Neuropsychological correlations with fornix and hippocampus in temporal lobe epilepsy demonstrated with diffusion tensor imaging

Abstract

Objective

Memory dysfunction is a common disabling comorbidity in temporal lobe epilepsy (TLE) and hippocampal sclerosis (HS). While bilateral fornix diffusion abnormalities have been reported in TLE using diffusion tensor imaging (DTI), the impact of fornix pathology on cognitive outcomes remains unclear. This study investigated the relationships between fornix DTI, hippocampal volume, and cognitive outcomes in TLE.

Methods

Thirty‐five TLE patients (21 unilateral HS, 11 non‐HS, 3 bilateral HS) and 22 controls underwent fornix FLAIR‐DTI (1.2 × 1.2 × 2 mm³ resolution) and hippocampal DTI (1 × 1 × 1 mm³) scans. Deterministic tractography was used to extract left and right fornix diffusion metrics including fractional anisotropy (FA). Whole hippocampi were manually segmented to yield volume and mean diffusivity (MD). In a subset of 23 TLE patients, linear correlations were assessed for verbal memory, non‐verbal memory, and processing speed index (PSI) versus imaging measurements.

Results

In the HS group, a significant linear correlation was found between both hippocampal volume (r = 0.58, p = 0.001) and MD (r = −0.42, p = 0.03) with fornix FA. Left hippocampal volume correlated with verbal memory (r = 0.74, p = 0.0001) but not with PSI (p > 0.05). Left fornix FA also correlated with verbal memory (r = 0.48, p = 0.025). In contrast, right hippocampal volume and right fornix FA showed no correlation with non‐verbal memory (p > 0.05). FA of both left and right fornix correlated with PSI (left: r = 0.45, p = 0.038; right: r = 0.44, p = 0.037).

Significance

The findings of correlations between the hippocampus and memory, but not processing speed, and fornix with both memory and processing speed in TLE suggest a specialized role of the hippocampus in memory and a broader role for the fornix in cognitive function, beyond its association with memory.

Plain Language Summary

The hippocampus is a brain structure involved in learning and memory. Patients with temporal lobe epilepsy (TLE), especially those with damage to the hippocampus, often experience memory and thinking problems. This study used advanced brain magnetic resonance imaging (MRI) to explore the role of a nearby structure called the fornix, which connects the hippocampus to other brain areas. We found that damage to both the hippocampus and fornix was linked to poor verbal memory, while damage to the fornix also affected thinking speed. These findings suggest that the fornix plays a broader role in brain function than previously thought.

Key points.

• Left hippocampal volume correlated with verbal memory but not processing speed, supporting its specialized role in memory.

• Left fornix FA correlated with both verbal memory and processing speed, indicating a broader cognitive role in TLE beyond memory.

• Right hippocampal volume and fornix FA showed no correlation to non‐verbal memory, consistent with diffuse right hemisphere memory networks.

1. INTRODUCTION

Memory dysfunction is a common disabling comorbidity of temporal lobe epilepsy (TLE) with hippocampal sclerosis (HS). Deficits in broader cognitive domains such as processing speed have been recognized in TLE suggesting a more widespread network dysfunction beyond the mesial temporal structures.^{1, 2} The severity of memory impairment in TLE has been shown to correlate with hippocampal atrophy on magnetic resonance imaging (MRI) ³ and neuronal loss on histological samples of the hippocampus.^{4, 5} Correlations between memory and several white matter structures in TLE have also been demonstrated with diffusion tensor imaging (DTI). However, surprisingly, correlations with the fornix, the main efferent output pathway from the hippocampus, were not observed. ⁶ Furthermore, the structural basis of impairment in other cognitive functions in TLE remains poorly understood. ¹

DTI has revealed white matter abnormalities extending beyond the temporal lobe and the hemisphere ipsilateral to the seizure focus.^{7, 8} Although bilateral fornix diffusion abnormalities in TLE patients with HS have been described,^{9, 10} the impact of fornix pathology on cognitive outcomes in TLE patients remains unclear. Previous DTI tractography studies of the fornix were hindered by lower resolution protocols (voxel volume >8 mm³). High‐resolution (1.2 × 1.2 × 2 = 2.9 mm³) along with fluid‐attenuated inversion recovery (FLAIR)‐DTI sequence has been shown to improve the measurement of diffusion metrics of the fornix microstructure. ¹¹ The purpose here was to assess the correlation between fornix DTI and hippocampus volume and their links to memory and processing speed.

