Ex vivo Magnetic Resonance Imaging of the Human Fetal Brain

The human cerebrum first appears as a neuronal tube during the embryonic period at the 4th to 5th post-fertilization weeks (Fig. 1, left) [1]. At around the 8th week, which is the beginning of the fetal period, the brain is more morphologically developed and crucial structures such as the cortical plate (CP), subcortical nuclei, and several commissure tracts come into being. The fetal brain experiences several developmental milestones before it turns into a structurally and functionally impact system, including the evolution of transient laminar zones in the telencephalic wall, the appearance of cortical gyri and sulci, and the emergence and growth of neuronal pathways. The underlying mechanisms of these profound developmental transitions have long been the interest of neurologists. Histology has been the most important method for investigating the intricate architecture of fetal brains [2‒4]. However, histological analyses require meticulous tissue dissection, visualizing only selected brain regions in 2-dimensional, and demand a high degree of professional expertise. Histological processing often leads to tissue deformation, thereby impeding the investigation of brain morphology.

On the contrary, magnetic resonance imaging (MRI) enables the 3D visualization of the entire brain with minimal invasion. MRI has been widely used to study the human fetal brain. Visual examples of fetal brain MRI scans are displayed in Figure 1. Despite the advances of in utero imaging techniques, ex vivo MRI still plays an irreplaceable role in visualizing fetal brain anatomy and microstructure for its superior image quality [6‒8]. Ex vivo MRI, benefiting from high-filed strength MRI scanners, longer scan durations, and custom-made RF coils that fit the specimen, can reach higher spatial resolution and signal-to-noise ratio. This enables the visualization of fine structures in the tiny fetal brain with only a few centimeters in length. Moreover, ex vivo imaging is free of the random and unpredictable fetal motion suffered by in utero MRI, reducing imaging artifacts. In addition, ex vivo brain samples can be used for genetic and histological analyses to investigate the underlying mechanisms correlating the MRI findings.

In this review, we first illustrated the commonly adopted ex vivo MRI techniques and their advances in fetal brain imaging, and proposed standards for postmortem fetal brain specimen collection and fixation. We then reviewed the ex vivo MRI-based studies on fetal brain development regarding the neuronal proliferative zones, cortex, and white matter (WM) fiber pathways, and summarized the utility and reliability of ex vivo MRI in these aspects. Finally, we listed limitations in using ex vivo MRI for fetal brain research and proposed several future directions in this field.

We searched for eligible research or review articles for this review on Web of Science with the following keywords: (“ex vivo” OR “postmortem” OR “in vitro”) AND (“MRI” OR “magnetic resonance”) AND (“fetus” OR “fetal”) AND (“brain” OR “cerebrum” OR “cerebral”) AND “development”. The search was limited to studies involving humans, yielding 98 articles. These articles were then manually reviewed to exclude those that were mistakenly selected or focused on unrelated methodologies, where MRI was only used as a supporting method without significant relevance to our review. Ultimately, we selected 58 articles. Additionally, we selectively discussed other studies that applied postmortem MRI to fetal brain malformations in section 5.

A crucial prerequisite for performing ex vivo MRI is the meticulous preparation and preservation of fetal brain specimens, which are more fragile than adult brains. Specimen collection, fixation, and precise time interval management collectively constitute critical factors in acquiring fetal brain specimens. Here, we present a comprehensive outline for the collection of fetal brain specimens (Fig. 2).

The proposed sample preparation pipeline integrates fixation and freezing techniques, establishing a standardized framework that can be readily adopted by clinicians and technicians. The procedures include the following steps.

• 1.

  Photograph the front and back of the fetal cadaver. Place a ruler during the photography process to enable subsequent observation of the fetal brain size.

• 2.

  Remove the fetal brain and spinal cord and photograph them from the frontal and dorsal perspectives.

• 3.

  Dissect the fetal brain and spinal cord. The dissection site is below the medulla and above the first cervical nerve root. Divide the fetal brain into left and right hemispheres for separate processing.

• 4.

  Fix the right hemisphere in formalin. Freeze the left hemisphere. Details of sampling the left hemisphere are provided below:

  • -

    For fetuses <15 gestational weeks (GW), freeze directly for preservation. If sampling is necessary, it is recommended to immerse the fetal brain in artificial cerebrospinal fluid to reduce tissue damage and improve sampling efficiency.

