Introduction

Traditionally, cerebellar physiological function has long been confined to its motor control features; however, it has now become evident that the cerebellum plays a crucial role for many non-somatic functions other than the coordination of movement, and is engaged also in the regulation of cognition and emotion [1–6].

The best clinical evidence in humans, supporting the role of the cerebellum in non-somatic functions, is the cerebellar cognitive affective syndrome [6].

This syndrome, which occurs following lesions of the cerebellar posterior lobe but not the anterior lobe, is characterized by a range of executive, autonomic, visuospatial, linguistic, and affective deficits.

This constellation of non-somatic symptoms has also been reported in subsequent studies in adults and children, and the observation that the syndrome results from lesions of the cerebellar posterior lobe has also withstood scrutiny [7–14].

Interestingly, CCAS may occur in the absence of motor impairment corroborating the concept of a sensorimotor-cognitive anatomo-functional dichotomy in cerebellar organization [14–17].

Unfortunately, in contrast to the well-known regional cerebellar anatomy for motor control [18, for review], the anatomical connectivity underlying cerebellar modulation of cognition and emotion remains unclear.

In addition to corticopontine afferents from motor and non-motor regions, the cerebellum also receives inputs from the medial mammillary bodies, engaged in memory [19, 20], and multimodal deep layers of the superior colliculus.

Moreover, the cerebellum has reciprocal connections with the hypothalamus [20, 21] and with brainstem areas (ventral tegmental area, periaqueductal gray, and locus coeruleus) that are connected with limbic and paralimbic regions [22].

In terms of functional cerebello-hypothalamic connectivity, several authors showed that electrical stimulation of the cerebellum produced evoked responses in the amygdala and hippocampus and altered abnormal discharges in the latter [23, 24]. Furthermore, Moruzzi [25] showed that stimulation of the cerebellum altered the phenomenon of sham rage derived from hypothalamic stimulation [25, 26], suggesting that the cerebellum may act on behavioral responses belonging to complex autonomic reaction.

In line with this hypothesis, Snider showed the presence of anatomical projections from the fastigial nucleus to the hypothalamus [27]. Hereafter, many other authors provided evidence of interconnections between the cerebellum, hypothalamus, and limbic system [27–30].

Interestingly, anatomical and physiological evidences have now identified vermis as the principal target of such limbic connections [3, 31, 32].

Indeed, stimulation of the cerebellar vermis modulates firing patterns in the hippocampus, amygdala, and septum [33–35] and it can ameliorate aggression [36].

In a previous diffusion tensor imaging (DTI) study, we found the existence of a direct cerebello-limbic connection between the vermis; cerebellar lobules VIII, IX, X, crus I, and crus II; and the hippocampus, thus providing a new anatomical framework for the comprehension of the non-somatic cerebellar functions [37].

DTI is a MRI technique that allows to visualize in vivo white matter fibers, estimating properties of preferential anisotropic diffusion of magnetically labeled water into myelinated axons [38, 39]. Constrained spherical deconvolution (CSD) is a robust diffusion modeling technique which allows to overcome some of the most important limitations of the DTI approach, such as partial volume effect and the inability to resolve complex fiber configurations within a single voxel [40, 41]. Using this technique, we were able to reconstruct a limbic, basal ganglia, and cerebellum connectome in both normal and pathological conditions [37, 42–46].

In the present paper, we would like to further extend our previous work on cerebello-limbic pathways by evaluating the connectivity between the cerebellum and hypothalamus using CSD tractography. We hypothesized that the cerebellum may receive limbic inputs from the mammillary bodies and/or may give an output to the hypothalamus, thus influencing autonomic functions, visuospatial orientation, and memory functions. The results of this study could improve our understanding of non-motor cerebellar functions and may shed new light on several diseases where clinically relevant impairments in navigational skills or memory may become manifest at early stages of the disease.

Materials and Methods

Subjects

We recruited 13 healthy humans with no history of any overt neurological or psychiatric diseases. The research followed the tenets of the Declaration of Helsinki; written informed consent was obtained from all included subjects, after explanation of the nature and possible consequences of the procedure. The study was approved by the institutional review board of IRCCS Bonino Pulejo, Messina, Italy (Scientific Institute for Research, Hospitalization and Health Care), protocol number 15/2012.

