High-resolution images through the hippocampi can detect the primary findings that establish a diagnosis of MTS as well as secondary abnormalities (Table 74.1). Images through the rest of the brain allow detection of dual pathology. Although imaging sequences may vary slightly from center to center, essential components are thin-section high-matrix T2 fast spin echo (FSE), thin-section fluid-attenuated inversion recovery (FLAIR) images angled perpendicular to the long axis of the hippocampus, and volumetric T1-weighted images with a partition size not exceeding 2 mm through the entire brain; a gradientecho T2 sequence is a high priority in adults. Symmetric alignment of the head improves the accuracy of visual assessments of asymmetric hippocampal volume. Some centers obtain short tau inversion recovery (STIR) sequences because of the excellent gray-white contrast and ability to assess the hippocampal signal. Our volumetric T1-weighted sequence provides higher resolution of the gray-white junction, and T2 FSE and FLAIR images evaluate the hippocampal signal. Spin-echo T2-weighted sequences are probably more sensitive to subtle increases in T2 signal but the long imaging times often necessitate cardiac gating to obtain adequate images.

The assessment usually begins with a close inspection of the hippocampi on oblique T2 FSE images, preferably with
magnification. Increased T2 signal, which reportedly has a sensitivity of 93% and a specificity of 74% (28), can be detected in specific sectors, along with blurring of the internal architecture (Fig. 74.1). In more severe cases, decreased undulations in the hippocampal head can often be identified (29). T2-signal increases in the collateral white matter and blurring of the gray-white junction are also visible on this sequence. Table 74.2 summarizes the radiographic findings of MTS (5,9,29,30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44).

Close inspection of the FLAIR images confirms T2 signal asymmetry. On FLAIR images, the hippocampus normally is brighter than neocortex, and the choroid plexus just superior to the hippocampi is bright. A comparison with T2 FSE images can avoid misdiagnosing bilateral disease or mistaking choroid plexus for hippocampal hyperintensity. Scrutiny of the coronal volumetric T1 images allows comparison of hippocampal and temporal lobe volumes. Secondary MTS findings such as volume loss in the circuit of Papez are best detected on this sequence, as are temporal lobe polymicrogyria and heterotopia. A gradient-echo T2 sequence may also detect dual pathology such as subtle cavernous venous, or cryptic venous, malformations or old hemorrhagic shear injury that would otherwise be missed. A normal or small T2-bright hippocampus makes the diagnosis in the appropriate clinical setting. In asymptomatic individuals, an increased T2 signal in the hippocampus (45,46) and a small hippocampus without signal abnormality (45) have uncertain significance. Increased hippocampal volume associated with increased T2 signal may suggest acute swelling because of recent severe seizure activity (47,48), infiltrating glioma, or cortical dysplasia. Although such acute swelling can progress to MTS, the differential diagnosis of an enlarged hippocampus does not include MTS (48,49). Normal variants not associated with seizures are hippocampal sulcal remnants and choroidal fissure cysts (29,50,51); both are equal to cerebrospinal fluid in signal
intensity on all sequences. Occurring in about 10% of the population, hippocampal sulcal remnants are 1- to 2-mm structures between the dentate gyrus and the cornu ammonis that represent a failure of normal embryogenic involution of the hippocampal sulcus. Choroidal fissure cysts, located just above the hippocampus, become more frequent with age (29,51).

