Lateralization of Interictal Temporal Lobe Hypoperfusion in Lesional and Non-Lesional Temporal Lobe Epilepsy Using Arterial Spin Labeling MRI

Abstract

Purpose:

Non-invasive imaging studies play a critical role in the presurgical evaluation of patients with drug-resistant temporal lobe epilepsy (TLE), particularly in helping to lateralize the seizure focus. Arterial Spin Labeling (ASL) MRI has been widely used to non-invasively study cerebral blood flow (CBF), with somewhat variable interictal alterations reported in TLE. Here, we compare temporal lobe subregional interictal perfusion and symmetry in lesional (MRI+) and non-lesional (MRI−) TLE compared to healthy volunteers (HVs).

Methods:

Twenty TLE patients (9 MRI+, 11 MRI−) and 14 HVs underwent 3T Pseudo-Continuous ASL MRI through an epilepsy imaging research protocol at the NIH Clinical Center. We compared normalized CBF and absolute asymmetry indices in multiple temporal lobe subregions.

Results:

Compared to HVs, both MRI+ and MRI− TLE groups demonstrated significant ipsilateral mesial and lateral temporal hypoperfusion, specifically in the hippocampal and anterior temporal neocortical subregions, with additional hypoperfusion in the ipsilateral parahippocampal gyrus in the MRI+ and contralateral hippocampus in the MRI− TLE groups. Contralateral to the seizure focus, there was significant relative hypoperfusion in multiple subregions in the MRI− compared to the MRI+ TLE groups. The MRI+ group therefore had significantly greater asymmetry across multiple temporal subregions compared to the MRI− TLE and HV groups. No significant differences in asymmetry were found between the MRI− TLE and HV groups.

Conclusion:

We found a similar extent of interictal ipsilateral temporal hypoperfusion in MRI+ and MRI− TLE. However, significantly increased asymmetries were found only in the MRI+ group due to differences in perfusion contralateral to the seizure focus between the patient groups. The lack of asymmetry in the MRI− group may negatively impact the utility of interictal ASL in seizure focus lateralization in this patient population.

1. Introduction

Epilepsy is one of the most common neurologic disorders [5]. Approximately one-third of patients with epilepsy continue to have seizures despite optimal medical management (drug resistant epilepsy, DRE), leading to significant morbidity and mortality [5]. Temporal lobe epilepsy (TLE) is a common cause of DRE, and epilepsy surgery has been shown to be more effective than ongoing medical treatment in these patients[38], with postoperative seizure freedom reported in up to 70% of TLE patients with clear epileptogenic lesions on MRI [35]. However, in the approximately 20-30% of TLE patients without MRI lesions, post-operative outcomes may be worse, with approximately 40% achieving seizure freedom in one study [35]. As a result, there is ongoing interest in adjunctive non-invasive imaging modalities to improve lateralization or localization of the seizure focus to guide surgical decision-making.

Arterial spin labeling (ASL) MRI is a non-invasive imaging method used to identify changes in cerebral perfusion. Unlike fluoro-D-glucose-positron emission tomography (FDG-PET) and ictal single-photon emission computed tomography (SPECT), two of the most widely used adjunctive imaging methods in epilepsy, ASL MRI does not require exogenous contrast agents or radiation and is logistically simple and relatively inexpensive to obtain when added on to widely available standard MRI studies [8]. For each of these adjunctive functional imaging methods, however, timing in relation to seizures significantly impacts the findings. Peri-ictal hyperperfusion and increased glucose metabolism is presumably related to transient increases in neuronal activity, local inflammation, or changes in the blood brain barrier [6, 26]. Hyperperfusion during or immediately after a seizure may be followed by a short period of significant post-ictal hypoperfusion, then by more subtle interictal hypoperfusion and regional hypometabolism. Indeed, interictal estimates of perfusion using ASL in TLE appear to be similar or highly correlated to estimates of metabolism obtained using FDG-PET [19, 18, 24, 2, 33, 9, 28, 37, 30].

