Material Didático


Volumetry and diagnostic MRI evaluation of hippocampal sclerosis



Fernando Cendes


Key Points


1.         Thin coronal MRI slices, perpendicular to the axis of the hippocampus, give the best images for determining hippocampal sclerosis (HS) and other subtle pathologies and for ascertaining anatomical detail.


2.         MRI features of HS, detectable by visual inspection of the images, are


                        - hippocampal smallness (atrophy) which is the most specific and reliable feature

                        - increased T-2 signal which in isolation may be insufficient to diagnose HS

                        - loss of internal structure


There may also be asymmetry of the horns of the lateral ventricles, which is variable and may lead to false lateralisation, and atrophy of the anterior temporal lobe which is non-specific.  T-2 mapping is an objective method for measuring abnormal T-2 signal which may be difficult to detect visually.


3.         Most patients with HS undergoing presurgical evaluation have one hippocampus which is clearly smaller than the other on visual inspection, and which has increased T-2 signal, along with a normal appearing contralateral hippocampus, so that volume measurement is not necessary for clinical purposes.  The visual binary paradigm breaks down in the presence of symmetric bilateral atrophy or mild unilateral disease.  In these cases volumetric MR analysis of the hippocampus and the amygdala are very sensitive and specific for identifying HS.


4.         MRI measurements of hippocampal volumes are a surrogate for histopathological methods of assessing the presence and severity of neuronal loss in each hippocampus allowing each to be classed as normal or abnormal.  This may give useful prognostic information concerning postoperative seizure control.  Surgical treatment of strictly unilateral HS should give >90% excellent outcome.





Imaging of temporal lobe structures



The temporal lobes can be divided into two compartments: The neocortex and medial temporal lobe structures that include the uncus, amygdala, hippocampus and parahippocampal gyrus. This subdivision follows the clinical-EEG classification of temporal lobe epilepsy: limbic or medial versus neocortical.

The lateral temporal cortex can be well evaluated by MRI. The anterior, superior, lateral and medial aspects are well established, but the posterior end joins the parietal and occipital lobes, without any clearly defined boundary, except the inconspicuous temporo-occipital incisure. The superior and inferior temporal sulci divide the lateral surface of the temporal lobe into three gyri: superior, middle and inferior (Figures 1 and 2) 1.

Figure 1. 1mm thick, sagittal images obtained with a Gradient Echo 3D T1-weighted sequence in a normal control subject.

A: Sagittal cut passing through the mid portion of the temporal lateral cortex, showing the superior (small black arrow) and inferior (white arrow) temporal sulci dividing the superior (T1), middle (T2) and inferior (T3) temporal gyri.

B: Sagittal slice 6mm medially from A. The margin of the planum temporale and planum polare below the insula (I) 

C: Sagittal slice 12mm medially from B.

 a: amygdala; h: hippocampus; fg: fusiform gyrus; c: collateral sulcus.

D: Sagittal slice 6mm medially from B. Note the amygdala lying anterior and superior to the head of hippocampus. The two structures are separated by a white line (alveus).



Figura 2. Sagittal T1 MRI at the level of hippocampus, showing the optimal orientation for obtaining coronal images for evaluating mesial temporal structures.


The superior temporal gyrus (T1) runs parallel to the lateral fissure. Its anterior end is a part of the temporal pole. The upper margin of the superior temporal gyrus forms the temporal operculum. It continues into the lateral fissure by a large cortical area, sometimes called the superior surface of the temporal lobe. This surface can only be observed when the superior overlying margin of the lateral fissure has been removed. From front to back the superior surface of the temporal lobe is divided into 3 parts: the planum polare, the transverse temporal gyri, and the planum temporale. The planum polare is separated from the insula by the inferior circular sulcus (Figure 1) 1,2.

The middle temporal gyrus (T2) runs from the temporal pole through the occipital lobe without any clear boundary.  The inferior temporal gyrus (T3) is not very visible on the lateral surface and is caudally separated from the occipital lobe by the temporo-occipital incisure (Figures 1 and 2).