  • -

    For fetuses between 15 and 24 GW, it is recommended to immerse the fetal brain in artificial cerebrospinal fluid for sampling and cut 6∼8 coronal slices from front to back.

  • -

    For fetuses ≥24 GW, sample directly and cut 6∼8 coronal slices from front to back.

  • -

    Slice the brainstem and medulla thinly. Freeze the first and third slices and fix the second and fourth slices in formalin.

  • -

    For the spinal cord of fetuses ≥15 GW, freeze 3 slices (cervical, thoracic, lumbar) and place the rest in formalin.

• 5.

  Fix the right hemisphere in formalin for 2 weeks. For fetuses <20 GW, obtain coronal slices from anterior to posterior. For those ≥20 GW, follow the specified collection standards and dissect the following tissue blocks (Fig. 3):

1) Optic chiasm.

2) Frontal cortex (area frontalis).

3) Visual cortex (area 17/18).

4) Corpora mamillaria + 3rd ventricle.

5) Putamen and globus pallidus.

6) Ventricular corner R/L (corpus callosum, corona radiata, cortex, caudate nucleus).

7) Hippocampus.

8) Mesencephalon (midbrain).

9) Pons.

10) Medulla oblongata above the obex.

11) Choroid plexus.

12) Cerebellar hemisphere.

• 6.

  Conduct staining according to the following criteria: hematoxylin and eosin staining, Nissl staining, and Luxol-PAS staining.

Ex vivo MRI typically includes structural and diffusion-weighted imaging, which provides macrostructural and microstructural information about the brain, respectively. Due to the immaturity of fetal brain tissue, it exhibits considerably longer T1 and T2 relaxation times as well as higher diffusivity compared to the adult human brain [10‒12]. These tissue properties change dramatically during fetal brain development, making it challenging to obtain desirable MRI contrast. Recent advances in MRI techniques have improved its performance in ex vivo fetal brain imaging.

T1- and T2-weighted (T1w and T2w) structural MRI are useful tools for investigating the macroscopic morphology of the whole brain and tissue blocks. Higher spatial resolution and tissue contrast have been achieved through the utilization of high-field MRI scanners and optimized scanning protocols. Studies have used MRI scanners with field strength ranging from 0.5T to 11.7T to visualize human embryos, achieving a finest spatial resolution of 36 μm × 44 μm × 117 μm in T1w images [13‒15]. This exceptional resolution enables the visualization of fine structures in the developing brain [15, 16]. For T2w imaging, while single-shot sequences are predominantly employed for in utero fetal MR imaging due to their shorter acquisition times and robustness against fetal motion, multi-shot turbo-spin-echo protocols can be adopted in postmortem T2w MRI scans to achieve down to 130 μm isotropic resolution [17], while concurrently enhancing tissue contrast. T2w MR images generally show better contrast of the fetal brain than T1w images [7].

Neuroimaging studies have successfully obtained high-quality ex vivo diffusion MRI (dMRI) for investigating fetal brain tissue microstructure. For example, Huang et al. [18] (2009) scanned fetal brain specimens of 13–16 GW using an 11.7T scanner with a 30 mm inner diameter coil, achieving 200 μm isotropic spatial resolution. Optimized scanning protocols such as multi-echo diffusion-weighted sequences have been adopted to improve signal-to-noise ratio [18]. High b values up to 8,000 s/mm² are often adopted in ex vivo dMRI due to decreased diffusivity of the fixed brain specimens [19]. Quantitative dMRI models, such as diffusion tensor imaging (DTI) and spherical harmonics have been adopted to depict tissue microstructure through the modeling of directional water diffusion [20]. For instance, the DTI-derived fractional anisotropy (FA) map reveals the anisotropic organization of radial glial fibers in the ex vivo fetal brain [21], providing complementary contrast to structural MRI. Moreover, dMRI-based tractography enables the 3D reconstruction of neuronal fiber pathways by leveraging the directional information of tissue [19].