Both peri-ictal and interictal changes in perfusion measured using ASL have been described in a wide variety of epilepsy syndromes. Peri-ictal hyperperfusion has been reported in hemimegalencephaly [1], idiopathic generalized epilepsy [3], status epilepticus [21, 22], and focal epilepsy [27, 36, 26, 10, 34, 25, 29]. Interictal ASL perfusion changes have been reported in idiopathic generalized epilepsy [32], tuberous sclerosis [40, 12], focal cortical dysplasia [39, 2], TLE [41, 20, 23, 31, 18, 37, 28, 11, 42], and more broadly, in drug resistant focal epilepsy [18, 24, 33, 11, 23, 9, 13, 13, 28, 37, 17, 16, 30]. Compared to healthy volunteers, immediate post-ictal hypoperfusion was found in 94% of TLE patients, compared to interictal hypoperfusion in 71.4-77.8%, consistent with the presumed seizure onset zone in 80% [10, 25]. These findings can be compared to a reported sensitivities in TLE of 85-90% using interictal FDG-PET[15], 75-97% using ictal SPECT, and approximately 50% using interictal SPECT [14].

In this study, we wished to further explore the differences in interictal perfusion changes in lesional versus non-lesional TLE. We hypothesized that lesional TLE patients may have more pronounced or more focal perfusion changes compared to non-lesional patients, who may have variable seizure focus locations or more diffuse alterations. In our small cohort, we found similar degrees of hypoperfusion ipsilateral to the seizure focus in both TLE groups, as in previous reports. However, we found significantly increased asymmetries only in the MRI+ TLE group. These findings are significant because they suggest that despite the presence of ipsilateral hypoperfusion, interictal ASL may be less useful for seizure focus lateralization in MRI− compared to MRI+ TLE.

2. Material and Methods

2.1. Study population

For this study, we included all patients who enrolled in a prospective non-invasive epilepsy imaging protocol at the NIH Clinical Center from May 2018, when this ASL protocol was instituted, through November 2020, when this data analysis began, who had 1) preoperative Pseudo-Continuous ASL (PCASL), 2) available 3T epilepsy protocol structural MRI, and 3) video-EEG confirmed temporal lobe epilepsy (TLE). We then excluded those with 1) destructive lesions on MRI (vascular malformation, tumor, large encephalomalacia), 2) bilateral or unclear temporal lobe seizure focus laterality on ictal EEG, and 3) low image quality on visual inspection (PCASL and/or structural) or poor alignment between scans. We obtained demographic and clinical information from the medical record. The time and type of most recent seizure was obtained just prior to scanning by study personnel and noted in a scan log.

Patients were identified as lesional (MRI+) or non-lesional (MRI−) based on the presence or absence of clinically identified abnormalities on structural MRI. The ipsilateral temporal lobe was defined as the side of seizure onset on EEG recordings. The control group consisted of all healthy volunteers (HVs) enrolled in the imaging protocol who had PCASL MRI scans. Enrolled HVs had no previous history of neurologic, psychiatric, or other significant illness that may affect the central nervous system. This protocol was approved by the National Institutes of Health Institutional Review Board (IRB) and all participants provided written informed consent prior to enrollment.

2.2. MRI acquisition

Research ASL and T1-weighted MPRAGE scans (T1wI) were collected in the NIH Nuclear Magnetic Resonance Center on a 3T GE Discovery MR750 scanner with a 32-channel head coil. A 3-D fast spin-echo (FSE) Pseudo-Continuous ASL (PCASL) sequence with background suppression was obtained with the following parameters: TR/TE: 4705/11.16 ms, PLD 1.525 s, label duration (τ): 1.45 s, slice thickness 3 mm, acquisition time 7:08 min, and saved as raw echo signals. Cerebral Blood Flow (CBF) was estimated by setting negative ASL values to zero, then proceeding according to the manufacturer manual (GE manual). For an assumed brain density of 1g/mL, CBF is reported in units of (ml of blood)/(100g of tissue)/minute.

Research T1wI were obtained with the following parameters: TR/TE 8.028/3.632 ms, flip angle 7, acquisition matrix 256 x 256, voxel size 1x1x1 mm. As part of our multimodality imaging protocol, a separate clinical 3D MPRAGE T1wI was acquired using a Philips Achieva 3T MRI scanner in the NIH Clinical Center Radiology Department with the following parameters: TR 6.8–7.2 ms, TE 3.2 ms, TI 900 ms, flip angle 90, voxel size 0.75 x 0.75 x 0.8 mm, acceleration factor 2 in slice direction.