2. METHODS

2.1. Participants demographics

This study recruited 22 healthy controls (mean age: 43 years; range: 18–67 years; 6 males) and 35 patients with TLE (42 years; 18–70 years; 17 males). In this cohort, 19 controls and 23 TLE patients were recruited from our previous studies.^{12, 13} Control subjects were screened for neurological diseases or contraindications to MRI. TLE patients were referred by their neurologist at the University of Alberta Epilepsy Clinic, based on ictal semiology, EEG video telemetry, and MRI findings consistent with a diagnosis of TLE. Patients showing evidence of an extratemporal seizure focus on clinical semiology, MRI, or ictal/interictal EEG were excluded from the study. This study was approved by the Health Research Ethics Board at the University of Alberta.

Patients with TLE were categorized into unilateral HS (n = 21), non‐HS (n = 11), and bilateral HS (n = 3) groups based on qualitative assessment of clinical MRI by a radiologist with expertise in epilepsy, using high‐resolution 3D T1‐weighted, coronal T2, and coronal FLAIR sequences with high in‐plane resolution (Table 1). For unilateral HS patients, EEG video telemetry lateralization was concordant with MRI findings. Non‐HS patients showed no evidence of hippocampal sclerosis or other structural abnormalities, except for one patient with a low‐grade ganglioglioma (confirmed histopathologically) and normal hippocampi on MRI without histological evidence of hippocampal pathology. Past medical history was notable for febrile seizures in eight TLE patients (7 unilateral HS, 1 bilateral HS). Four patients had a history of meningitis (3 unilateral HS, 1 bilateral HS), and one bilateral HS patient had a history of encephalitis.

TABLE 1.

Demographic and clinical characteristics of controls and patients with temporal lobe epilepsy (TLE) in this study.

TLE subtype	Control (n = 22)	TLE (n = 35)
	n/a	Non‐HS (n = 11)	Unilateral HS (n = 21)	Bilateral HS (n = 3)
Age (years, range)	43 (18–67)	38 (18–53)	43 (20–70)	49 (36–60)
Sex	M:6, F:16	M:7, F:4	M:9, F:12	M:1, F:2
Handedness	R:21, L:1	R:9, L:2	R:18, L:4	R:3
Telemetry Laterization	n/a	R:4, L:4, Bil:2, None:1	R:10, L:11	R:1, Bil:1, None:1
Age of seizure onset (years, range)	n/a	24.1 (5–51)	22.0 (7–53)	13.6 (0.7–35)
Disease Duration (years, range)	n/a	15.3 (1–40)	21.1 (2–55)	31.8 (19–54)

Note: No significant differences were observed in age or disease duration between groups (p > 0.05). Given the small number of patients in the bilateral hippocampal sclerosis (HS) group, comparisons were limited to the control, non‐HS, and unilateral HS subgroups. Telemetry results indicate seizure lateralization by video‐EEG monitoring: R = right, L = left, Bil = bilateral, None = non‐lateralized.

A subset of 23 TLE patients (15 unilateral HS: one left‐handed, 8 left‐lateralized and 7 right‐lateralized; 6 non‐HS: one left‐handed, 3 left‐lateralized, 2 right‐lateralized, and 1 non‐lateralized; and 2 bilateral HS: both right‐handed, 1 bilateral‐lateralized and 1 right‐lateralized) who were undergoing presurgical evaluation for epilepsy surgery underwent neuropsychological testing, including the Rey Auditory Verbal Learning Test delayed recall (RAVLT), Rey Complex Figure Test and Recognition Trial delayed recall (RCFT) for non‐verbal memory, and Wechsler Adult Intelligence Scale (WAIS‐IV) Processing Speed Index (PSI).