Data Acquisition

We performed the following MRI sequences with a 3T Achieva Philips scanner using a 32-channel SENSE head coil:

  1. 1

    Three-dimensional high-resolution T1-weighted fast field echo (FFE) sequence was acquired using the following parameters: repetition time (TR) = 25 ms, echo time (TE) = 4.6 ms, flip angle = 30°, field of view (FOV) = 240 × 240 = mm2, reconstruction matrix = 240 × 240 voxel, voxel size = 1 × 1 × 1 mm, and slice thickness = 1 mm.

  2. 2

    Three-dimensional high-resolution T2-weighted turbo spin echo (TSE) sequence was obtained using the following parameters: TR = 2500 ms, TE = 380 ms, FOV = 250 × 250 mm2, reconstruction matrix = 312 × 312 voxel, voxel size = 0.8 × 0.8 × 0.8 mm, and slice thickness = 0.8 mm.

  3. 3

    A dual-phase encoded pulsed gradient spin echo diffusion-weighted sequence [47] using 60 gradient diffusion directions was chosen following an electrostatic repulsion model [48]. The other sequence parameters were as follows: diffusion-weighting b-factor = 1000 s/mm2, TR = 11,884 ms, TE = 54 ms, FOV = 240 × 240 mm2, reconstruction matrix = 120 × 120 voxel, in-plane voxel size = 2 × 2 mm2, axial slice thickness = 2 mm, and no interslice gap.

Preprocessing and Coregistration

SPM8 MATLAB tool (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/) was used to correct diffusion images for motion as well as for eddy currents distortion artifacts; a rotational part of transformations was later applied to individual gradient directions. High-resolution T1 and T2 images were then coregistered to preprocessed DWIs using a pipeline outlined in [49] and used in our previous work [44]. In brief, using new segment option of SPM8, we extracted cerebrospinal fluid (CSF) component probability masks from b0 image and structural scans. CSFs of structural scans were subsequently warped to match the CSF estimated from the b0 image using FLIRT and FNIRT FSL utilities (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/). Estimated warping fields were eventually applied to structural scans. Thanks to abovementioned workflow we were able to align structural scans onto diffusion images with high precision, thus reducing possible misalignment biases coming from usage, in diffusion images’ space, of region of interest (ROI) segmented in space of structural scans.

Segmentation

On T1-weighted images, segmentation of the mammillary bodies was manually performed by a skilled neuroradiologist as follows: firstly, the individual volumes obtained from the 3D T1-weighted FFE sequence were loaded into the viewer; secondly, the contrast values in the viewer were set to maximally increase the visibility of the areas; thirdly, the axial view was zoomed to facilitate the drawing of the mammillary bodies mask; and fourthly, the mammillary bodies were visualized in coronal and sagittal views to better define it.

The segmentation of the dentate nucleus was manually performed on T2-weighted TSE images, which permitted to obtain high-resolution images with a relatively short acquisition time. At the same time, this sequence allowed to obtain a fine representation of the iron-loaded nuclei due to a T2* effect linked with the use of a very long echo time. The right and left dentate nuclei were depicted as hypointense bean-shaped structures and were traced individually on axial slices of the coregistered MR images and then saved as ROIs using the viewer provided with the MRtrix3 software.

In addition, we segmented the substantia nigra and red nucleus bilaterally to be set as regions of avoidance (ROAs).

Volumetric segmentation of the cerebellum was performed on the coregistered T1 images with the FreeSurfer image analysis suite by using the Desikan-Killiany atlas, which is documented and freely available for download online (http://surfer.nmr.mgh.harvard.edu/). Successively, the segmentations obtained from each individual were visually inspected and, if needed, manually edited.

Tractography

We modeled diffusion signal by using a modified high angular resolution diffusion imaging (HARDI) technique, namely non-negative CSD, that is able to estimate the fiber orientation distribution function (fODF) directly from deconvolution of DW signal with a reference single-fiber response function [50]. Response function was estimated from the data on the basis of voxels having a fractional anisotropy (FA) index above 0.7, by using MRtrix software (http://jdtournier.github.io/mrtrix-0.2/index.html). Probabilistic streamline fiber tracking was performed by calculating the fODF peak direction closest to the previous stepping direction using trilinear interpolation.

