies have begun to explore additional contrasts, revealing potential subfield-specific structural and microstructural alterations in MDD. Preliminary evidence points to localized abnormalities in CA1 and DG, but it remains unclear whether such differences are associated with functional network alterations or clinically meaningful outcomes. To date, no studies have investigated hippocampal subfield functional connectivity at 7 T in MDD–highlighting a critical, yet addressable, gap in the literature.

Towards a multimodal and mechanistically informed understanding of hippocampal subfields in depression

Expanding the scope of subfield-specific research

In depression research, subfield analyses remain largely focused on volumetrics, limiting insight into how structure, function, and tissue properties converge to shape clinical phenotypes and treatment response. To date, only three studies have explored non-volumetric analyses in MDD using UHF MRI [120, 125, 131], despite multiple demonstrations in healthy controls of the feasibility of measuring subfield-specific functional connectivity [132, 133], task-based activation [133, 134], neurite density [135], and blood perfusion [114]. While this underutilization may be due to underreporting null results, it stands in contrast to progress made in other clinical populations. For example, in epilepsy, subfield-specific functional connectivity analyses have been used to detect lateralized subfield-specific disruptions associated with clinical subtype and side of seizure focus [136]. Similarly, QSM has been used to quantify subfield-specific iron deposition and distribution related to memory performance in early Alzheimer’s disease [137]. These findings highlight the potential of subfield-specific analyses at UHF and underscore the need to expand beyond volumetric analyses in depression research. As detailed in Overcoming limitations of conventional MRI for subfield imaging, future work should leverage whole-brain resolution and increased sensitivity to microstructural, vascular, and functional contrasts to investigate subfield-specific alterations underlying depression phenotypes and treatment response.

Converging evidence across preclinical, conventional MRI, and UHF MRI studies implicates CA1 and DG as key subfields in the pathophysiology and treatment response in MDD. However, additional research is needed to disentangle their relationships with clinical heterogeneity and treatment outcomes. Preclinical studies indicate that CA1 is particularly susceptible to stress, and conventional imaging suggests this may have structural and functional consequences. Yet, these findings have not been clearly replicated at UHF, aside from preliminary evidence of altered CA1 relaxation times suggesting microstructural tissue changes [131]. In contrast, DG findings span multiple measures at UHF, including decreased structural connectivity [120, 125] and ECT-related volumetric increases [128, 129]. In conventional imaging studies, preliminary ECT-associated microstructural alterations in the DG [138] additionally converge with preclinical models linking DG plasticity to stress responses and antidepressant mechanisms. These patterns suggest that CA1 may serve as a hub of network-level dysfunction, while DG may represent a plasticity-sensitive subfield–each offering a distinct mechanistic and translational target in depression. However, the unique functional roles of these subfields and their involvement in hippocampal circuits, particularly in supporting cognitive, affective, and reward processing, have yet to be resolved in humans.

Functional specialization likely exists not only between subfields, but also along the hippocampal longitudinal axis, with converging evidence from preclinical, postmortem, and human imaging studies for anterior-posterior gradients in connectivity and function [8, 139]. Although this organization is well established across species, the exploration of gradients within subfields and their relevance to depression remains an emerging area of investigation. Early evidence suggests that anterior and posterior portions of hippocampal subfields may be differentially affected by antidepressant treatment and ECT [140, 141]. UHF MRI may increasingly allow investigations of subfield and axis specific effects, representing an important area for future exploration.

Translating subfield insights into targeted interventions

Isolating the contributions of key subfields may enable the development of subfield-specific therapeutic strategies. For subfields with strong afferent connectivity and potential influence on network dysfunction, such as CA1 and subiculum, one particularly promising, but unexplored, direction is real-time fMRI (rt-fMRI) neurofeedback [142, 143]. With this strategy, patients may learn to modulate activity in specific hippocampal subregions during task-based or naturalistic paradigms, such as movie viewing or autobiographical recall [144]. While prior studies in healthy individuals have demonstrated that rt-fMRI neurofeedback can modulate hippocampal activation [145], its application in MDD with subfield-specific targeting remains untested. Complementary strategies using noninvasive stimulation, such as transcranial magnetic stimulation (TMS), may further enable targeted network intervention. Recent work combining TMS with concurrent fMRI has shown that theta-burst stimulation targeting hippocampal networks can selectively enhance hippocampal activity during memory encoding and improve subsequent performance [146]. Extending these protocols to UHF MRI could allow researchers to evaluate whether modulation of CA1 or subiculum, and their associated prefrontal connectivity, alters task-evoked activation or connectivity patterns, and in turn, improve clinical symptoms.