Patients with HS had an average education of 12.0 ± 3.5 years, RAVLT T‐score 32.6 ± 12.9, RCFT T‐score 27.2 ± 14.7, and PSI score 87.8 ± 17.9. Non‐HS patients had an average education of 11.5 ± 5.9 years, RAVLT 42.7 ± 13.7, RCFT 29.7 ± 11.6, and PSI 92.2 ± 10.6. T‐scores <40 and PSI <85 were considered below average based on normative data.

2.2. Image acquisition

All brain scans were acquired on a Siemens Prisma 3 T. To allow full coverage of the fornix, FLAIR‐DTI images were acquired over 35 slices, 5 b = 0 s/mm², 20 diffusion directions with b = 1000 s/mm², 1.2 × 1.2 × 2 mm³ interpolated to 0.64 × 0.64 mm² in‐plane resolution, TR = 9000 ms, TE = 69 ms, TI = 2300 ms in 4:11 min. The hippocampus DTI scan was acquired with single‐shot EPI, 20 slices at 1 × 1 × 1 mm³ resolution, b = 500 s/mm², 10 averages of 10 gradient directions and 10 b = 0 s/mm² in 5:18 min as per Treit et al. ¹⁴ Hippocampal DTI scans were not available for one control and two unilateral HS patients.

All DTI images were denoised with the non‐local and spatial angular matching (NLSAM) algorithm^{15, 16} and tensor parameters were estimated after correcting for Gibbs‐ringing, eddy current, and motion in MRtrix3.

2.3. Fornix and hippocampus segmentation

Deterministic tractography in MRtrix3 ¹⁷ used a seed point resolution and step size matching the voxel size, FA threshold of 0.15, turning angle of 30°, and minimum fiber length of 10 mm. An ‘AND’ ROI was placed in the fornix body (midway between crus and anterior commissure) to include tracts through pre‐ and post‐commissural columns. ‘NOT’ ROIs excluded anterior commissure and other callosal or thalamic fibers. In HS patients, tractography was interrupted at the fornix crus curvature because of low FA around this region. To overcome this issue, tractography was repeated with two additional ‘AND’ ROIs around each hippocampus and ‘NOT’ ROIs as needed. The full fornix tract was reconstructed for all subjects, including those with HS. To conduct lateralized analysis, left and right fornix tracts were separated using the tckedit function with a midline ROI at the anterior and posterior junctions where the fornix body splits into the pre/postcommissural columns and bilateral crura, respectively. The volume, mean diffusivity (MD), fractional anisotropy (FA), axial diffusivity (AD), and radial diffusivity (RD) were then extracted separately for the left and right fornix.

Whole hippocampi were manually segmented on mean‐DWIs using ITK‐snap v4.2.2 as previously described. ¹² The volume and MD were obtained separately for each hippocampus.

2.4. Statistical analysis

Statistical tests in SPSS v29 (SPSS Corp, 2021) used analysis of covariance (ANCOVA) to assess group effects on continuous variables (fornix volume, FA, MD, AD, RD, and hippocampal volume, MD), controlling for age. The groups included control (bilateral fornix/hippocampus of controls), non‐HS (bilateral fornix/hippocampus of non‐HS and contralateral fornix/hippocampus of unilateral HS patients), and HS (ipsilateral fornix/hippocampus of unilateral HS and bilateral fornix/hippocampus of bilateral HS patients). For ANCOVAs with significant main effects, pairwise post hoc comparisons were conducted with Sidak adjustment for multiple comparisons. Adjusted p < 0.05 was considered significant.

Fornix FA and hippocampus volume/MD were selected for partial correlations with each other in the control, non‐HS, and HS groups based on their significant group effects in the above ANCOVA.

Memory deficits in TLE patients correspond to lateralization of HS (left or dominant hemisphere HS associated with verbal memory deficit and right or nondominant HS associated with figural memory deficits).^{1, 18} Hypothesis‐driven correlations between TLE neuropsychological scores and fornix/hippocampus measures were tested in all 23 TLE patients combined including: left fornix FA and left hippocampal volume versus RAVLT, right fornix FA and right hippocampal volume versus RCFT, and bilateral fornix FA and hippocampal volumes versus PSI. Partial correlations controlling for age and education years were used and p < 0.05 was set as significant. Of the 23 patients with neuropsychological data, one did not undergo hippocampus imaging and was excluded from hippocampal–neuropsychological correlations but was included in fornix–neuropsychological analyses.

