As we have noted throughout the previous section, visual analysis is an excellent method of detecting HS if optimized images are acquired and the reporting specialist is expert in the interpretation of these images. There remain issues of what is the optimal method; for example, do we need high-resolution images to examine the hippocampal internal structure? What sequences are necessary, and how much imaging time is required? We have addressed all these issues above. Now we turn to quantitative measures of hippocampal pathology.
For many reasons it is useful to have a quantitative and objective number to describe pathology. A truly objective measurement would take the subjective skill level of the person reporting the pathology out of the list of variables we have to consider. Also, we can determine the degree of abnormality as a continuous variable. This is important for many research studies designed to unravel the neurobiology of many of these conditions, and may help us to understand normal functions.
On the other hand, subtle atrophy as determined by quantitative analysis cannot be simply assumed to have the same meaning as visually assessed atrophy and signal change.
Also, just because a measurement is reported as a number does not necessarily mean that it is a totally objective measurement or that a high degree of subjective skill is not needed for the acquisition of that number. Similarly, once a number is derived, we can become totally oblivious to the assumptions and compromises that went into acquiring that number, and mistakenly confuse it with objective reality. What this means, really, is that if you use numbers to make decisions, you need to know something about how they were acquired, what they mean, and what their limitations are. This is also true of such statistically based methods as voxel-based morphology and relaxometry (discussed below). In the context of temporal lobe epilepsy these numbers (hippocampal volumes and T2 relaxation times) are a means of reaching a diagnosis, and trying to standardize information and objectify the thresholds at which significant abnormality is diagnosed. As in all medical measurements, the clinical context and overall pattern are important.
To summarize, the MR features of HS are atrophy, an increased T2 and a decreased T1 signal, and disruption of the internal structure (175). The atrophy and signal change can be quantified. Quantitative MR protocols, including hippocampal volumetric studies (207, 218, 235-240) and T2 quantification (130, 241-243), are essential in research protocols and for selected clinical cases. These two measurements are complementary and allow a full description of the spectrum of HS (130, 244). Hippocampal atrophy correlates with neuronal cell loss, mainly in the CA1 hippocampal subfield (210, 245-247). The pathologic correlate of hippocampal T2 is gliosis, which most likely correlates with hippocampal T2 (136, 247).
Since volume measurement of the hippocampus was first described, the technique has been widely used, with many modifications of technique that have become largely center specific. Despite this the principles and use have changed little since these early descriptions. With the advent of statistical approaches to analysis there have been enormous developments in advanced analysis methods of volume data, which are dealt with further in Chapter 8.
Jack et al. (235) first demonstrated in-vivo volume loss in the hippocampus in the context of epilepsy. Volume loss was soon seen as a sensitive and specific indicator of HS in the clinical context of epilepsy (112, 116, 161, 163, 206-211, 245, 248, 249). Hippocampal volume estimates permitted definition of minor volume asymmetries in these patients and also gave confidence to reporting specialists when making visual judgments. The detection of bilateral atrophy remains a problem for studies of this type, which generally compare the two sides. Absolute measures have a large inherent variability due to biological and measurement reasons. The range of normal hippocampal volumes is large, and because of this absolute volumes are not a reliable measure of pathology, particularly of mild degrees of abnormality. Test-retest and interobserver variation in absolute volumes can also be up to 7%. This is only slightly better at high field strengths (250).
Because of the relative anisotropy of the hippocampus, slices 3 mm thick or less are probably necessary for accurate estimation of hippocampal volume (207). Van Paeschen showed that a sampling strategy of one in every three slices is accurate to the degree required for the diagnosis of HS (130, 218, 247, 250-252). Volumetry is a simple, reliable method of detecting hippocampal asymmetries that can be carried out in almost all centers with little or no advanced image processing or technical expertise. In centers with excellent expertise in visual assessment of optimized images acquired for the purpose of assessing the hippocampus, the sensitivity from visual diagnosis is such that volumetric assessment usually adds little. Without this expertise, volumes are a very sensitive means of detecting hippocampal abnormality. There are some pitfalls, however, and we will discuss these now.
