Abstract

Introduction

Drug-resistant epilepsy (DRE) has been defined as the ‘failure of adequate trials of two tolerated, appropriately chosen and used antiseizure medications to achieve seizure freedom’. ¹ Several studies have shown that children with mesial temporal lobe and extra-temporal lobe epilepsy could have hippocampal volume reduction compared to controls.^2–8 Decreased volume of the hippocampal head, body and tail and total hippocampal volume were identified even when the hippocampus appeared normal by visual inspection on magnetic resonance imaging (MRI) in children with non-lesional focal epilepsy. ³ Hippocampal volume has not been shown to be associated with age at seizure onset or duration of epilepsy in children with focal epilepsy.^4,9 However, in adults with temporal lobe epilepsy, volume reduction in the hippocampus has been shown to correlate with the duration of epilepsy.^10–12 Most studies that have assessed hippocampal volumes in children with epilepsy have evaluated total or regional hippocampal volume; that is, hippocampal head, body and tail, and have been cross-sectional in design.^{4–6,8,11,13,14} Hence, it is uncertain if hippocampal volumes decline over time with ongoing seizures.

The aims of this study were to determine if there was a change in total or subfield hippocampal volume over time in children with focal DRE, and whether seizure control influenced total or subfield hippocampal volume.

Methods

FreeSurfer automated hippocampal subfield segmentation

Automated segmentation of the whole hippocampus and hippocampal subfield volumes including cornu ammonis (CA)1, CA2/3, CA4, subiculum, presubiculum, parasubiculum, molecular layer and dentate gyrus volumes was carried out using FreeSurfer, v6.0 (http://surfer.nmr.mgh.harvard.edu). FreeSurfer uses probabilistic information estimated from a large training set of expert measurements to assign automatically a neuroanatomical label to each voxel in the brain. ¹⁵ The technical details of the FreeSurfer whole brain segmentation procedure have been described in detail in previous publications.^16–19 In short, FreeSurfer was used to detect automatically the grey matter and white matter of all brain structures.^20,21 After that, the subroutine of the hippocampal subfield detection ²² was implemented to generate subfield volumes including CA4, CA3+CA2 (delineated in combination by the algorithm), CA1, hippocampal tail, subiculum, presubiculum, parasubiculum, molecular layer hippocampus (HP), granule cell and molecular layer of the dentate gyrus (GC-ML-DG), hippocampal amygdala transition area (HATA), fimbria and total hippocampus (Figure 1). FreeSurfer segmentations of the whole hippocampus and hippocampal subfields were visually inspected for accuracy.

Open in a separate window

Figure 1.

15-year-old boy with right hemispheric seizure, later classified as Engel III. Axial volumetric T1 weighted sequence (a), with sagittal (b) and coronal (c) reformats and (d) colour-coded overlay of hippocampal subfield volumes.

Results

Subjects

A total of 140 children with epilepsy and no epileptogenic lesion older than 5 years of age were screened for baseline and follow-up MRI brain. Of these, 103 children did not have follow-up brain MRI. Thus, 37 children (12 girls and 25 boys) were included in this study. The mean age was 10.31±3.68 years (range 3.79–17.34 years) at first MRI. Sixteen patients had left-sided epilepsy and 21 had right-sided epilepsy. Nineteen patients had frontal lobe epilepsy (nine left-sided and 10 right-sided), seven had temporal lobe epilepsy (two left-sided and five right-sided), six had parietal lobe epilepsy (four left-sided and two right-sided) and five had multilobar epilepsy (one left-sided and four right-sided). The mean age at seizure onset for all patients was 6.25±4.18 years and the mean duration of epilepsy at first MRI was 4.07±3.47 years. The median number of antiepileptic drugs at baseline for all patients was two (range one to four).