2.3. Image processing and analysis

For further analysis and for incorporation into our multimodal imaging pipeline, which includes functional and structural MRI, diffusion tensor imaging, MEG, intracranial electrode locations and surgical resection masks, the research T1wI and ASL images had previously been aligned to each other and then to the patient’s clinical T1wI using the Analysis of Functional NeuroImages (AFNI) software package [4]. ASL images were co-registered to the research T1wI using AFNI’s align epi anat.py function with a rigid body transformation, the “edge method”, and an NMI cost function. The resulting registrations were visually assessed for alignment quality and found to be poor in seven subjects. For these subjects, alternate cost functions yielded acceptable registrations (LPA in 4 subjects, LPA+ZZ in 2 subjects, and NMI in 1 subject). The research and clinical T1wI were then co-registered using 3dAllineate and an NMI cost function, with the resulting transformation applied to the ASL images.

To allow for group level analysis of cerebral blood flow across subjects, we created brain masks for each subject and aligned the subjects’ masked ASL data to MNI space. To create a brain mask, we used in-house software to classify voxels in the T1wI as gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF). The T1wI was skull stripped and dilated by 5 voxels. A brain mask was created consisting of the intersection between GM and WM voxels and the skull stripped dilated T1 mask. The resulting brain masks were visually inspected for accuracy to ensure they covered the entirety of the brain and did not incorporate skull or large regions of CSF. The clinical T1wI were aligned to the MNI152 2009 template using linear followed by nonlinear transformation using 3dQwarp. We applied the same transformation to the aligned ASL and brain mask images.

2.4. Generating regions of interest (ROIs) and asymmetry indices

For this study, we created ROIs based on parcellations from the Brainnetome atlas [7], including the lateral temporal lobe (LTL), further subdivided into anterior and posterior subregions (aLTL, pLTL), and the medial temporal lobe (MTL), subdivided into the hippocampus, amygdala, and parahippocampal gyrus (PhG). For each subject, the median GM CBF was calculated using all GM voxels within the brain mask. To adjust for possible global perfusion differences across patients due to physiologic state or medication effects, we then normalized the CBF values at each voxel by dividing by the median GM voxel value, and computed the average normalized CBF (nCBF) for each ROI. We then calculated absolute asymmetry indices (aAI) for the average CBF within each ROI using the below equation [18]

$$\text{aAI}=100\cdot \frac{\begin{array}{c}\hfill \hfill \\ \hfill \text{ipsilateral}-\text{contralateral}\hfill \\ \hfill \hfill \end{array}}{0.5\cdot (\text{ipsilateral}+\text{contralateral})}$$ (1)

where ipsilateral and contralateral were defined relative to the patient’s clinically determined seizure focus.

2.5. Statistical Analysis

We performed statistical analysis using Python. Across group differences in age and epilepsy duration were computed using a unpaired t-test. For each group, nCBF and aAI were averaged across subjects for each ROI. For median CBF and aAI, results were compared across groups using a Kruskal-Wallis test. When significant differences were identified, a post-hoc Dunn’s test with Bonferroni adjustment was used to identify which pairs of group asymmetry indices were significantly different. For comparisons of nCBF across regions, Mann-Whitney U test was used for unpaired across group comparisons, while Wilcoxon rank-sum test was used for within group side-to-side comparisons.

2.6. Data and code availability statement

Anonymized imaging data will be made available upon reasonable request. Code used for data analysis in this study is available on GitHub (https://github.com/InatiLab/ASLproc).

3. Results

3.1. Study participants

We identified sixty-three consecutive patients who completed PCASL MRI imaging from 2018-2020. Of these, forty-one patients had video-EEG confirmed TLE. Four patients were excluded due to the presence of destructive lesions, fifteen for bilateral or unclear seizure onsets, and two due to poor PCASL or structural MRI image quality. The remaining twenty patients and fourteen healthy volunteers met the inclusion criteria for this study. Group level information about the clinical characteristics of the included subjects is summarized in Table 1, and individual patient characteristics are detailed in Supplementary Table S-1. There were no significant differences in clinical characteristics across groups except for a significantly increased duration of epilepsy in the MRI+ compared to the MRI− group (p=0.0273). In the MRI+ group, seven patients had mesial temporal sclerosis (MTS), one had mesial temporal focal cortical dysplasia and one had periventricular nodular heterotopias. Laterality of ictal EEG onset on video-EEG monitoring was concordant with structural lesions in all MRI+ patients. Four of nine patients in the MRI+ group underwent surgical resection, all with Engel 1 seizure outcomes at one year follow up. Seven of eleven in the MRI− group underwent surgical resection, with one year seizure outcomes of Engel 1 in three patients and Engel 2 in four patients.