MRS noninvasively measures the integrity and function of neuronal tissue; ³¹P-MRS evaluates cerebral energetics and pH. Studies at 1.5 T and 4.1 T have shown reductions of 50% in the ratio of phosphocreatine/inorganic phosphate (PCr/Pi) in the affected lobe of patients with temporal lobe epilepsy and of 24% in the contralateral temporal lobe (52, 53, 54), thought to indicate persistent impairment of energy metabolism. Despite these intriguing changes, a low signal-to-noise ratio, poor spatial resolution, and widespread unavailability for routine clinical studies limit the clinical application of ³¹P-MRS. ¹H-MRS, in contrast, can be performed on most clinical 1.5-T scanners with spatial resolutions between approximately 1 and 8 cc, depending on whether multivoxel or single-voxel techniques are used. ¹H-MRS provides biochemical information about neuronal function (N-acetyl aspartate [NAA]), membrane turnover (choline), and total energy stores (creatine [Cr]) as well as the presence of cerebral lactate. Its sensitivity and specificity in temporal lobe epilepsy ranges from 60% (55) to 97% (56) and is greater with higher-spatial-resolution techniques (57). High-quality routine MRI of the hippocampus and ¹H-MRS for localization have not been directly compared but are probably complementary. Because absolute quantification of NAA, choline, and Cr is difficult, the sum of creatine and phosphocreatinine is assumed to be constant in adults and ratios of NAA and choline to Cr are usually calculated. Ipsilateral decreases in NAA/Cr occur in about 90% of MTS cases, and the side of functional abnormality corresponds highly with the side of seizure onset on EEG (58). The loss of NAA and the decreased NAA/Cr ratio are associated with dysfunctional mitochondrial metabolism and neuronal loss (38,39,59, 60, 61) but reflect neuronal as well as glial function (62). Combined with MRI findings, MRS improves predictions of surgical outcomes (63) and may help with lateralization in MRI-negative patients with temporal lobe epilepsy (64). With voxels typically including at most only part of the hippocampus, these MRS findings probably characterize the extrahippocampal, temporal lobe metabolic abnormalities in patients with mesial temporal lobe epilepsy. In fact, although NAA/Cr levels and ipsilateral hippocampal volumes can be statistically related, they share only a small percentage of variance and may be related but distinct (65,66). Temporal lobe MRS findings may also indicate a more widespread underlying pathologic condition (67). NAA/Cr decreases in the contralateral hemisphere may predict a poor surgical outcome (68). Lactate levels are normally absent in the brain or present in very low levels. The lactate resonance consists of two distinct peaks at 1.3 ppm and indicates a disturbance of the cellular oxidative mechanism. Lactate has been identified in the region of seizure activity during status epilepticus and epilepsia partialis continua and in the hippocampus within 24 hours after temporal lobe seizures (69, 70, 71, 72). Lactate may also alter the excitability of local neurons (73).

Because acute and transient changes have also been reported in MRS (71,74), results should be interpreted in conjunction with recent seizure activity. The 1H-MRS signal also may be influenced by antiepileptic medication and even by metabolic interventions such as a ketogenic diet (57,75,76).

Hippocampal volumetry and T2-relaxometry show increased sensitivity over visual analysis in the detection of hippocampal sclerosis ranging from about 80%-90% to 90%-95% (29,38,39,60,77, 78, 79). However, T2-relaxometry requires an additional imaging sequence that lengthens an already protracted protocol, and both techniques require manual segmentation of the hippocampus and post-processing to calculate T2 values and volumes. Although the added sensitivity has not yet prompted their incorporation into clinical protocols for mesial temporal lobe epilepsy, these techniques have enhanced our understanding of MTS. Hippocampal volumetry has disclosed a spectrum of volume loss in MTS (80,81), and total loss of volume has been correlated with reduced neuronal density (82, 83, 84). Longitudinal volumetric MRI studies have shown progressive hippocampal volume loss in patients with intractable mesial temporal lobe seizures, whereas patients with new-onset well-controlled temporal lobe epilepsy exhibited no change over the same period (85). No volume loss had been identified in extratemporal epilepsy, emphasizing the specificity of volumetric studies. Automated methods for calculating hippocampal volumes are being developed (86,87) and, if validated, could be used routinely.

T2-relaxometry measures the decay in signal intensity at different echo times in a series of T2-weighted images acquired at the same slice. Techniques range from the use of a 16-echo sequence (88,89) or two echoes (90, 91, 92) and time-efficient sequences (79,91); all reliably measured the T2 signal in the hippocampus and the amygdala (93). Boundaries with cerebrospinal fluid must be carefully avoided in this operator-dependent technique. During measurement of the T2-relaxation time along the long axis of the hippocampus in healthy individuals, a study described a characteristic profile that is disturbed in patients with hippocampal sclerosis (94). The hippocampal relaxation time correlates well with the glial and neuronal ratio, particularly in the CA1 subregion (80,84).

Essential components are 3-mm or less high-matrix T2 FSE images obtained through the lobe of highest suspicion, coronal volumetric T1-weighted images no larger than 2 mm through the entire brain, and a gradient-echo T2 sequence (Table 74.1). Contrast should be used in all adults and in children with obvious lesions, but if time is an issue, the high-resolution sequences take precedence. A patient can always return for a contrast study if a small lesion is identified; however, if omitting the thin sections misses a small nonenhancing lesion, the disease may be labeled nonlesional. Positioning is important; symmetric alignment of the head enhances the visual assessments of gyral asymmetries. STIR sequences may be obtained, but our center uses a volumetric T1-weighted sequence to evaluate the gray-white junction. Spin-echo T2-weighted sequences can detect a subtle increased signal in the cortex or white matter associated with cortical dysplasias, but time constraints preclude their routine use.