Clarifying the clinical relevance of human hippocampal subfields would also provide a stronger foundation for back-translating findings to preclinical models. Hippocampal subfields differ markedly in cytoarchitecture, receptor density, and neurochemical sensitivity [12]. Defining the relevance of these subfields in humans could inform the development of pharmacological interventions that target subregion-specific mechanisms. Such biologically grounded strategies could help bridge molecular and circuit-level approaches to the treatment of MDD. A particularly compelling target for next-generation antidepressants is the DG, where interventions could be designed to selectively engage receptor systems–such as α₁-adrenergic or muscarinic receptors–that are differentially enriched in this substructure [12], paving the way for precision antidepressant therapies.

Considerations for implementation

Despite its promise, several practical and technical challenges continue to constrain the application of UHF MRI in depression research. Access to 7 T scanners is steadily improving, but remains far more limited than access to machines at conventional field strengths [117, 118, 147]. In addition, the technical trade-offs of UHF imaging–such as increased susceptibility to artifacts, participant ineligibility, and greater discomfort during the scan–can result in higher exclusion rates, particularly in clinical populations [90, 117, 148]. UHF image processing also demands specialized expertise and tailored preprocessing pipelines to mitigate motion-related degradation and correct for intensity inhomogeneities, spatial distortions, and sequence-specific artifacts. Building on decades of methodological refinement, ongoing initiatives continue to address these challenges by developing dedicated motion-correction tools and preprocessing frameworks for UHF imaging [149–152].

Current studies using UHF to investigate hippocampal subfields in individuals with depression have employed varied acquisition protocols, segmentation procedures, and analysis strategies. Rather than a limitation, this methodological heterogeneity likely reflects the early, more exploratory phase of applying 7 T MRI to increasingly small structures. Diverse approaches allow the field to probe different tissue contrasts and acquisition strategies, optimize delineation methods, and evaluate clinical effects from multiple angles. However, studies relying solely on standard whole-brain T1-w images at 0.7-1 mm resolution may not fully capitalize on the spatial precision afforded by 7 T. Incorporating dedicated UHR T2-w sequences ($\le$0.6 mm in-plane resolution) into imaging protocols will be critical for improving anatomical fidelity [19]. Moreover, without more systemic consideration of temporal factors, such as illness duration and treatment timepoints, the potential of UHF imaging to refine subfield-level volumetric findings will likely remain limited, regardless of its anatomical precision.

Given this methodological variability, transparent reporting–particularly of segmentation procedures and representative segmentation outputs–is essential to facilitate reproducibility and comparability across studies, as represented in recent studies [124, 125,127]. Emerging harmonization and open-source quality control initiatives are working to improve standardization and reporting transparency, supporting reproducible multi-site research and broad-scale clinical applicability [153, 154].

As access improves and adoption expands, UHF MRI offers a unique opportunity to pair anatomical precision with mechanistic hypotheses, enabling more targeted investigations of hippocampal subfields in relation to depression symptomatology and treatment response. Longitudinal approaches will be essential to characterize the temporal course of subfield alterations and to determine whether these changes precede, accompany, or follow clinical improvement. Ultimately, reproducible frameworks will be critical for ensuring that insights from UHF MRI translate into robust, mechanistically grounded models of depression [155].

Conclusion

Now that UHF MRI enables in vivo investigations of hippocampal subfields with greater specificity and across MRI contrasts, the next step is to harness this capability to broaden the scope of scientific questions we ask. Rather than focusing solely on hippocampal subfield volumetrics, future UHF neuroimaging research should integrate functional, structural, and microstructural assays to test subfield-specific hypotheses that are mechanistically grounded and clinically relevant. Preliminary findings from the small subset of UHF studies already highlight the relevance of examining hippocampal properties beyond volume. Importantly, 7 T findings should be viewed as complementary to the extensive body of 3 T research, offering greater anatomical and microstructural specificity that enables deeper mechanistic insight into how subfield-level alterations contribute to depression and treatment response. Future work must establish the robustness and reproducibility of preliminary findings and determine how subfield alterations converge across imaging contrasts and integrate within broader networks to shape symptom expression and treatment response. Ultimately, such a framework may pave the way toward biologically informed interventions that target specific hippocampal circuits based on an individual’s symptom profile.

Author contributions

JJ conducted the literature review, synthesized the findings, and drafted the manuscript. The overall direction of the review was developed through discussions between both authors. YJ provided conceptual guidance, critical revisions, and contributed funding and data resources. Both authors reviewed and approved the final manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-026-03870-5.