3. RESULTS

3.1. Fornix and hippocampus DTI in the groups

Tractography of the fornix demonstrated regional elevations of MD (>1.4 × 10⁻³ mm²/s) and reductions of FA (<0.30) in the fornix ipsilateral to HS (body and crus) of HS patients (Figure 1C,D) compared to controls and non‐HS subjects (Figure 1A,B). The sclerotic hippocampus of the same HS TLE patients (Figure 1C,D) demonstrated widespread elevated MD (>1.1 × 10⁻³ mm²/s) which corresponded to higher MD and lower FA in the fornix.

FIGURE 1.

Regional hippocampus MD maps (row 1) and fornix DTI tractography with MD (row 2) and FA (row 3) color coded in a representative (A) control (51 years old), (B) non‐HS (32 years old), (C) unilateral left HS (43 years old), and (D) bilateral HS (54 years old) subjects. The non‐HS hippocampi and fornix are comparable to the control. The ipsilateral hippocampus of unilateral left HS (indicated by *) and sclerotic hippocampi of bilateral HS demonstrate widespread elevated MD (above 1.1 × 10⁻³ mm²/s) which corresponds to higher MD (above 1.4 × 10⁻³ mm²/s) and lower FA (below 0.30) in the body and crus of the fornix.

There was a significant effect of group for hippocampus volume (F = 36.7, p < 0.001) and MD (F = 63.1, p < 0.001) and for fornix volume (F = 26.7, p < 0.001), FA (F = 17.1, p < 0.001) and MD (F = 17.9, p < 0.001), AD (F = 12.2, p < 0.001), and RD (F = 20.9, p < 0.001) (Table 2 provides post hoc comparisons).

TABLE 2.

Hippocampus and fornix volume and DTI measures represent the average of left and right hippocampi and fornix in controls (n = 22 × 2 = 44).

	Control	TLE non‐HS	TLE HS
Whole hippocampus MRI
Volume (cm3)	2.53 ± 0.38	2.55 ± 0.43	1.59 ± 0.63*
MD (×10−3 mm2/s)	0.79 ± 0.03	0.79 ± 0.03	0.91 ± 0.07*
Fornix
Volume (cm3)	3.30 ± 0.62	2.91 ± 0.69	2.14 ± 0.62*
FA	0.35 ± 0.02	0.34 ± 0.02	0.28 ± 0.02*
MD (×10−3 mm2/s)	1.06 ± 0.07	1.09 ± 0.08	1.12 ± 0.12*
AD (×10−3 mm2/s)	1.47 ± 0.08	1.49 ± 0.09	1.59 ± 0.13*
RD (×10−3 mm2/s)	0.86 ± 0.06	0.88 ± 0.07	0.98 ± 0.09*

Note: The non‐HS group includes bilateral hippocampi and fornix from non‐HS patients and contralateral hippocampi and fornix from unilateral HS patients (n = 11 × 2 + 21 = 43). The HS group includes ipsilateral hippocampi and fornix from unilateral HS patients and bilateral hippocampi and fornix from bilateral HS patients (n = 21 + 3 × 2 = 27).

^*Significant difference (p < 0.05) from controls based on pairwise comparisons from ANCOVA controlling for age.

3.2. Fornix versus hippocampus correlations

There was a significant linear correlation between hippocampal volume and fornix FA (r = 0.58, p = 0.001) in the HS group whereas there were none in the controls (r = 0.26, p = 0.089) and non‐HS groups (r = 0.067, p = 0.67) (Figure 2A–C). Similarly, hippocampal MD showed a linear correlation with fornix FA in the HS group (r = −0.42, p = 0.03) and not in the controls (r = −0.16, p = 0.29) and non‐HS groups (r = −0.011, p = 0.94) (data not shown).

FIGURE 2.

Linear correlations between hippocampus volume and fornix fractional anisotropy (FA) in (A) controls—non‐significant, (B) non‐HS—non‐significant, and (C) HS—significant. (D, F) The Rey auditory verbal learning test delayed recall (RAVLT) correlated with both left hippocampus volume and left fornix FA in the entire TLE cohort. (E, G) In contrast, the left fornix FA but not left hippocampus volume showed linear correlations with PSI.