The inferior surface of the temporal lobe consist of three gyri: the inferior temporal gyrus (T3), the fusiform gyrus (T4) and parahipocampal gyrus (T5). The fusiform gyrus is well delimited by the lateral occipito-temporal sulcus laterally, the collateral or medial occipito-temporal sulcus medially, and the anterior and posterior transverse collateral sulci rostrally and caudally. The fusiform gyrus does not extend to the temporal pole. The parahippocampal gyrus (T5) together with the lingual gyrus form the medial occipito-temporal gyrus. The parahippocampal gyrus is separated from the fusiform gyrus by the collateral sulcus and can be divided into two segments (Figures 1 and 2): (a) the posterior segment is narrow and its flat surface, or subiculum, is separated from the hippocampus by the uncal sulcus, and (b) the anterior segment is also called piriform lobe and includes the anterior part of the uncus and the entorhinal area1,2.

The anatomy of the medial temporal lobe structures can be studied using high resolution MRI. Thin coronal slices, obtained perpendicular to the long axis of the hippocampus, provide the optimal images for anatomical details and determining subtle structural pathologies often associated with temporal lobe epilepsy.


Optimal technique for imaging the medial temporal lobe structure - Directed towards diagnostic evaluation of patients with intractable temporal lobe epilepsy.


The ideal sequence for MRI acquisition should be that resulting in excellent spatial resolution and contrast in a short period of time. Unfortunately, these are mutually exclusive due to limitations imposed by physical principles that are beyond the scope of this paper.

An imaging protocol for the investigation of partial epilepsies should include T1 and T2-wheitghed sequences in three, or at least two, orthogonal planes using thin slices 3,4. Contrast (gadolinium) enhancement is usually not necessary, but it is important to increase specificity in lesional epilepsies. The ideal imaging of temporal lobe structures, particularly the hippocampus, depends on image orientation and sequences optimized to display the anatomy and signal abnormality of hippocampal sclerosis and other pathologies of the temporal lobe. Nowadays three dimensional imaging (3D) is becoming part of the routine for patients with partial seizures. The advantage of 3D image is that, since the entire volume of the image has been acquired, images in any plane can be generated afterwards. In addition, the interslice interval is generally very small because it is determined by the space between the voxels.

In the following paragraphs we will discuss briefly an example of MRI protocol for patients with partial epilepsies.

First, T1-weighted sagittal images are acquired covering both hemispheres. Given the characteristics of T1 images, this sequence demonstrates the normal anatomy and allows the visualization of different types of pathology. These sagittal images are also important to ensure the optimal plane for acquisition of the coronal oblique images (Figure 2).

The second sequence consists of spin-echo T1-weighted inversion recovery (IR) coronal oblique images. These are obtained perpendicular to the long axis of the hippocampi using a sagittal image for planning the acquisition. Slice thickness should be of 3mm or less in order to avoid partial volume and improve definition of anatomical details (Figures 2 and 3). It has been advocate that this sequence may provide information not seen on conventional T1-weighted or 3D images 3, due to an excellent contrast between gray and white matter (acquisition parameters that can be used are: matrix of 256x256 or 256x128; FOV=250mm; TE=12-30; TR IR 1600-3200; TI=428-800) (Figures 4 and 5).

Figure 3. T1-weighted coronal images passing through the head and anterior portion of the body of hippocampi in a normal control subject. These images were reconstructed from a 3D acquisition obtained in the sagittal plane.

A: amygdala, AG: angular gyrus; ES: enthorinal sulcus; UC, uncal cleft; TI: tentorial indentation; LVi: temporal horn of lateral ventricule; CSi: circular sulcus of the insula. Note the symmetry and oval shape of the healthy hippocampi. Note also the alveus (white line) separating the amygdala from hippocampus.



Figure 4. T1-weighted inversion recovery (IR) coronal images (3mm thick) passing through the anterior segment of the uncus (A) and through the head and posterior portion of the body of hippocampi (B,C) in a normal control subject. These images illustrate the excellent contrast between gray and white matter and great definition of anatomical landmarks that can be obtained with IR sequences. The amygdala boundaries can be well delineated on panel A, above the ventricular horn and separated from the head of hippocampus by the uncal recess of the inferior horn of the ventricle and the alveus covering the hippocampal digitations. Note the symmetry and oval shape of the healthy hippocampi and the normal (mild) asymmetry of the temporal horns.