Given the progressions in imaging techniques and the advantages of ex vivo MRI, it has been widely utilized to explore the spatiotemporal patterns in human brain prenatal development. These studies mainly focused on several brain structures with the most notable developmental events during gestation, including the transient cellular compartments, the cortex, and WM pathways (Table 1). In this section, we summarized the existing ex vivo MRI studies to investigate the developing fetal brain and their major findings. It should be noted that these studies have utilized different terminologies of fetal age. Gestational age or GW is measured from the first day of the woman’s last menstrual cycle to the scanning date. By contrast, the post-fertilization, post-ovulation, or post-conception weeks are based on the time since conception which is 2 weeks less than GW. In this context, we have chosen to uniformly employ the term “GW” to define fetal age.

The human fetal brain undergoes a series of intricate cellular processes throughout gestation, during which the aggregation of neuronal cells in different developmental stages forms transient compartments. These compartments have dynamic temporal patterns as the cells proliferate and migrate, which have been successfully revealed by ex vivo MRI, as well as the evolution of their linked neuronal pathways. Visual examples are provided in online supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000542276).

One of the most noticeable changes in the fetal brain is the progressive evolution of laminar zones in the telencephalic wall. The laminar compartments include the ventricular zone (VZ), periventricular zone (PVZ), subventricular zone (SVZ), intermediate zone (IZ), subplate (SP), CP, and the marginal zone (MZ) [54]. These compartments change in size and component during neuronal proliferation and migration and ultimately develop into brain tissue including the basal ganglia, WM, and the cortex.

Ex vivo MRI can identify four to seven laminar zones of the cerebral wall during different developmental stages, starting from as early as 10 GW, with more pronounced patterns during the mid-gestation [17]. Their organization and MRI contrast change drastically due to neuronal proliferation, aggregation, and migration. The performance of laminar delineation also depends on the MRI techniques adopted. As reported by Zhang et al. [34] (2011), T1w has higher image contrast than T2w on 3.0T MRI, while T2w has higher image contrast on 7.0T MRI. On T1w 3.0T MRI, the layers could only be obscurely visualized at 14 GW and appeared clearer after 18 GW. On 7.0T MRI, four zones could be clearly recognized at 14 GW and all seven layers can be detected during 15–22 GW. Using DTI-derived FA maps, only three layers (CP, SP, and one inner layer) can be detected [35, 36].

• Ventricular zone

The VZ or germinal matrix is a cell-packed proliferative zone [55], which is the site of neuroblast differentiation [46]. It can be detected as early as 9 GW by 4.7T T1w MRI [24] and at 13 GW on 1.5T clinical MRI [46]. VZ has high T1w and low T2w signal intensity, which corresponds to the highly cellular compartment delineated by Nissl-stained sections [13, 23, 34]. Its volume exponentially increases by 23 GW and then sharply decreases at 28 GW [24]. It is almost invisible by the time of 30 GW.

• Periventricular zone

The PVZ has low T1w and high T2w signal intensity on MRI, which largely corresponds to the periventricular fiber-rich zone [23, 34]. From 18 to 20 GW, it appeared as a continuous layer encircling VZ. As gestational age increases, the periventricular fiber crossroads appear more apparent and form punctiform regions with low signal intensity that stand out in the PVZ [34]. The fibrous PVZ appears discontinuous on 3.0T MRI starting from 22 GW and continues to expand. After around 26 GW, it merges with SVZ and becomes a low T1w band.

• Subventricular zone

The SVZ is also a cell-packed proliferative zone that has low T2w and moderate T1w signal intensity compared to VZ. Between 15 and 27 GW, the intensity of neuronal proliferation decreases drastically and the relative volume of VZ and SVZ compartments shows a pronounced decline [25]. As reported by [34], the most of SVZ disappeared at 22 GW and could not be visualized at 26 GW. Its disappearance may be related to the expansion of the fibrous PVZ, which merges with SVZ during 26–34 GW and becomes a low T1w signal intensity band that continues to thicken in volume [23].

• Intermediate zone

The IZ has high T2w and moderate T1w signal intensity compared to VZ from 17 to 22 GW [23, 34]. It contains AchE-reactive thalamocortical (TH-C) fibers during this period and is almost the same intensity as SVZ on T1w MRI [23]. It is also characterized by high FA, low ADC values, and tangential orientation on DTI. It cannot be separated from other inner layers by DTI-derived maps [35]. IZ appears wide and typical on structural MRI until 24 GW when it becomes hard to discriminate from the SP. It eventually differentiates into the fetal brain WM [34].