Table 1: Summary of demographic and clinical information by group.

N=number of participants. HS=hippocampal sclerosis. FCD=focal cortical dysplasia. EEG Lat=seizure focus lateralization based on video electroencephalography monitoring. ^a One participant had HS and FCD, ^b One participant had combination of early HS, FCD, and microdysgenesis. ^c One participant had bilateral foci with right temporal seizures and independent left frontal seizures (only right temporal focus was considered for this analysis).

Characteristic	Lesional (MRI+)	Non-Lesional (MRI−)	Healthy Volunteers
N	9	11	14
Age (years)
Median Range	30.2 [18.1-45.8]	35.6 [20.6-56.3]	25.3 [21.3-55.5]
Sex, n (%)
Female	5 (55.6)	7 (63.6)	8 (53.3)
Epilepsy duration (years)
Median Range	18.5 [2.1-27.5]	8.6 [1.3-18.6]	-
Surgery, n (%)	4 (44.5)	7 (63.6)	-
Pathology, n
HS FCD	2a	1b 1b	-
Microdysgenesis	2a 1	4b 3	-
Gliosis	-		-
EEG Lat, n (%)
Left Right	6 (66.7)	6 (54.5)	-
	3c (33.3)	5 (45.5)	-

Based on patient interview just prior to the scan, no patients had generalized tonic-clonic seizures within 48 hours of the scan. Two patients (one in each group) had frequent brief focal retained awareness seizures, so may have had one the day of scanning, although this was not specifically noted. One patient in the non-lesional group had a brief focal impaired awareness seizure several hours prior to the scan. No other patients had known seizures within 24 hours of the scan, and no seizures were noted during the scanning session.

3.2. Individual Subject Normalized CBF and Asymmetry Estimates

Based on previous literature, we expected interictal ipsilateral temporal lobe hypoperfusion in TLE patients relative to HVs, as well as to their own contralateral temporal regions, leading to increased perfusion asymmetry. In Figure 1, representative images including those used in our group level analysis are shown for several individual subjects. Visual inspection of the images demonstrates hypoperfusion (and negative asymmetry) in the ipsilateral anterior temporal lobe (best visualized in the more inferior axial slices) and the mesial temporal region, with more complex regional changes in the more superior and posterior neocortical temporal regions. Of note, for comparison across subjects and groups, GM CBF was median normalized within each subject. No significant difference in overall median GM CBF was noted across groups (median (IQR) in mL/100g/min for HVs: 44.32(9.62), MRI+: 47.93(12.56), MRI−: 46.06(14.09), p=0.4026), suggesting that any subsequent differences noted across groups are due to regional and not global changes in cerebral blood flow.

Figure 1: Representative Cases.

Rows 1 and 2: MRI− TLE patients with gliosis (Row 1) and microdysgenesis (Row 2) on pathology. Row 3: MRI+ TLE patient with mesial temporal sclerosis and adjacent focal cortical dysplasia on pathology. Row 4: healthy volunteer. Two representative axial slices through the temporal lobe are shown (left box more inferior). Columns from left to right: pre-operative T1-weighted images, post-operative T1-weighted image (for patients), raw CBF map, normalized CBF map, averaged within each Brainnetome parcel, and negative asymmetry indices (indicating hypoperfusion relative to the other side of the subject’s brain for each parcel)

3.3. Group Level CBF Alterations: Medial and Lateral Temporal Lobes

Across groups, we initially compared perfusion broadly across the medial and lateral temporal lobe subregions (MTL and LTL) (Fig. 2). We found significant differences in both asymmetry (aAI) and normalized perfusion (nCBF) between the three groups in both the LTL (p=0.00642) and MTL (p=0.0296). Pairwise comparisons showed that the MRI+ group had significantly greater aAI in the LTL compared to both the HV (p=0.0167) and MRI− groups (p=0.00344), with no significant aAI difference between the MRI− and HV groups. In the MTL, there were no significant aAI differences between the HV and patient groups, although MRI+ patients had significantly higher aAIs compared to MRI− patients (p=0.0109).

Figure 2: CBF Alterations: Lateral and Mesial Temporal Lobes.