By using CSD, we were able to overcome the partial volume effects associated with DTI and also to improve, comparing with other types of HARDI acquisitions while discarding QBI (Q-balls imaging) and DSI due to long acquisition times required [38]. On the one hand, if higher b values permit to resolve smaller angles among fibers [50, 51], they increase the probability to find motion and eddy currents related artifacts due to longer acquisition times. For this reason, we preferred a lower b value in order to obtain the best quality speed trade-off.

We performed probabilistic CSD-based whole-brain tractography, using the following parameters: spherical harmonic degree = 8, step size = 0.2 mm, maximum angle = 10°, maximum fiber length = 150 mm, minimum fiber length = 40 mm, and minimal fODF amplitude = 0.15.

CSD combined with a probabilistic approach allows to track streamlines even in voxels with complex fiber geometries, though the number of false positive tends to increase. For this reason, we used a more conservative approach with respect to usual standards setting the minimal fODF amplitude = 0.15, thus keeping to a minimum false-positive fiber bundles. In this way, the tracking procedure involved only voxels with high probability to belong to the white matter [52, 53].

Before running tractography, we applied a small dilatation to WM masks in order to allow streamlines to reach our ROIs, placed in GM, for subsequent analyses. Target ROIs were moderately dilated to include gray/white matter boundaries, thus ensuring that streamlines were able to reach target ROIs for subsequent connectivity analysis.

For visualization purposes, we reconstructed a color-coded map in which red, blue, and green colors indicate the principal streamline directions [54]. Specifically, red color indicates a left–right pattern, green color an anterior–posterior pattern, and blue color a caudal–cranial pattern.

Connectivity Analysis

We extracted, for each subject, streamlines connecting the right and left mammillary bodies with inclusion masks. Connectivity was defined as the percentage of streamlines ending in each ROI (dentate nuclei and cerebellar cortices). Such numbers are used as markers of connectivity density, in both healthy and pathological conditions [55–58].

Since it is documented that streamlines are more likely to connect with larger ROIs or closer ROIs, we took these into account for volume and distance biases. Hence, we performed weighted connectivity defined as the density of the tracts connecting a couple of ROIs, scaling the number of streamlines by the sum of inverse of the fiber length and the mean volume of two ROIs [59].

Statistical Analysis

Connectivity, FA, mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), and tract length mean are the parameters considered for each pathway. A non-parametric analysis was carried out because the results of the Shapiro-Wilk normality test indicated that most of the target variables were not normally distributed. The numerical data are presented in median and first–third quartiles for non-normal distributions. The Wilcoxon signed-rank test was used in order to compare, for each pathway (mammillary bodies-ipsilateral cerebellar cortex, mammillary bodies-contralateral cerebellar cortex, mammillary bodies-ipsilateral dentate nucleus, mammillary bodies-contralateral dentate nucleus), the results between the right and left sides (intra-pathway analysis) in each parameter. Spearman’s rank correlation was used to assess whether there was a relationship between connectivity indices and other parameters. Analyses were performed using an open source R3.0 software package. A 95 % of confidence level was set with a 5 % alpha error. Statistical significance was set at p < 0.05.

Results

In total, 13 subjects (seven males and six females) were enrolled with a mean age of 35.46 ± 11 years.

In all subjects, we traced bilaterally fiber bundles connecting the cerebellum with the mammillary bodies. In particular, we isolated fiber bundles between the dentate nucleus and mammillary bodies, passing through the ipsilateral superior cerebellar peduncle (Fig. 1), ascending the medial mesencephalic tegmentum at the level of the inferior colliculi, surrounding the red nucleus, bending around the medial face of the substantia nigra, and finally reaching the mammillary body. Figure 2 shows in axial slices the anatomical course of the connections between the dentate nucleus and mammillary bodies. We identified both ipsilateral and contralateral components, the latter decussating at the level of midbrain (Fig. 3).