Because of the large range of variation in normal hippocampi, both within series and between different groups (often using slightly different techniques), absolute abnormality is difficult to define. The correction of hippocampal volume for total intracranial volume is a useful method of improving the usefulness of volume measurements. Corrections for influence of height, gender, body mass, and age have also been suggested. Females have smaller hippocampi than males and this is largely corrected by correcting for total brain size (253). Hippocampal volumes have also been shown to change with the effects of seizures that affect that temporal lobe (254, 255). Thus volumes may reflect the original pathology as well as the effects of seizures projecting to the temporal lobe. Subtle abnormalities may only reflect seizure effects and should not be used on their own to infer the site of seizure origin.
The basic finding is that the side of hippocampal volume loss as measured by side-to-side asymmetry, in the context of intractable epilepsy, correlates with both the presence of HS as found in the post resection specimen and the side of seizure onset.
An important group of patients is those in whom the findings are more difficult: those with marginal MR abnormalities that approach the limits of normal variation. Spencer and colleagues (211) investigated a population with diffi-cult-to-localize epilepsy (but not difficult MRI findings) and found, even in this group, that volume measurements of the hippocampus were 75% sensitive to and 64% specific for medial temporal seizure onset as recorded with intracranial electrodes. This was the most useful of all noninvasive localizing tests in their patient group.
The ability of a quantitative method to measure the degree of hippocampal asymmetry enables the comparison of the degree of pathology to the degree of memory impairment postoperatively (112, 210, 256) and to outcome following temporal lobe surgery (257).
For the actual technique of hippocampal volume measurement, the estimation of the hippocampal boundaries is an important issue (Fig. 4.29). Jack (258) pointed out that even the difference between including the pixels under the drawn boundary and excluding those pixels under the drawn boundary can make a difference of up to 30% in total hippocampal volume estimation. It is the surface pixels that create the greatest variation, and unfortunately the hippocampus has a large surface-to-volume ratio. Because of this, the exact boundary of the hippocampus can make a large difference in the total estimated volume, and the variation and placement of this region between observers can sometimes be great. Similarly, the test-retest measurement in the same observer can differ considerably until considerable skill is acquired in the use of the technique.
The normal variation of volume measurements of the hippocampus also depends on the subjective drawing of the boundary of the hippocampus in multiple slices. This is a problem for the comparison of hippocampal volumes across centers. Not all centers define the boundaries of the hippocampus in the same way, and this can have a considerable difference in estimating total hippocampal volume. For example, Jack (258) defined the boundaries of the hippocampus in the coronal plane to be the inplane boundaries and the anterior posterior boundaries. The areas to be included are the CA1-CA4 sectors of the hippocampus, the dentate gyrus, and the subiculum. This includes the pyramidal cell outflow tracts
(the alveus) and uses this as a high-contrast boundary of the hippocampus.
In the body and tail regions of the hippocampus, the boundaries of the hippocampus are easily determined, being defined by CSF in the temporal horn, CSF in the carotid fissure, and CSF in the uncal and ambient cisterns. Interiorly, the gray-white matter junction between the subiculum and white matter of the parahippocampal gyrus is also easy to define. The boundary between the hippocampus and the subiculum where it is continuous with that of the parahippocampal gyrus is a more subjective issue (259-264). Jack's studies used the boundary as the line from the angle formed by the most medial extent of both the subiculum cortex and the parahippocampal cortex (see Chapter 3). It is important to have an understanding of the issues involved in defining the boundaries of the hippocampus (161).
In the head of the hippocampus (pes hippocampi), the distinction of the head of the hippocampus from the overlying amygdala may be more difficult. In some cases it may be easier to make this determination than in others. For example, if the uncal recess of the temporal horn is patent this can provide a superior landmark for the head of the hippocampus. Otherwise the thin line formed by the alveus, which divides the fused hippocampus and amygdala, can also usually be distinguished and used as a boundary. In some cases, however, neither of these landmarks is easily seen, and a straight horizontal line is sometimes drawn between a mid portion of the ambient gyrus medially and the most superior medial portion of the temporal horn laterally.
FIG. 4.29. Volumetric measurements of the hippocampus. A. Manual outlining of the hippocampus, which is usually done In the coronal plane. B. Semi-automated segmentation of gray and white matter structures can be achieved with segmentation algorithms, although some manual completion in the mesial temporal regions is usually required. This image shows the results of such segmentation.
Was this article helpful?