(A) Lateral (LTL) and Medial (MTL) temporal lobe asymmetry index absolute values for Healthy Volunteers (HV), Lesional Patients, and Non-Lesional Patients. (B) Lateral (LTL) and Medial (MTL) normalized CBF for healthy volunteer (HV), lesional ipsilateral (LI) and contralateral (LC), and non-lesional ipsilateral (NI) and contralateral (NC). Values are plotted as mean and standard error of the mean across participants in each group. *between group or between ROI significant differences (corrected p<0.05)

We then directly compared ipsilateral and contralateral nCBF findings in each group. Compared to HVs, we found ipsilateral hypoperfusion in both the MRI+ and MRI− TLE groups in the LTL (p=0.0224 and p=0.0379 respectively) and MTL (p=0.0102, p=0.0351 respectively), with no significant difference in the ipsilateral perfusion between patient groups (LTL p=0.3233, MTL p=0.4033). Contralaterally, we found hyperperfusion in the MRI+ group compared to healthy volunteers in the LTL (p=0.0384) but not the MTL (p=0.3130), and relative to the contralateral LTL (p=0.0024) and MTL (p=0.0482) in the non-lesional group. There were no significant perfusion differences between perfusion in the healthy volunteers and the contralateral non-lesional LTL (p=0.4082) or MTL (p=0.889). As expected based on the asymmetry evaluation, there was significant hypoperfusion ipsilaterally versus contralaterally in the lesional LTL (p=0.0023), and MTL (p=0.0070) but not in the non-lesional group (LTL 0.1580, MTL 0.8182).

3.4. Group Level CBF Alterations: Medial Temporal Subregions

As nCBF differences were found in the MTL across groups, we further investigated which subregions contributed to these changes. Using the Brainnetome Atlas, the MTL was divided into three subregions: hippocampus, amygdala, and parahippocampal gyrus (PhG) (Fig. 3). We found significant group level differences in the absolute asymmetry indices (aAI) in the PhG (p=0.0221), with pairwise comparisons showing increased aAI in the MRI+ group compared to both the HV (p=0.0322) and MRI− groups (p=0.00555), with no significant aAI difference between the MRI− and HV groups. We found no significant between group differences in the hippocampal (p=0.188) or amygdalar (p=0.512) aAIs.

Figure 3: CBF Alterations in MTL Subregions.

(A) Hippocampus (Hipp), Amygdala (Amyg), and Parahippocampal Gyrus (PhG) asymmetry index absolute values. (B) Hippocampus, Amygdala, and PhG CBF for healthy volunteer (HV), lesional ipsilateral (LI) and contralateral (LC), and non-lesional ipsilateral (NI) and contralateral (NC) regions of interest. Values are plotted as mean and standard error of the mean across participants in each group. * between group or between ROI significant differences (corrected p<0.05)

Assessing nCBF directly, in the hippocampal ROI compared to HVs we observed significant hypoperfusion in the ipsilateral hippocampus in both the MRI+ and MRI− groups (p=0.0055 and p=0.0100 respectively). In the contralateral hippocampus, there was significant hypoperfusion compared to healthy volunteers in the MRI− group (p=0.0475) but not the MRI+ group (p=0.8180). Although there were no significant differences in the hippocampal aAIs, in this analysis there was significant hypoperfusion in the ipsilateral compared to the contralateral side in the MRI+ group (p=0.0118). There was no significant side-to-side difference in hippocampal nCBF in the MRI− group (p=0.6224), suggesting that the lack of asymmetry in the MRI− group derives from bilateral hippocampal hypoperfusion, rather than lack of changes in perfusion.

In the PhG, pairwise comparisons to HVs showed significantly lower nCBF in the ipsilateral PhG in the MRI+ group (p=0.0186) but not in the MRI− group (p=0.1382). While there were no significant differences between TLE groups in PhG perfusion ipsilaterally (p=0.1286), there was significant relative hyperperfusion in the MRI+ compared to MRI− groups contralaterally (p=0.0098). Similarly to the aAI analysis, there was a significant difference in PhG perfusion between sides in the MRI+ group (p=0.0071) but not the MRI− group (p=0.8696). There were no significant perfusion differences in the amygdala ROIs within or across groups.

3.5. Group Level CBF Alterations: Lateral Temporal Subregions

Given our findings of lateral temporal interictal hypoperfusion and based on visual inspection of the aAI and nCBF maps, we wished to quantify the extent to which LTL perfusion abnormalities appeared in the anterior compared to posterior lateral temporal ROIs (Fig. 4).

Figure 4: CBF Alterations in LTL Subregions.