Fig. 1
figure 1

Ipsilateral dentate nucleus-mammillary connections. Parasagittal views of left (A) and right (B) pathways connecting the dentate nucleus (orange Volume of Interest (VOI)) and ipsilateral mammillary body (purple VOI) via the superior cerebellar peduncle. The enlarged views better depict streamlines avoiding the substantia nigra (cyan VOI) and red nucleus (red VOI)

Full size image
Fig. 2
figure 2

Course of ipsilateral dentate nucleus-mammillary connections. Ten axial T2-weighted images show the course of the connections between the dentate nucleus and mammillary bodies. On the right side, fibers are red colored and, on the left side, blue colored. The first three images depict tracts leaving the dentate nuclei, and then fibers pass through the superior cerebellar peduncles, surround red nuclei, passing medially to the substantia nigra, and finally reach the mammillary bodies

Full size image
Fig. 3
figure 3

Contralateral dentate nucleus-mammillary connections. A Representation in coronal view of the pathway connecting the right mammillary body (purple VOI) with contralateral dentate nucleus (orange VOI). B Coronal view shows the connection between the left mammillary body (purple VOI) and contralateral dentate nucleus (orange VOI). The inset in mid position focuses on the fibers exiting the superior cerebellar peduncles and decussating at the level of midbrain, below, and medially to red nuclei (red VOIs)

Full size image

In addition, we found a less strong connectivity between the cerebellar cortex and the mammillary bodies. Starting our description from the cerebellum, these tracts originated in the anterior cerebellar cortex (vermis; lobules VIIB, VIIIA, VIIIB, X, crus I, and crus II), reached the medial part of the cerebellum, ascended the ipsilateral middle cerebellar peduncle, and finally reached the mammillary bodies (Fig. 4). Representation of the course of the connections between the cerebellum and mammillary bodies is depicted in axial slices in Fig. 5. Also for this pathway, we demonstrated the presence of both ipsilateral and contralateral connections, decussating at the level of the upper pons, as shown in Fig. 6.

Fig. 4
figure 4

Ipsilateral cerebellar cortex-mammillary connections. Parasagittal views of left (A) and right (B) pathways connecting the cerebellar cortex and ipsilateral mammillary body (purple VOI) via the middle cerebellar peduncle. In this case, fibers avoid the dentate nucleus (orange VOI). The enlarged views better depict streamlines avoiding the substantia nigra (cyan VOI) and red nucleus (red VOI)

Full size image
Fig. 5
figure 5

Course of ipsilateral cerebellar cortex-mammillary connections. Ten axial T2-weighted images show the course of the connections between the cerebellar cortex and mammillary bodies. On the right side, fibers are red colored and, on the left side, blue colored. The first row shows streamlines leaving the cerebellar cortex and passing through the middle cerebellar peduncles. In the second row, at the level of midbrain, fibers encircle red nuclei, pass medially to the substantia nigra, and finally reach the mammillary bodies

Full size image
Fig. 6
figure 6

Contralateral cerebellar cortex-mammillary connections. A Coronal view shows the connection between the right mammillary body (purple VOI) and contralateral cerebellar cortex, decussating at the level of the pons. B Coronal view shows the connection between the left mammillary body (purple VOI) and right cerebellar cortex, decussating at the level of the pons. Note that these pathways also run externally to dentate nucleus (orange VOI). The enlarged view magnifies streamlines crossing the midline at the level of the pons and bypassing the red nuclei (red VOIs) and substantia nigra (cyan VOIs)

Full size image

Figure 7 reports the box plot of the pathway for each parameter. Intra-pathway analysis highlighted no significant differences for side factor in the pathway of each parameter (p > 0.05) (Table 1). A significant correlation between the left mammillary bodies-ipsilateral cerebellar cortex connectivity and RD (r = −0.621, p = 0.026) as well as a trend toward the correlation between the left mammillary bodies-contralateral dentate nucleus connectivity and FA (r = −0.527; p = 0.067) were observed (Fig. 8).

Fig. 7
figure 7

Representation by box plot of parameters for each pathway in the right and left sides: a connectivity, b fractional anisotropy, c mean diffusivity, d axial diffusivity, e radial diffusivity, and f tract length mean. I_CC_R right mammillary bodies-ipsilateral cerebellar cortex, I_CC_L left mammillary bodies-ipsilateral cerebellar cortex, C_CC_R right mammillary bodies-contralateral cerebellar cortex, C_CC_L left mammillary bodies-contralateral cerebellar cortex, I_DN_R right mammillary bodies-ipsilateral dentate nucleus, I_DN_L left mammillary bodies-ipsilateral dentate nucleus, C_DN_R right mammillary bodies-contralateral dentate nucleus, C_DN_L left mammillary bodies-contralateral dentate nucleus

Full size image
Table 1 Intra-pathway differences of pathway values in the right and left sides
Full size table
Fig. 8
figure 8

Significant correlation between the connectivity and diffusion parameters. a Scatter plot of connectivity and radial diffusivity for mammillary-ipsilateral cerebellar cortex pathway in the left side. b Scatter plot of connectivity and fractional anisotropy for left mammillary-contralateral dentate nucleus pathway

Full size image

Discussion

In the present study, we showed the existence of connections between the mammillary bodies with both the cerebellar cortex and dentate nucleus in the human brain.