3.3. Hippocampal and fornix correlations with memory and processing speed

Left hippocampal volume linearly correlated with RAVLT (r = 0.74, p = 0.0001) but not with PSI (r = 0.33, p = 0.14) in TLE patients (Figure 2D,E). Right hippocampus showed no correlation with RCFT (r = 0.081, p = 0.72) and PSI (r = 0.31, p = 0.17). Left fornix FA showed a significant correlation with RAVLT (r = 0.48, p = 0.025) (Figure 2F) while right fornix FA showed no correlation with RCFT (r = 0.05, p = 0.83). PSI was significantly correlated with both the left (r = 0.45, p = 0.038, Figure 2G) and right fornix FA (r = 0.44, p = 0.037).

4. DISCUSSION

Since the demonstration of anterograde amnesia from bilateral hippocampal damage, ¹⁹ many studies have shown that the memory circuitry includes not only the hippocampus but also other temporal and extra‐temporal regions, such as neocortical areas, reflecting the distributed nature of memory. ²⁰ This study showed the expected correlation between left hippocampus volume and verbal memory but not with PSI in TLE, supporting a specialized role of the hippocampus in memory encoding rather than broader cognitive domains.

Our findings show no correlation between right hippocampal volume or fornix FA with non‐verbal memory, consistent with previous DTI studies in TLE.^{6, 21, 22, 23} This is expected as non‐verbal memory tasks are often less sensitive to lateralized dysfunction than verbal memory assessments. ²⁴ Nonverbal memory may not localize as strongly to the right hemisphere ²⁵ and the right hemisphere memory network appears diffuse, relying on the hippocampus and extending to posterior cortical and subcortical structures, making it harder to detect reliable correlations involving right‐sided structures. ²⁶

Given the known primary circuitry of the limbic system, this study used high‐resolution FLAIR‐DTI to demonstrate microstructural correlations between the fornix and hippocampus in TLE. The observed correlation between left fornix FA and verbal memory supports its role as the principal efferent pathway of the hippocampus, where downstream degeneration is expected in hippocampal injury in TLE. While these results align with two previous studies,^{21, 22} other studies have shown no correlations between fornix DTI measures and memory.^{6, 23} Differences in DTI acquisition methods, particularly resolution, lack of partial volume correction, tractography approaches, and subject variability may account for these discrepancies.

This study demonstrated a significant correlation between bilateral fornix and PSI in TLE, consistent with previous studies.^{22, 23} Processing speed is a marker of general cognitive efficiency, with psychomotor slowing frequently found in TLE patients.^{27, 28} Considering the link between PSI and white matter connectivity, ²⁷ fornix correlation with PSI supports its role in connecting various nodes of the limbic circuitry with both limbic and frontal regions involved in information processing, including the prefrontal cortex, anterior cingulate gyrus, and potentially more complex downstream cognitive networks.

The fornix is classically considered the primary efferent pathway from the hippocampus to the mammillary bodies via postcommissural fibers and part of the memory network. ²⁹ However, the fornix also contains precommissural fibers carrying efferents from the hippocampus to the septal area and prefrontal cortex and afferents from subcortical regions back to the hippocampus. ³⁰ While pre‐ and postcommissural or afferent/efferent fibers could not be separately analyzed here, these fibers may serve different roles (e.g., cognitive control vs. memory consolidation).^{31, 32} This anatomical complexity highlights the fornix's involvement in diverse cognitive functions within the limbic system. The broader role of the fornix in cognitive processes as opposed to the more specific role of the hippocampus supports the hypothesis that the correlation between fornix FA and PSI reflects microstructural integrity in pathways other than the hippocampal efferent to the mammillary bodies.

AUTHOR CONTRIBUTIONS

S.A.A. processed all experimental data, performed the analyses, and drafted the manuscript with input from all authors. J.U. conducted the cognitive assessments and provided guidance on the analysis and interpretation of neuropsychological data. C.B. and D.W.G. conceived the project and oversaw design, direction, and execution.

CONFLICT OF INTEREST STATEMENT

None of the authors have any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon request.