(A) Anterior and posterior lateral temporal lobe (LTL) asymmetry index absolute values for healthy volunteers, lesional patients, and non-lesional patients. (B) Anterior and posterior LTL normalized CBF for healthy volunteers (HV), lesional ipsilateral (LI) and contralateral (LC), and non-lesional ipsilateral (NI) and contralateral (NC) groups. Values are plotted as mean and standard error of the mean across participants in each group. * between group or between ROI significant differences (corrected p<0.05)

We found significant between group differences in asymmetry (aAI) in the anterior (aLTL) (p=0.0116) but not the posterior lateral temporal regions (pLTL) (p=0.193). Pairwise comparisons in the aLTL showed significantly higher asymmetry in the MRI+ group compared to both the HV (p=0.0098) and MRI− (p=0.0070) groups. No significant aAI differences were found between the MRI− and healthy volunteer groups.

Direct assessment of perfusion (nCBF) in the LTL subregions again yielded somewhat distinct results compared to the asymmetry findings. In the aLTL, pairwise comparisons with the HV group showed significantly lower ipsilateral perfusion in both the MRI+ (p=0.0068) and MRI− (p=0.0100) groups. No significant differences were observed between HV and either TLE group in the contralateral aLTL or pLTL. In the MRI+ group, we found significant side-to-side differences, with relatively decreased nCBF in the ipsilateral compared to contralateral aLTL (p=0.0054) and pLTL (p=0.0054). There were no significant side-to-side nCBF asymmetries in the anterior or posterior LTL in the MRI− group. Similar to our other results, compared to the contralateral aLTL and pLTL in the MRI+ group, there was relative hypoperfusion MRI− group (aLTL p=0.0227; pLTL p=0.0275).

4. Discussion

Using interictal ASL MRI in our cohort of patients with TLE, we add to the limited literature assessing regional patterns of altered perfusion in lesional versus non-lesional TLE. Compared to healthy volunteers, in both lesional and non-lesional TLE, we found hypoperfusion in the ipsilateral mesial and lateral temporal lobes, as well as in the hippocampal and anterior lateral temporal neocortical subregions, therefore not confined to lesional regions. Despite having similar extents of ipsilateral hypoperfusion, we found increased perfusion asymmetries only in the lesional TLE group. The differences in asymmetry findings arose from differences in contralateral perfusion, which was relatively preserved in the MRI+ group, and was significantly higher than perfusion in the corresponding regions in the MRI− group. These findings are clinically important in that they support the presence of interictal perfusion changes in TLE, but with a less lateralized pattern of changes non-lesional TLE patients. Accordingly, asymmetry indices may fail to detect changes in non-lesional TLE patients.

Adjunctive studies in TLE are frequently used to aid in lateralization of the seizure focus and are particularly of clinical utiity in patients without clear abnormalities on structural MRI. The presence of asymmetries is particularly important in this context, with the side of the seizure focus typically visually identified by identification of hypoperfusion in comparison to the contralateral side. Indeed, reliance on asymmetry indices may underlie some of the inconsistencies in previous studies using interictal ASL for seizure focus lateralization in TLE. While some studies have reported increased medial temporal perfusion asymmetry [41, 19, 23], others found non-significantly increased mesial and lateral temporal perfusion asymmetry [18], or no significant hippocampal and amygdalar perfusion asymmetries [31]. However, similar to our findings, the latter studies that failed to find significant asymmetries did identify significant ipsilateral hypoperfusion in the mesial temporal lobe [18], and in the temporal neocortex, amygdala and olfactory cortex [31] when comparing directly to healthy volunteer normative data. This suggests while assessment of asymmetry is important in visual assessment of images and is convenient as no control population is required, it may be less sensitive to disease-related changes in perfusion than direct comparison to normative data, particularly in the setting of more widespread or bilateral perfusion changes.

Another cause of inconsistency in ASL studies in TLE may lie in the variable mix of underlying pathologies, with most studies including a mix of lesional and non-lesional patients. While identification of perfusion changes in lesional TLE may be important in understanding pathology-related changes in physiology, it has less utility in clinical practice, as lateralization is typically already established based on the MRI findings. Only a few studies have specifically accounted for the lesional status of the included patients. One study limited to SEEG-proven unilateral mesial TLE with unilateral MTS found significant ipsilateral hippocampal ASL hypoperfusion [37]. Another study compared ASL findings in non-lesional patients compared to a mixed group, reporting a decreased ability to lateralize the seizure focus in the non-lesional patients [31]. Another study of apparently non-lesional TLE with focal cortical dysplasia found no areas of significant hypoperfusion using an SPM-based analysis, although both visual inspection and asymmetry analysis suggested some alterations in temporal perfusion in many of these cases [28]. These studies suggest that ASL perfusion abnormalities may be more subtle in non-lesional compared to lesional TLE, but did not include direct comparisons between the groups.