These findings are in line with previous studies performed in rodents and primates [60–62]. In particular, afferent connections from the lateral mammillary nucleus to vermis lobule IX (uvula) and the cerebellar anterior lobe in exemplars of three shrew (Tupaia glis) were reported, while no reciprocal efferent projections between the cerebellar cortex and cerebellar nuclei to the mammillary bodies were found [63]. Similar results have been reported in a prosimian primate (Galago crassicaudatus) [64]. Finally, studies performed on the squirrel monkey described a slightly different network, where hypothalamo-cerebellar fibers mainly reach lateral mammillary and supramammillary nuclei, connecting them with paraflocculus, dorsal paramedian lobule, and lobule V of the anterior lobe, while the ansiform lobule is reached by projecting fibers from medial mammillary and supramammillary nuclei. As described in previous experiments, these connections have both ipsilateral and contralateral components, with an ipsilateral preponderance with few sparse efferent connections arising from globose and fastigial nuclei directed to the contralateral medial mammillary nucleus [20]. Finally, a study focused on connections between the cerebellar nuclei and hypothalamus performed on Macaca fascicularis revealed that all the deep cerebellar nuclei (dentate, fastigial, globose, and emboliform nuclei) received afferents from lateral and medial mammillary, supramammillary, and tuberomammillary nuclei; on the other hand, no significant efferent projections to the mammillary bodies were reported [65]. Pooling together the literature, it looks like that there is a predominant hypothalamic-cerebellar direction. A comparison between data from studies in animals and our data is shown in Table 2.

Table 2 Overview of cerebellar-mammillary connections in animals
Full size table

In the present study, we found that the mammillary bodies have a strong connectivity with the dentate nucleus and a less strong connectivity with the cerebellar cortex. In particular, fibers terminated in lobules VI, VII, VIII, crus I, and crus II and vermis in the posterior part of the cerebellum.

This is in line with the physiological view of a functional compartmentalization of the cerebellar cortex where the anterior lobe is related to sensorimotor functions while the posterior lobe (in particular, lobules VI and VII) may have a predominant cognitive function, and the posterior vermis may be involved in affective processes [66].

The existence of these parallel distinct networks may explain why in patients with cerebellar lesions cognitive and affective symptoms can even occur in the absence of motor symptoms [67–69].

The hypothesis of an “emotional” cerebellum involving posterior vermian and paravermian regions is also suggested by functional neuroimaging studies that have shown a significant posterior cerebellar activation during recognition of (i) facial emotions [70], (ii) pleasant vs unpleasant visual stimuli [71], (iii) subjective mood of sadness and anxiety [72, 73], (iv) positive mood states [74], and (v) fear conditioning and fear learning [75].

Physiological Relevance of Cerebellar-Hypothalamic Connectivity

Despite we cannot establish the directionality of cerebellar-mammillary connections, the present work may open a window to some physiological speculations.

It can be hypothesized that the cerebellum may receive limbic inputs from the mammillary bodies and/or may give an output to the hypothalamus, thus influencing vegetative and memory functions.

The involvement of the cerebellum in regulating autonomic functions has been reported in many animal studies (control of blood pressure [76], control of visceral, vegetative, and immune functions, reviewed in [77]). On the other hand, the role of the cerebellum on autonomic functions, in humans, is still a matter of debates, although some data are now emerging.

For instance, Schmahmann et al. reported that hunger significantly increased the regional cerebral blood flow (rCBF) in a number of brain regions of human subjects including the hypothalamus, limbic structures, as well as bilateral lobule V of the cerebellar cortex in the paravermian regions [78]. In addition, in a functional magnetic resonance imaging (fMRI) study, the cerebellum, supplementary motor area, somatosensory cortex, anterior cingulate, and orbitofrontal cortices were activated in a minute after glucose acquisition through a per-oral rubber tube [79].