Figure 5. T1-weighted inversion recovery (IR) coronal and T2-FSE images (3mm thick) in a patient with right TLE showing an atrophic right hippocampus with loss of internal structure, and hyperintense T2 signal (arrow) as well as a smaller temporal lobe on the right side.





The third sequence is a T2-weighted set of images, ideally both in the coronal and axial planes Figures 4 and 5). This can be obtained using a spin-echo (SE) sequence with dual echo, i.e., T2-weeighted and proton density images, or a fast spin-echo sequence (FSE). FSE has replaced conventional SE for T2-weithed scans in many institutions. The shorter imaging times and equal or superior lesion conspicuity of the T2 FSE compared with conventional T2 SE has been established by a number of studies. These sequences are important for better definition of dysplastic and tumoral lesions, as well as for the identification of hyperintense T2 signal often present in hippocampal sclerosis 3-6 (Figure 5).

T1-weithed 3D acquisition, as discussed above, can be implemented as one of the routine acquisitions. This allows for generation of images in different planes and directions, as well as for quantitative studies, co-registration with other imaging modalities and automatic segmentation analyses. This can be acquired routinely in every patient, and in this case the scanning time for this sequence should be less than 10-12 minutes.


Hippocampal sclerosis: Magnetic resonance imaging features


Visual analysis



Images need to be optimized for evaluation of features indicating hippocampal pathology. Image orientation is crucial. Coronal slices are mandatory and they need to be obtained on a plane perpendicular to the long axis of the hippocampus guided by a sagittal scout image. The slices need to be thin to allow appreciation of fine details of the different portions of hippocampal anatomy. Ideally, the slice thickness should be 3mm or less, and never more than 5 mm. T2-weighted images are important to assess qualitatively the signal intensity, either using conventional SE or FSE sequences (see above). The technique known as FLAIR (fluid attenuation inversion recovery) can be an alternative. A recent evaluation of Fluid Attenuated Inversion Recovery (FLAIR) imaging sequences demonstrated an accuracy of 97% for determination of abnormalities associated with hippocampal sclerosis on pathology 7. To evaluate volume, shape, orientation and internal structure, high resolution T1-weighted images, particularly inversion recovery (IR), are essential.

Visual discrimination of a normal from an abnormal hippocampus is straightforward when one is clearly normal and the other is grossly abnormal, but the visual binary paradigm breaks down in the presence of symmetric bilateral disease or mild unilateral disease. In order to accurately determine the presence and severity of hippocampal atrophy in both hippocampi, absolute quantitative measurements are therefore necessary.  The presence and severity of hippocampal sclerosis in both hippocampi may provide useful prognostic information about both postoperative seizure control and memory outcome 4,8-10.

A majority of patients with hippocampal sclerosis undergoing presurgical evaluation will have a clear cut unilateral atrophic hippocampus with increased signal and a normal appearing contralateral hippocampus (Figure 5).  Several studies have shown volumetric MRI analysis of the hippocampus and amygdala to be very sensitive and specific in the identification of hippocampal sclerosis in this setting 4. However, simple qualitative visual analysis is also quite sensitive in this task, especially if the MR images are carefully and properly acquired 11. Measurements of hippocampal volume are unnecessary in this situation for clinical purposes. However, only a few studies have done a direct comparison between quantitative volumetric MRI of the hippocampus and qualitative visual assessment of the same MR images for the signs of hippocampal sclerosis. In the original work by Jack et al 12, they found volumetric MRI to be slightly more sensitive than qualitative image analysis (76 vs 71 per cent, respectively). However, another investigation using high resolution MR techniques found that volumetric MRI measurements had a sensitivity of 92 per cent compared to 56 per cent for qualitative visual inspection blinded to clinical information 13. Therefore, volumetric MRI appears to offer a significant improvement in the detection rate of hippocampal abnormalities, particularly bilateral hippocampal volume loss. Volumetric MRI is much more time-consuming and must be done correctly to be accurate and reliable. The greatest utility for volumetric MRI may be in the field of clinical research. MRI-based volumetric studies generate numerical data that permit better comparisons of the degree of atrophy of medial temporal structures in various subgroups of patients.  The findings can be statistically correlated with various clinical parameters and thereby lead to better discrimination and understanding of the underlying condition.