The mean interval between the two MRI scans was 2.59±1.25 years. Five patients were free of disabling seizures, five had rare disabling seizures, three had worthwhile improvement in seizures and 24 had no worthwhile improvement in seizures. Eleven children (29.7%) had a less severe seizure burden at baseline. Of these, two patients were in the group 1 category. Twenty-six patients (70.3%) had a severe seizure burden at baseline. Of these, 11 patients were in the group 1 category.

Mean age at first and second MRI and interval between the scans were not significantly different between groups 1 and 2 (all P>0.002). Table 1 summarises patient characteristics according to their group.

Table 1.

Characteristics of 37 patients.

	Group 1 (n=13, 35%)	Group 2 (n=24, 65%)	P value
Engel class I	5	N/A
Engel class II	5	N/A
Engel class III	3	N/A
Engel class IV	N/A	24
Female:male (absolute and ratio)	3:10 (25%:40%)	9:15 (75%:60%)
Mean age±SD at first MRI (years)	11.33±4.65	9.77±2.88	0.22
Mean age±SD at second MRI (years)	14.17±5.15	12.22±2.88	0.15
Mean interval±SD between the scans (years)	2.84±1.38	2.45±1.56	0.46
Right-sided:left-sided EEG findings	9:4 (43%:25%)	12:12 (57%:75%)
Mean duration of epilepsy±SD at first MRI (years)	2.74±2.62	4.79±3.65	0.08
Median number and range of antiepileptic medications	2 [1-4]	2 [1-3]
Seizure severity at baseline MRI, severe:less severe	11:2 (42%:18%)	15:9 (58%:82%)

Open in a separate window

EEG: electroencephalography; MRI: magnetic resonance imaging; SD: standard deviation.

Total and subfield hippocampal volumes: whole cohort

The baseline and follow-up total and subfield hippocampal volumes for the whole cohort are shown in Table 2. There was no significant difference in left versus right hippocampal and hippocampal subfield volumes for the whole cohort (all P>0.002). No significant differences were found in the total hippocampal and hippocampal subfield volumes for the whole cohort at follow-up compared to baseline (all P>0.002).

Table 2.

Whole cohort comparison of volumes in mean mm³ ±SD at baseline and at follow-up MRI.

	1st MRI	2nd MRI
Right
Hippocampal tail	476.50±64.69	474.96±63.96
Subiculum	418.28±46.59	421.68±49.17
CA1	634.16±82.03	632.03±87.31
Presubiculum	300.80±40.98	300.38±44.01
Parasubiculum	65.98±11.91	66.38±15.88
Molecular layer HP	559.15±62.72	558.39±63.15
GC-ML-DG	294.22±32.17	292.58±28.21
CA3	208.00±30.92	206.04±30.67
CA4	253.06±26.50	249.93±23.76
Fimbria	88.79±19.64	92.68±21.73
HATA	58.42±8.59	59.72±8.70
Whole hippocampus	3357.36±341.72	3354.76±351.79
Left
Hippocampal tail	482.63±70.73	479.14±68.38
Subiculum	424.40±47.58	427.21±49.98
CA1	613.58±88.52	616.00±85.64
Presubiculum	314.20±34.39	316.15±36.91
Parasubiculum	68.78±11.54	69.47±13.73
Molecular layer HP	548.18±66.68	551.88±65.67
GC-ML-DG	280.51±33.96	282.56±33.00
CA3	190.15±35.92	189.38±35.08
CA4	239.16±27.96	241.06±28.51
Fimbria	92.76±20.44	94.37±18.17
HATA	55.91±8.82	56.79±8.49
Whole hippocampus	3310.25±370.63	3324.02±357.92
Whole brain volume of 1st and 2nd MRI	1,184,294.38±102,951.96	1,174,004.70±102,542.59

Open in a separate window

FreeSurfer results are shown for hippocampus and hippocampal subfield segmentations in mm³ (mean ±SD).