In line with a role of the cerebellum on autonomic functions, it was shown that a fear-conditioned heart rate increase was impaired in patients with medial cerebellar lesions [80]. It has been postulated that direct cerebello-hypothalamic projections may modulate cardiovascular regulation by acting on brainstem neurons sensitive to baroreflex and osmotic pressure [81].

Regarding emotion and cognition, Reiman et al. revealed an activation in lateral parts of both cerebellar hemispheres (lobule VI) as well as in the other central structures subserving emotional expression such as the hypothalamus, amygdala, hippocampus, and occipito-temporo-parietal and anterior temporal cortices, in human subjects passively viewing film clips that generated feelings of happiness, sadness, or disgust [82]. Moreover, several authors reported that the cerebellum is implicated in sensory, emotional, attentional, and cognitive processes independently from motor involvement [83–86].

Relevance of Cerebello-Mammillary Connections on Visuospatial Orientation and Memory

Clinical evidences in patients with cerebellar lesions indicated that visuospatial ability was impaired in patients with cerebellar damage [87].

This notion is supported from recent neuroimaging data showing cerebellar activation during simple spatial tasks such as line bisection judgment [88] and during the mental rotation of objects [89].

In addition, cerebellar circuits play a pivotal role for acquisition of the procedural components required for spatial learning [90].

Neuroimaging findings in young individuals and lesion studies have identified a complicate network involved in human navigation, including the hippocampus, parahippocampal gyrus, cerebellum, parietal cortex, and retrosplenial cortex [91–93].

A specific role of the cerebellum in spatial navigation has been recently suggested by Rochefort and colleagues who hypothesized a direct pathway linking cerebellum and hippocampus [94].

In keeping with this hypothesis, we have recently demonstrated in normal humans, by using CSD tractography, the presence of a direct connection between the cerebellum (vermis; lobules VII, IX, X, crus I, and crus II; and fastigial nucleus) and the hippocampus [37].

Recent advances in our understanding of the mammillary bodies and their connections have shed new light on the possible role of these structures in high-order functions, such as spatial memory, thus challenging the prevalent hippocampal-centric view of mammillary body function and considering a wider network of structures interconnected with mammillary bodies [95]. In this perspective, a direct cerebello-mammillary pathway may be another piece of this complicate circuit implicated in visuospatial orientation.

Although the mammillary bodies have been implicated in amnesia perhaps for longer than any other single brain region, their role in memory has remained elusive.

This uncertainty persists despite the mammillary body pathology has been linked to the amnesic Korsakoff’s syndrome [96]. Indeed, mammillary body damage is not sufficient alone to account for all of the memory deficits that are associated with Korsakoff’s syndrome, considering that mammillary body degeneration is sometimes found in the absence of amnesia. On the other hand, it has been reported that anterior thalamic nucleus pathology was the best predictor of memory loss in patients with Korsakoff’s syndrome, Wernicke’s syndrome, and alcoholism [97, 98].

Limitations

It is known that tractography suffers from inherent technical limitations. First of all, tractography is not able to offer a direct visualization of axons, whose diameter is typically less than 10 μm, but only a possible reconstruction of their trajectory according to local water diffusion anisotropy. Although our results are anatomically plausible compared with previous anatomical descriptions of these circuits, tractographic findings should be carefully considered and interpreted with caution. In the present work, we decided to use more conservative tracking criteria in respect to usual standard [44, 52, 99], at the cost of an underestimation issue, thus making our tractographic findings more reliable. Another important drawback of this technique is the lack of information regarding the directionality of the signal transmission [100, 101] and, therefore, whether the pathway that we revealed in the present study is hypothalamo-cerebellar, cerebello-hypothalamic, or reciprocal. Finally, tractography is unable to detect synapses and gap junctions, and thus, it does not allow to determine if a neural connection is mono- or polysynaptic. Therefore, further studies based on functional connectivity approach and postmortem microsurgery dissection and/or anterograde and retrograde tracer studies in animals are recommended in order to confirm definitively the existence of this pathway and to give more insights about its functions. Anyway, tractography still remains the only existing technique allowing to investigate neural connections in the human brain in vivo and non-invasively.

Conclusions

In conclusion, we demonstrated, in vivo, in humans, the presence of direct cerebello-mammillary connections in keeping with animal data, described in cats, rodents, and primates. These findings might be potentially useful to explore pathological conditions affecting cerebellar-limbic connections involved in autonomic functions, visuospatial orientation, and memory.