The determination of abnormalities of medial temporal structures increases with the experience of the examiner and knowledge of the anatomic details of this region. The following criteria are important for the visual analyses for evaluation of hippocampal sclerosis:

            1. Atrophy of the anterior temporal lobe: When the patient’s head is well aligned, the tip of the atrophic temporal lobe starts at a level posterior to the opposite side. The volume of the white matter is reduced compared to the contralateral homologous area. This finding isolated is not very sensitive or specific, since there is a lot of variation in the normal population, unless when it is pronounced, which is usually associated with medial temporal atrophy.

            2. Asymmetry of the temporal horns of the lateral ventricles: This is an indirect sign of hippocampal atrophy, often used in the practice. However, studies 4, have shown that the size of temporal horns are extremely variable, both in normals and in patients with hippocampal sclerosis, that may lead to false lateralization. If the damage occurred early in live, there will be a hypotrophy of the temporal structures without dilatation of the ventricular horn. Enlarged ventricular horn is most often found in patients who had sustained medial temporal damage after 4 years of age, particularly in the context of a meningoencephalitis, severe head trauma or hypoxia and in this case it is often bilateral.

Hippocampal atrophy: This is the most specific and the most reliable isolated finding of MTS in patients with TLE. The qualitative diagnosis of hippocampal atrophy is established (qualitatively) by comparing the hippocampal circumference on each side on all available coronal slices. Small asymmetries can be present due to normal variation or tilted position in the scanner, and should not be considered as abnormal. It is important to evaluate the shape of the hippocampus as well. A normal hippocampus is oval in shape, and in the presence of hippocampal sclerosis it becomes flattened and usually inclined (Figure 5). The diagnosis of MTS can be made in the majority of cases due to a significant unilateral or asymmetric hippocampal atrophy (Figure 5), usually associated with other findings, such as loss of internal structure and signal changes. Mild MTS or bilateral symmetrical hippocampal atrophy can be missed by visual analysis.

Hyperintense T2-weighted signal: This is usually easy to identify when the atrophy is pronounced, and there is a hypointense T1-weighted signal (Figure 5). Hyperintense T2-weigheted signal alone is not sufficient for diagnosis of MTS, although FLAIR imaging sequences seem promising for detecting abnormalities associated with mild hippocampal sclerosis 51. Studies have shown that quantitative T2 map (relaxometry) can improve the diagnosis of hippocampal sclerosis 14. Hyperintense T2 signal is caused by an increased concentration of free water in the abnormal tissue, and it has been postulated that this is due to gliosis 5. However, one study has shown that the hyperintense T2 signal in MTS is not directly correlated with glial cell density and has a different neuropathological basis than the hippocampal volume loss 6. It is also important to differentiate the intense T2 signal produced by the CSF and choroid plexus from the abnormal signal inside the hippocampus (Figure 5).

Loss of internal structure: This is usually associated with atrophy and hyperintense T2 signal. The loss of normal internal hippocampal structure is a consequence of neuronal loss and gliosis with a collapse of pyramidal cell layers (CA1,CA3, hilus) that is characteristic of hippocampal sclerosis. This abnormality is better seen on T1-weighted inversion recovery images (Figures 5); and exceptionally, it can be present without hippocampal atrophy 15.