CA: cornu ammonis; HP: hippocampus; GC-ML-DG: granule cell and molecular layer of the dentate gyrus; HATA: hippocampal amygdala transition area; MRI: magnetic resonance imaging; SD: standard deviation.

Individually, 23 patients have an increase of more than 10% in subfield hippocampal volumes, and seven patients have an increase of more than 10% in total hippocampal volumes. Twenty-five patients have a decrease of more than 10% in subfield hippocampal volumes, and two patients have a decrease of more than 10% in total hippocampal volumes. One patient had right frontal epilepsy and showed a 12.1% decrease in the size of the right hippocampus, including a 51.43% decrease of the parasubiculum, 24.76% decrease of the presubiculum and a 16.16% decrease of the fimbria. The other patient had left parietal epilepsy and showed a 11.51% decrease in the size of the left hippocampus, including a 25.2% decrease of the hippocampal tail, a 19.27% decrease of the CA3 and a 15.65% decrease of the CA1.

Seizure control versus total and subfield hippocampal volumes

The right and left total and subfield hippocampal volumes of group 1 and group 2 at baseline and follow-up are shown in Figure 2. There were no significant associations between total or subfield hippocampal volumes at follow-up and seizure control group, after adjusting for baseline volume, age, seizure severity at baseline and time interval between the MRI scans (all P>0.002). Regression coefficients and P values are summarised in Supplementary Table 1. From group 1, nine patients demonstrated a hippocampal subfield volume decrease greater than 10% including the fimbria (five patients), hippocampal tail (two patients), parasubiculum (two patients), HATA (one patient), CA1 (one patient), presubiculum (one patient) and CA3 (one patient). From group 2, 15 patients demonstrated a hippocampus or a hippocampal subfield volume decrease greater than 10% including parasubiculum (eight patients), fimbria (five patients), CA3 (four patients), HATA (three patients), presubiculum (three patients), hippocampal tail (three patients), subiculum (three patients), CA1 (two patients), CA4 (two patients), whole hippocampus (two patients), GC-ML-DG (one patient) and molecular layer HP (one patient) (see Supplementary Table 2).

Open in a separate window

Figure 2.

(a) Right and (b) left total and subfield hippocampal volumes group 1 (Engel I–III seizure outcome), and (c) right and (d) left total and subfield hippocampal volumes in group 2 (Engel IV seizure outcome).

Discussion

We did not find a difference in total or subfield hippocampal volume at follow-up compared to baseline for the whole cohort. Further, we found that there were no associations between total or subfield hippocampal volumes at follow-up and with seizure control (group 1 vs. group 2), after adjusting for baseline volumes, age and time interval between the MRI scans.

Polli et al. ²⁶ evaluated the hippocampal volumes in rats treated with kainic acid or pilocarpine to induce status epilepticus, and their hippocampal volumes were measured on MRI at 3, 6 and 9 months. Both groups had a time-specific reduction in MRI hippocampal volume compared to controls, but the hippocampal volume reduction did not progress over time. The authors concluded that hippocampal atrophy once detected remains stable, despite continuous seizures. ²⁶ Similar findings were reported in an adult study with 179 patients. ²⁷ Thirty-three of the 179 patients were diagnosed with cryptogenic temporal lobe epilepsy. No significant correlation was observed between the frequency of convulsive seizures, partial seizures, epilepsy duration, antiepileptic drug use and change in ipsilateral hippocampal volume. ²⁷ Our findings were in line with the literature, in that we did not find a significant volume loss of total or subfield hippocampal volumes at follow-up compared to baseline, nor volume loss based on seizure control over time in children with focal DRE.