• Subplate

The SP zone has low T1w signal intensity from 17 to 20 GW, when it is rich in extracellular matrix (ECM) with a few “waiting” TH-C axons and synapses neurons, revealed by Nissl and acetylcholinesterase staining [23]. From 21 to 28 GW, the SP widens due to the growth of callosal and cortico-cortical fibers [25]. Its T1w intensity increases and the contrast between SP and its subjacent WM decreases, while the superior part of SP remains visible as a narrow band of low MRI signal intensity below CP [25, 34]. The developmental peak of SP is around 32 GW. From 33 to 38 GW, the T1w signal intensity of SP further increases and becomes almost the same as WM. At these ages, SP is no longer visible at the bottom of the sulci and in primary gyral regions. The increase in MRI signal intensity in SP correlates with the ingrowth and accumulation of TH-C and cortico-cortical afferents, revealed by histochemical staining [23]. The DTI-derived FA value in SP is persistently low due to its richness in the ECM with nondirectional water diffusion [36].

In addition, Pogledic et al. [48] (2020) observed a thin, hyperintense layer below the CP in the upper SP portion on T2w MR images of fetal human brains from 19 to 24 GW, which has an anatomical correspondence to the histologically established superficial SP [56].

• Cortical plate

During early- to mid-gestation, CP contains densely packed radially organized neurons and has high T1w and FA intensities [35, 36]. From 17 to 20 GW, CP is identified as a band of high T1w signal intensity above SP, aligning with the layer revealed by Nissl staining [23]. During 26–34 GW, synapse formation initiates in the deep portion of CP [57]. After 35 GW, the CP shows decreased T1w and FA intensities when it becomes most of the cortical gray matter [34]. The disappearance of radial neuronal organization in the CP is due to dendritic arborization and the growth of axonal fibers, leading to a declined FA [29, 52]. The decrease varies across the CP, suggesting regional differences in maturation patterns [36].

• Marginal zone

The MZ is a cell-sparse and narrow layer above CP [41]. On T2w 7.0T MRI at 14 GW, MZ could be visualized as a high signal layer exterior to CP at the frontal, occipital lobes, and hippocampal regions (Fig. 4b). It could also be vaguely delineated on T1w 7.0T MRI at 14 GW [34]. However, it is hard to be separated from CP on 2.0 T MRI in early GW [23]. During 17–24 GW, the hippocampal MZ enlarges and becomes visible on MRI [7].

The GE is the site of origin for the basal ganglia, certain thalamic and olfactory neurons, and GABAergic interneurons of the cerebral cortex, characterized by high cell density [26]. On ex vivo MRI, the GE is characterized by high T1w signal intensity in a localized thickening of VZ [23], which is very prominent in the mid-fetal period (Fig. 4b) [7]. From 9 to 14 GW, the GE is broader at the rostral terminus aligning with the future anterior horn of the lateral ventricle, and narrows at the caudal extremity facing the thalamus [15]. At 19 GW, GE shows high FA and low ADC near the ventricular surface, and relatively low FA and high ADC at its center [35].

The basal ganglia can be recognized by ex vivo MRI starting from the second trimester. From 12 to 15 GW, the caudate nucleus is recognizable as an area of low T1w signal intensity [7]. During 17–20 GW, a larger number of basal structures can be easily recognized on T1w MRI, including the thalamus, caudate nucleus, and putamen, which all display a similarly moderate intensity (Fig. 5). They are separated by the fiber-rich internal capsule (IC), which displays lower T1w intensity [23]. After 20 GW, the dorsal thalamus, lenticular nucleus, and caudate nucleus can be differentiated on T1w images, which have increased intensities. The T1w intensity is lower in the anterior limb of IC and is lower in its posterior limb. The external capsule has lower T1w intensity compared to IC [37].

Neurons generated in the proliferative compartments migrate through certain pathways to their final destinations. These pathways can be visualized by dMRI-based tractography.