Quantitative analysis


MRI volumetric studies


Technical Aspects

In order to maximize the precision and reproducibility of MRI-based hippocampal volume measures, the technical parameters employed when acquiring the images themselves should reflect the following guidelines 13,62:  1) Spatial resolution should be maximized.  In practical terms this means that the imaging sections (or slices) should be made as thin as possible (while preserving signal to noise) in order to avoid volume averaging artifacts in the direction of voxel anisotropy.  2) In order to optimally display hippocampal boundaries, the contrast to noise ratios between gray matter, white matter, and CSF should be high enough to permit reliable discrimination of hippocampal boundaries.  3) The image acquisition time should be short enough (less than 10 minutes) that high quality images free of motion artifact may be acquired in the vast majority of patients being screened.  The preceding criteria lead to two logical choices for the optimum type of MR imaging sequence to be employed for subsequent volume measurements.  The most commonly employed approach is a 3D volumetric pulse sequence. This results in an image dataset which is useful not only for hippocampal volume measurements but provides whole head anatomic coverage for routine diagnostic purposes.

Images acquired in the sagittal plane can be retrospectively reformatted into the coronal plane for hippocampal volume measurement, and the advantage of this approach is that the most narrow dimension of the head generally is in the sagittal direction.  This in turn permits whole head anatomic coverage with a 3D volumetric sagittal acquisition composed of 1 or 1.5 mm thick imaging voxels that are essentially isotropic (as opposed to the thicker anisotropic voxels necessary to cover the whole head if the section selection direction is in the coronal orientation).  The disadvantage of a sagittal image acquisition however is that in order to visually compare the hippocampi for the presence of relative side to side atrophy for diagnostic purposes, the images must be secondarily reformatted in the coronal plane, and thus the native or raw MR images as they are acquired are not suitable for the clinical visual evaluation of hippocampal atrophy.  Some authors have actually outlined the hippocampus for volumetric determination in the sagittal plane.  While this approach works well for most of the hippocampal borders, portions of the hippocampal border are optimally displayed only in the coronal plane, not the sagittal plane, particularly the medial subicular-parahippocampal boundary, the medial boundary between the hippocampal head and the amygdala/ambient gyrus, and the posterior border of the hippocampus.

After the image dataset has been acquired, it must be processed to produce volume measurement information.  This step requires great attention to detail in order to produce precise and accurate hippocampal and amygdaloid volume measurements.  This is generally done by transferring the MR images to a computer workstation and manually tracing hippocampal and amygdaloid borders on serial planimetric slices with a manual interactive device.  Manual tracing of hippocampal and amygdaloid borders creates a volume of interest.  The voxels inside the volume of interest are then automatically counted by the computer and multiplied by the number of mm3 per image voxel to generate hippocampal and amygdaloid volumes in mm3.  Discrepancies between the way different software programs handle the counting of border pixels in a traced volume of interest are a likely cause of the discrepancies among various sites for the “normal” absolute volume of the right and left hippocampi and amygdalae in normal subjects 16.  The second likely source for interinstitutional variability in reported “normal” hippocampal and amygdaloid volumes are the neuroanatomic boundary criteria used to define hippocampal borders 17. Rigorous standardized criteria which have a solid neuroanatomic basis must be followed when tracing the borders of the amygdala and hippocampus in order to insure precise and reproducible volume measurements. Detailed protocols for MRI volumetry may be obtained in reference 4 (for 3mm coronal images) and in reference 17 (for 3D images). 


MRI-based volume measurements of the right and left hippocampus (or amygdala) may be interpreted in two ways - relative or absolute.  To date, the relative approach, in which the right and left hippocampus in a given patient are compared with each other either by taking a right to left hippocampal ratio or the difference between the two sides, has been employed more commonly.

Evaluating hippocampal volume in absolute terms is more complex because a number of different variables affect hippocampal volume in normal individuals, such as head size, age, gender, and hemisphere. Ideally, therefore, atrophy of the right or left hippocampus (or both) in any individual should be established by comparing those values to normative percentiles in an age and gender matched control population for that hemisphere, and after adjustment for head size.  Hippocampal atrophy in any given patient as a marker of hippocampal sclerosis would then be expressed in terms of the percentile of adjusted volume in normals.  As a rule, studies in epilepsy in which hippocampal volume has been analyzed in absolute terms have not taken into account age effects, because age related effects on hippocampal volume are found primarily in the very young, due to growth and development, and in older individuals, due to age related atrophy. The effect of gender is small in comparison to that of head size.  Therefore, several studies employing absolute volumetric quantitation in epilepsy have adjusted hippocampal volume only by intracranial volume.  This adjustment can take several forms.  The two most popular are either dividing hippocampal volume by total intracranial volume to create a ratio 4 (Table 1) or a covariance approach. When using 3D acquisition, the images can be normalized to a standard space, thus reducing the variation among individuals 17



Table 1.