Previous studies have evaluated hippocampal total or subregion volumes such as hippocampal head, body and tail volume, and demonstrated reduced hippocampal volumes in children with focal epilepsy.^3–8,28,29 Reduced total and regional hippocampal volumes were identified in those with non-lesional epilepsy and have normal appearing hippocampus, both ipsilateral and contralateral to seizure focus. ³ Those studies were cross-sectional in design. To our knowledge, there has been no longitudinal study that has evaluated total or subfield hippocampal volumes in children with focal epilepsy. It is possible that hippocampal volume reduction may be time specific, occurring around the time of seizure onset, but does not progress with time, similar to that observed in an animal model. ²⁶ Alternatively, there may be microstructural changes in the hippocampal subfields that may not be detected using volumetric measures. A prior study has also shown that there was no significant volume loss longitudinally in total, cerebral or hemispheric grey or white matter in children with focal epilepsy, ³⁰ suggesting that volume loss does not occur despite ongoing seizures.

A prior systematic review and meta-analysis has shown that neuronal loss was observed in hippocampal subfields CA1 to CA4 in patients with hippocampal sclerosis. ³¹ In vivo MRI of hippocampal body subfield identified abnormality in subfield volume in patients with hippocampal sclerosis, which corresponded with International League Against Epilepsy (ILAE) hippocampal sclerosis subtypes based on histology. ³² Voets et al. ³³ evaluated patients with MRI-negative temporal lobe epilepsy and found hippocampal subfield atrophy in nine out of 12 (75%) patients, commonly affecting CA3. Studies that have assessed hippocampal subfields have done so in adults with temporal lobe epilepsy, and were cross-sectional studies comparing hippocampal subfield volumes on MRI to histological assessment. Our study has evaluated hippocampal subfield volumes longitudinally in children. Although we did not find a difference in total or subfield volumes at follow-up relative to baseline for the whole cohort, 25 patients demonstrated a reduction in total and/or subfield volume by more than 10%. It is possible that there are individuals who may be more susceptible to injury related to ongoing seizures than others, and average group differences may not be sufficiently sensitive to identify subtle changes in hippocampal volumes.

Our study has limitations. We used FreeSurfer’s automated segmentation of the hippocampal subfields. Manual segmentation of the hippocampus was reported to be more sensitive in detecting hippocampal atrophy than automated segmentation,^13,17,34 but it is time consuming and prone to operator error. Automated segmentation of the hippocampal subfields using FreeSurfer has the potential to provide rapid and accurate assessment of the hippocampal volume, and has been shown to be sensitive at detecting hippocampal atrophy.^13,17,35,36 Whelan et al. showed that FreeSurfer version 6 is a reliable tool for segmenting hippocampal subfields, with intraclass correlation of 0.90 or greater. ³⁷ Herten et al. recommended utilising both T1 and T2-weighted images for FreeSurfer’s automated segmentation of the hippocampus to increase accuracy. ³⁸ However, volumetric T2-weighted images were not available for this study. We have included children with a variety of epilepsy types. It is uncertain whether those with temporal lobe epilepsy would be more likely to demonstrate hippocampal volume loss over time relative to extra-temporal epilepsy. Bernasconi et al. showed that in adults with temporal lobe epilepsy, hippocampal, amygdala and entorhinal atrophy was associated with the duration of epilepsy. ³⁹ In addition, it is possible that with an increase in the time interval between baseline and follow-up, total and subfield hippocampal volumes could demonstrate significant volumetric changes. Other limitations are the moderate number of patients at various ages, the single-centre nature of this study and the relative lack of controls with longitudinal MRI. Recruiting and retaining healthy children as controls in longitudinal research is challenging. In a previous cross-sectional study, we have compared the hippocampal volumes in children with non-lesional focal epilepsy to healthy controls and found smaller hippocampal subregion volumes in patients relative to controls. ³

In conclusion, our study did not show a difference in total or subfield hippocampal volumes at follow-up compared to baseline, although 25 patients showed a 10% reduction in total or subfield hippocampal volumes at follow-up compared to baseline. We also found that seizure control did not predict total or subfield hippocampal volumes at follow-up. Future study is required to validate our findings.

Supplemental Material

Footnotes

References