DTI shows anisotropic organization along the GE as early as 13 GW, revealing the first appearance of the periventricular fiber systems [59]. A clear radial pattern originating from the VZ and SVZ, and a tangential organization originating from the GE can be revealed at around 15 GW (Fig. 6), which are consistent with neuronal migration pathways identified by histology [19, 31]. As the migration process completes, the tangential organization of GE gradually dissolves [19]. As reported by Rados et al. [7] (2006), the GE itself cannot be seen between 24 and 28 GW. At the same time, the dominant radial organization associated with VZ and SVZ gradually diminished, starting from the dorsal parieto-occipital region and extending to the ventral frontotemporal parts [19].

The asymmetry patterns of these migratory pathways were also revealed by ex vivo dMRI. In brain specimens aged 22 GW or younger, the volumes of the left and posterior radial migration pathways were observed larger than the right and anterior ones [32]. By contrast, no such pattern was observed in the tangential pathways, suggesting distinct developmental trajectories of the radial and tangential neuronal migration pathways.

Throughout gestation, the fetal brain cortex evolves from an initially smooth surface into a complexly folded structure of intricate morphology. The time of emergence of primary sulci resembles among subjects, rendering them invaluable markers of normative development. Ex vivo MRI has helped the delineation of early sulcation of the fetal brain.

On T1w and T2w MRI, few sulci were present before 12 GW, except for the Sylvian fissure that forms early between 11 and 15 GW [25]. At 16 GW, major sulci such as the central sulcus, superior frontal sulcus, calcarine fissure, and parieto-occipital sulcus were present on ex vivo MRI (Fig. 7) [38]. By 22 GW, the lateral sulcus, cingular sulcus, posterior part of the superior temporal sulcus, and olfactory sulcus were also present. The anlage of the inferior temporal sulcus, collateral sulcus, and orbital sulcus emerged as well [38]. But the postcentral sulcus and intraparietal sulcus were still absent [37]. The sulci were at first shallow and narrow and progressively became deeper and more distinguishable, along with the emergence of secondary or tertiary branches. Based on sulcal patterns on ex vivo MRI, the maturation of the fetal brain can be evaluated and can be further utilized for gestational age estimation [60].

Quantitative methods have assisted in the assessment of fetal brain morphological maturation. Using structural MRI-based segmentation and surface reconstruction methods, the thickness, surface area, and curvature of the fetal brain cortex or CP, as well as their developmental trajectories can be characterized. During the second trimester (12–22 GW), the CP surface area and fetal brain volume exhibit a linear increase with GW, with different growth rates between brain regions [38‒40]. The volume, gyrification index, and surface area of the CP show exponential growth near term [25]. These quantitative values can be further subjected to advanced processing and analysis for additional evaluation of fetal brain structural maturation. For example, Xu et al. [40] (2021) analyzed the fetal cortico-cortical structural covariance network based on these measures.

Other subcortical patterns associated with gyrification can also be observed with dMRI, such as the regional variation in the migration pathways regression that the radial organization existed longer under the gyri than the sulci [19], as well as the emergence of cortico-cortical connections under the gyri. These observations suggest that cortical morphogenesis is highly associated with the evolution of subcortical neuronal activities, which can be explored in future investigations.

The cortico-cortical fiber pathways experienced phased developmental events before resulting in the complex architecture at birth. Histological studies demonstrated the growth and accumulation of subcortical fibers [61], the ingrowth of cortical neurons, and the emergence of cortico-cortical fibers with the development of primary gyri and sulci of the neocortex [62].

High-resolution ex vivo dMRI enables the 3D visualization of these complex fiber connections. Takahashi et al. [62] (2014) performed tractography in postmortem fetal brain samples and observed that short-range u-shaped superficial fibers emerge before gyrification (Fig. 6). The early forms of a few long-range association fibers can be observed at 17 GW, becoming more evident by term.

The cortico-cortical fibers link different cortical regions to support the information transfer between them and form a complex structural connectivity network (SCN) of the brain [63], which can be quantitatively modeled by dMRI-based fiber tractography. By defining parcellated brain regions as nodes and traced streamlines as edges, a whole-brain SCN can be established. Graph theory analysis can provide insight into the topological properties of brain connectivity. Song et al. [53] (2017) utilized a combined MRI dataset of ex vivo fetal brain specimens and in vivo preterm or term-born neonatal brains to study the prenatal developmental pattern of brain SCN. Higher network efficiency and network strength were observed during 20–40 GW. High-resolution ex vivo MRI helped distinguish the non-WM neural fibers, which could be a major contributor to the SCN during early gestation according to their results.