“Normalization” of Volumetric MRI Measurements of the Hippocampal Formation (HF) and Amygdala (AM)

·         Obtain the mean “Total Intracranial Volume (TIV)” of the normal control group

·         “Normalize” the volume of each of the structures measured (e.g., HF or AM) for individual variation in head size, using the formula:

·         “Normalized” HF (or AM) Volume  =  R  x  HF (or AM) Volume

·         Where, R  =  mean TIV of the controls
                              patient’s TIV



The investigator performing the volumetric measurements of the amygdala and hippocampus must have a detailed knowledge of the anatomy of the medial temporal region in order to obtain accurate and reliable results. In addition, the structures must be measured consistently according to a predetermined and standardized protocol.  When the boundaries of the hippocampus and amygdala are measured by a knowledgeable investigator following a predetermined and standardized protocol, the accuracy and reproducibility of the measurements are quite high. 

In addition to validating the accuracy and reproducibility of volume measurements, each center must also establish the range of normal values present in their patient and control populations.  A number of factors enter into the absolute values obtained at each institution, and therefore discrepancies between institutions are to be expected.  This requires each institution to create its own normal database.

The following protocol is adapted from that proposed by Watson et al 4


Anatomical aspects


Amygdala volume: The amygdala is an ovoid mass of gray matter situated in the superomedial portion of the temporal lobe, partly above the tip of the inferior horn of the lateral ventricle.  It occupies the superior part of the anterior segment of the uncus and partially overlies the head of the hippocampus, being separated from that structure by the uncal recess of the inferior horn of the lateral ventricle.  On the superomedial surface of the uncus, the amygdala forms a distinct protrusion, the semilunar gyrus, which corresponds to the cortical amygdaloid nucleus.  It is separated from the ambient gyrus by the semianular or amygdaloid sulcus, which forms the boundary between the amygdala and the entorhinal cortex.  The entorhinal cortex extends into the ambient gyrus and forms most of its surface.  The amygdala is separated from the substantia innominata by a deep fold, the endorhinal sulcus, which is lined on the amygdaloid side by the medial nucleus of the amygdala.  The superior rim of the ambient gyrus, lying in the fundus of the semianular sulcus, is related to the so-called corticoamygdaloid transition area which probably represents periamygdaloid cortex.  The medial surface of the ambient gyrus often shows a marked indentation, the tentorial indentation (also sometimes called the uncal notch or the intrarhinal sulcus), produced by the free edge of the tentorium cerebelli (Figure 3).

The anterior end of the amygdala is arbitrarily and consistently measured on the MRI section at the level of the closure of the lateral sulcus to form the endorhinal sulcus.  When using thin MRI slices it is possible to determine the outline of the amygdala starting at one or two slices anterior to the closure of the lateral sulcus. The medial border of the amygdala is covered by part of the entorhinal cortex which forms the surface of the ambient gyrus in this region.  The entorhinal cortex inferior to the tentorial indentation is excluded from the amygdaloid measurement.  If the tentorial indentation is poorly defined or not visible in the anterior amygdaloid region, the line of demarcation between the amygdala and the adjacent entorhinal cortex that occupies the ambient gyrus is defined by a line drawn in direct continuation with the inferior and medial border of the amygdala within the substance of the temporal lobe.  By proceeding in this manner a small amount of the superior extent of the entorhinal cortex is included in the amygdaloid volume, as is the case when the tentorial indentation is used as the landmark.  The inferior and lateral borders of the amygdala are formed by the inferior horn of the lateral ventricle or white matter. To define the superior border of the amygdala, we draw a straight line laterally from  the endorhinal sulcus to the fundus of the inferior portion of the circular sulcus of the insula, or just follow the boundaries of the amygdalar gray matter when the MRI contrast and resolution allow to do so.  More posteriorly, the optic tract is utilized as a guide to the lateral extension of the crural cistern into the transverse cerebral fissure.  This locates the medial aspect of the posterior amygdala and is used as the point of departure for defining the medial and superior borders of the structure posteriorly.