The major WM pathways develop starting from the second trimester of gestation [64]. This developmental process is characterized by the progressive formation of axons and their subsequent fasciculation, establishing long-distance WM fiber bundles, followed by oligodendroglial activities and myelination [65, 66]. These WM connections play an integral role in supporting the neural networks. Tractography based on dMRI is a powerful tool for visualizing fiber pathways of the brain. Ex vivo dMRI studies have enabled the early detection of emerging WM pathways and their developmental patterns across different stages (Fig. 8) [36, 27].

Limbic fibers are among the earliest developing ones. The fornix and the stria terminalis can be visualized on DTI-derived colored FA maps at 13 GW [27]. Parts of the fornix can be revealed by DTI tractography at 13 GW [29], and more complete at 21 GW [19, 59]. The cingulum bundle is only visible after 17 GW [29, 27].

The middle cerebellar peduncle (MCP) is visible on DTI at 13 GW and becomes clearer during 15–21 GW [18, 27]. The corpus callosum (CC) appears in DTI-based tractography at 15 GW [27], agreeing with histological studies [4, 67]. In the flowing weeks, fibers from CC spread in both anterior and posterior directions partly reached the cortical surface at 17 GW [68]. At 19 GW, the genu of CC is more evident than the splenium, indicating its anterior-to-posterior order of development [36].

The TH-C fiber system can be detected by DTI tractography at 13 GW [30]. At 15 GW, DTI tractography reveals the extension of TH-C fibers into the deep portion of CP [30], corresponding to the ingrowth of TH-C afferents revealed by histology [69]. They continue to develop and the anterior and middle thalamic pathways converge with cortical pathways between 17 and 21 GW [19, 68]. After 22 GW, posterior thalamic pathways converge with cortical pathways and TH-C fibers can be traced by DTI tractography within the anterior, superior, and posterior thalamic radiations [36]. In some cortical regions, the TH-C fibers converge with CC [68]. The corticospinal fiber tract (CST) can be observed by DTI at 15–17 GW and increases its size from 17 to 23 GW [59, 27].

Several long-range association fibers emerge in the fetal brain during the second to third trimester. Before 15 GW, DTI only reveals parts of the developing fronto-occipital fasciculus [59]. A more complete inferior fronto-occipital fasciculus (IFOF) is visible at 17 GW on DTI [27]. At the same time, the uncinate fasciculus (UNC) can be observed connecting the frontal and temporal lobes [19]. The inferior longitudinal fasciculus (ILF) can be traced starting from 19 GW and clearer at 31W [19]. Using traditional DTI-based tractography, the superior longitudinal fasciculus (SLF) cannot be detected during the second trimester and is not a significant event at birth [36]. However, parts of the SLF can be identified by HARDI tractography in 26 GW [45], suggesting the superior ability of HARDI compared to DTI in the detection of small fiber structures.

The sections above have summarized a series of studies demonstrating the ability of ex vivo MRI to investigate normal fetal brain development. The revealed developmental patterns may serve as a normative reference for the early detection of fetal brain abnormalities. On the other hand, postmortem fetal brain MRI has found application in the direct assessment of aberrant fetal brain conditions, in which its reliability as an evaluative tool has been tested [70].

Conventional autopsy has been the reference standard for postmortem elucidation of the cause of fetal death and for investigating relevant pathologies. In some situations, however, a detailed examination of the intricate connections and anatomical structure of the brain is best interpreted by 3D visualization.

Various studies have compared postmortem MRI to conventional autopsy in detecting and evaluating fetal brain abnormalities. MRI showed a high ratio of agreement with autopsy with high overall sensitivity and specificity and even provides additional diagnostic information [71‒77]. Arthurs et al. [78] (2015) tested the diagnostic accuracy of postmortem MRI for identifying intracranial hemorrhage, small isolated intraventricular bleeds, and ventricular dilation in fetuses and children. They found that postmortem MRI has a sensitivity of >87% for detecting brain malformations, intracranial bleeds, and neurological causes of death. It also provides important diagnostic information in over 50% of the fetuses where conventional brain autopsy is non-diagnostic. This suggested that ex vivo MRI of the fetal brain may serve an important role in the investigations of the neurological cause of fetal death and relevant pathologies. As a matter of fact, MRI is already part of routine postmortem fetal brain examination in some centers [79].

In addition, quantitative MRI techniques have comparable performance to traditional autopsy in evaluating fetal brain abnormalities. The fetal brain maceration score is associated with ADC as well as conventional autopsy [80, 81]. Fetal brain weight estimated by postmortem MRI shows good correspondence with conventional autopsy [77].

MRI outcomes were also leveraged for imaging-pathology correlation analysis in a fetus with cobblestone-like brain malformation, which helped the detection of a gene-related brain abnormality in a sibling [82]. This study shows the capability of ex vivo MRI to combine with pathological and genetic analyses in the investigation of fetal brain abnormalities.

Ex vivo fetal brain, MRI offers invaluable insights into the early stages of brain development. Our comprehensive review underscores how ex vivo MRI has significantly contributed to the exploration of diverse aspects of the developing human brain, including brain morphology, neuronal proliferative compartments, neuronal fibers, and the evaluation of prenatal neurological abnormalities. The alignment between ex vivo MRI findings and histological observations [13, 23, 36] attests to its efficacy in unraveling the patterns of fetal brain development.

Nevertheless, there are several limitations regarding the stability and sensitivity of ex vivo MRI studies. First, postmortem interval (PMI) and fixation duration can affect the results of ex vivo MRI. The water-rich fetal brains can be easily damaged if not properly handled. The difficulties in specimen collection and fixations, as well as the scarcity of subjects lead to the limited sample size of postmortem fetal brain studies. Second, postmortem MRI cannot completely replace autopsy or histology involving intricate microscopic observations of organs and tissues [83]. Most ex vivo MRI protocols cannot achieve comparable resolution to histology, thus limiting its capability in studying brain microstructure. Although DTI-based tractography is useful in visualizing macroscopic fiber bundles, it cannot always capture the emergence of neuronal pathways as revealed by histological staining, especially from a microscopic point of view [29]. It is difficult to distinguish axons, glial fibers, or aligned cell components in MRI tractography, for it only captures the anisotropies of the imaging signals but overlooks the underlying cellular and molecular components. Therefore, careful histological validations are important in ex vivo MRI analysis.

In the future, integration of multimodal MRI data can promote the study of specific scientific questions. Takahashi et al. [19] (2012) observed that the emergence of cortico-cortical connectivity coincided with the timing of cortical folding, which raised the question of whether the development of connectivity is related to gyral formation. Solving such questions requires information from both structural and dMRI. In addition, better integration of ex vivo MRI with histological and genetic analysis should be emphasized. According to our review, most studies compared MRI and histological findings qualitatively, without a systematic correlation between MRI and histological measurements. Better imaging-histology and imaging-pathological correlations may provide stronger support for the clinical use of ex vivo MRI. The genetic correlation of ex vivo MRI findings has only been investigated in a limited number of studies [14, 82], and thus needs further exploration. Moreover, ex vivo MRI scans can serve as anatomical references in genetic analyses, e.g., in the transcriptional landscape of the prenatal human brain [84], thereby giving invaluable support for related inquiries.

The authors have no conflicts of interest to declare.

This work was supported by the Ministry of Science and Technology of the People's Republic of China (2021ZD0200202), the National Natural Science Foundation of China (81971606, 82122032), and the Science and Technology Department of Zhejiang Province (202006140, 2022C03057).

Each named author has substantially contributed to conducting the research and drafting this manuscript. Jing Zhang and Dan Wu: conceptualization; Ruike Chen, Chen Tian, Keqing Zhu, Guoliang Ren, Aimin Bao, Yi Shen, and Xiao Li: design; Ruike Chen and Chen Tian: investigation, visualization, and writing – original draft; Ruike Chen, Chen Tian, Keqing Zhu, Guoliang Ren, Aimin Bao, Yi Shen, Xiao Li, Yaoyao Zhang, Wenying Qiu, Chao Ma, Jing Zhang, and Dan Wu: interpretation and writing – review and editing; Dan Wu and Jing Zhang: supervision and project administration; and Dan Wu: funding acquisition. All authors of this paper have read and approved the final version submitted. Ruike Chen and Chen Tian contributed equally to this work.

Ruike Chen and Chen Tian contributed equally to this work.