At its posterior end, the amygdala occupies the medial half of the roof of the inferior horn of the lateral ventricle, and care must be taken to exclude the tail of the caudate nucleus, the overlying globus pallidus and putamen, and the lateral geniculate body.

Hippocampal Volume:   The hippocampus is a complex structure consisting of an enlarged anterior part which has been called the pes, but perhaps is better termed the head of the hippocampus.  This portion of the hippocampus exhibits three or four digitations and turns medially to form the posterior segment of the uncus. As it turns medially the hippocampus and the dentate gyrus run in the roof of the uncal cleft (also sometimes called the uncal notch, the uncal sulcus, and, erroneously, the hippocampal sulcus 2), the sulcus-like cleft that separates the uncus above from the parahippocampal gyrus below (figure 3). Once the hippocampus and dentate gyrus reach the medial surface of the uncus, they turn up and form the posterior one-third of the medial and superomedial surface of the uncus. Macroscopically the dentate gyrus is discernible as a narrow elevation, the band or limbus of Giacomini. This is interposed between the intralimbic gyrus, which forms the posterior pole of the uncus and corresponds to sector CA3 of the hippocampus, and the uncinate gyrus, which extends anterior to the band of Giacomini and corresponds partially to sector CA1 and the subiculum. There is no macroscopically visible border between the uncinate gyrus and the ambient gyrus. The floor of the uncal cleft is formed by the presubiculum. The body of the hippocampus curves around the upper midbrain and is concave medially (Figure 1). The anatomy in this region is much less complex (Figure 4). Posteriorly, the hippocampal body tapers into the tail which turns medially just anterior to and below the splenium of the corpus callosum. The tail of the hippocampus gives rise to the fasciola cinerea which ultimately passes around the corpus callosum to continue on its upper surface as the indusium griseum. 

It is obviously most difficult to define the boundaries of the hippocampus in its most anterior portion, the hippocampal head (Figures 3 and 4). The most reliable structure separating the head of the hippocampus from the amygdala in this region is the inferior horn of the lateral ventricle. This is especially true if the ventricular cavity extends into the deep part of the uncus anterior to the head of the hippocampus, thereby forming the uncal recess of the inferior horn. However, portions of the uncal recess are often obliterated, especially medially, and the hippocampal digitations are fused to the amygdala across the ventricular cavity. When this is the case, three guidelines are used to outline the hippocampal head and separate it from the adjacent amygdala. If an obvious semilunar gyrus is present on the surface of the uncus, a line is drawn connecting the inferior horn of the lateral ventricle to the sulcus at the inferior margin of the semilunar gyrus (i.e., the semianular or amygdaloid sulcus). It is also useful to use the alveus covering the ventricular surface of the hippocampal digitations to distinguish the hippocampus from the amygdala. If neither the semianular sulcus nor the alveus is obvious, a straight horizontal line is drawn connecting the plane of the inferior horn of the lateral ventricle with the surface of the uncus. The inferior margin of the hippocampus is outlined to include the subicular complex and the uncal cleft.  The border separating the subicular complex from the parahippocampal gyrus is defined as the angle formed by the most medial extent of those two structures.  Unless significant atrophy is present, no attempt is made to outline the gray matter on the superior and inferior banks of the uncal cleft as it is usually quite narrow.  The gray matter of the entorhinal cortex or parahippocampal gyrus is excluded from this measurement.

In the hippocampal body, the delineation of the hippocampus includes the subicular complex, hippocampus proper, dentate gyrus, alveus, and fimbria. The border between the subicular complex and the parahippocampal gyrus is defined in the same manner as in the hippocampal head.  Therefore, the cortex of the parahippocampal gyrus is once again excluded from the measurement.

In the hippocampal tail, measurement again includes the subicular complex, hippocampus proper, dentate gyrus, alveus, and fimbria.  Excluded at this level are the crus of the fornix, isthmus of the cingulate gyrus, and parahippocampal gyrus. The most posterior section measured is the section with the crus of the fornix clearly separating from the hippocampus and its fimbria when using 3mm thick slices, or 2 sections posterior to that when using 1mm thick slices. 

Assuming a total anterior-posterior length of the hippocampus of approximately 40 mm, these guidelines should result in a volume measurement of 90 to 95% of the total hippocampal formation.



1. Gloor P. The temporal lobe and limbic system. New York:Oxford University Press, 1997

2. Duvernoy HM. The human brain: surface, three-dimensional sectional anatomy and MRI. Wien:Springer-Verlag, 1991

3. Kuzniecky RI, Jackson GD. Magnetic resonance in epilepsy. New York:Raven Press, 1995: 107.-183.

4. Watson C, Jack CR, Jr., Cendes F. Volumetric magnetic resonance imaging. Clinical applications and contributions to the understanding of temporal lobe epilepsy. Arch Neurol 1997;54:1521-1531.

5. Van Paesschen W, Connelly A, King MD, Jackson GD, Duncan JS. The spectrum of hippocampal sclerosis: a quantitative magnetic resonance imaging study. Ann Neurol 1997;41:41-51.

6. Van Paesschen W, Revesz T, Duncan JS, King MD, Connelly A. Quantitative neuropathology and quantitative magnetic resonance imaging of the hippocampus in temporal lobe epilepsy. Ann Neurol 1997;42:756-766.

7. Jack CR, Jr., Rydberg CH, Krecke KN, Trenerry MR, Parisi JE, et al. Mesial temporal sclerosis: diagnosis with fluid-attenuated inversion-recovery versus spin-echo MR imaging. Radiology 1996;199:367-373.

8. Jack CR, Jr., Sharbrough FW, Cascino GD, Hirschorn KA, O'Brien PC, Marsh WR. Magnetic resonance image-based hippocampal volumetry: correlation with outcome after temporal lobectomy. Ann Neurol 1992;31:138-146.

9. Cendes F. Progressive hippocampal and extrahippocampal atrophy in drug resistant epilepsy. Curr Opin Neurol 2005; 18(2):173-177.

10. Cendes F. Febrile seizures and mesial temporal sclerosis. Curr Opin Neurol 2004; 17(2):161-164.

11. Kuzniecky RI, Bilir E, Gilliam F, Faught E, Palmer C, Morawetz R et al. Multimodality MRI in mesial temporal sclerosis: relative sensitivity and specificity. Neurology 1997; 49(3):774-778.

12. Jack CR, Sharbrough FW, Twomey CK, Cascino GD, Hirschorn KA, et al. Temporal lobe seizures: lateralization with MR volume measurements of the hippocampal formation. Radiology 1990;175:423-429.

13. Cendes F, Leproux F, Melanson D, Ethier R, Evans A, et al. MRI of amygdala and hippocampus in temporal lobe epilepsy. J Comp Assist Tomogr 1993;17:206-210.

14. Coan AC, Kobayashi E, Li LM, Cendes F. Quantification of hippocampal signal intensity in patients with mesial temporal lobe epilepsy. J Neuroimaging 2003; 13(3):228-233

15. Jackson GD, Kuzniecky RI, Cascino GD. Hippocampal sclerosis without detectable hippocampal atrophy. Neurology 1994; 44:42-46.

16. Bonilha L, Kobayashi E, Cendes F, Li LM. Effects of method and MRI slice thickness on entorhinal cortex volumetry. Neuroreport 2003; 14(10):1291-1295

17. Bonilha L, Kobayashi E, Cendes F, Min LL. Protocol for volumetric segmentation of medial temporal structures using high-resolution 3-D magnetic resonance imaging. Hum Brain Mapp 2004; 22(2):145-154.