To our knowledge, only two previous studies have directly compared lesional to non-lesional TLE interictal ASL perfusion findings. These studies differ in regions of interest selected as well as analysis methods, making comparisons difficult. In the hippocampal region, both previous studies reported increased perfusion asymmetry in both MRI+ and MRI− TLE [11, 42]. In contrast, we found asymmetry only in the MRI+ group, with bilateral relative hypoperfusion in the MRI− group and a resultant lack of asymmetry. Several studies have specifically reported perfusion changes in the amygdala, possibly more prominent in lesional TLE [31, 11], which we did not observe in our cohort. We observed significant parahippocampal gyrus asymmetry in the MRI+ TLE group, which was not specifically addressed in other studies. Further studies specifically in lesional or non-lesional patient groups will likely be required to further investigate these subregional changes.

Finally, we observed differences in contralateral perfusion between lesional and non-lesional patients across multiple subregions. These findings most consistently suggested relatively preserved or even increased perfusion in lesional patients, with relative hypoperfusion in non-lesional patients suggesting more widespread or bitemporal pathology in our MRI− cohort. As there were no across group differences in median gray matter perfusion, these changes are more likely regional than global, extending at least to the anterior lateral temporal neocortex based on our findings. Similarly, Guo et al. found more widespread cortical hypoperfusion in the non-lesional TLE group [11]. In contrast, Zhang et al. reported more widespread cortical abnormalities in the lesional compared to the non-lesional TLE group [42].

This study has several limitations. As with many similar studies, we include a relatively small number of patients with heterogeneous underlying pathologies, each of which may have variable effects on cerebral perfusion. Only approximately half of these patients have undergone epilepsy surgery to date, limiting definitive confirmation of the seizure focus based on seizure outcomes. Additionally, some of the ASL differences between the MRI+ and MRI− groups could be attributable to the longer duration of epilepsy in the MRI+ group. Timing of interictal ASL scans in relation to previous seizure activity was not specifically accounted for in this study and may affect the findings. However, this reflects typical clinical practice when interictal ASL is used as an add-on screening tool to previously scheduled clinical structural MRI scans, which tend to be scheduled well in advance without regard to timing of last seizure. Finally, in this analysis, we examined group-level findings but did not attempt to identify single subject-level areas of abnormality. This is a future direction requiring a larger number of patients who have undergone epilepsy surgery, and likely a larger cohort of healthy volunteers to better account for subject to subject variability in perfusion findings.

5. Conclusions

In summary, we found ipsilateral hypoperfusion in the medial and lateral temporal regions in both lesional and non-lesional TLE using interictal ASL MRI. However, we found significant asymmetries only in the lesional TLE group. The difference between lesional and non-lesional perfusion patterns seemed to be driven largely by differences in perfusion contralateral to the presumed seizure focus, with relative preservation in the lesional group, and comparative hypoperfusion in the non-lesional TLE group. These findings are significant because they suggest that direct comparison to normative data may be more sensitive in identifying interictal perfusion changes in TLE than asymmetry indices, particularly in non-lesional TLE. The lack of asymmetries in non-lesional TLE is clinically important, as standard radiologic interpretation of images in clinical practice typically relies upon visual identification of side-to-side changes, which may be absent in these patients. Therefore, while our findings provide additional support for the presence of interictal ipsilateral perfusion abnormalities in both lesional and non-lesional TLE, the diagnostic utility of ASL as an adjunctive tool for seizure focus lateralization may be more limited in non-lesional patients.

Highlights:

• There is ipsilateral mesial and lateral temporal hypoperfusion in MRI+ and MRI− TLE.

• There is Ipsilateral anterior temporal hypoperfusion in MRI+ and MRI− TLE.

• There is ipsilateral hippocampal and parahippocampal hypoperfusion in MRI+ TLE.

• There is bilateral hippocampal hypoperfusion in MRI− TLE.

• Temporal lobe perfusion asymmetries are present in MRI+ but not MRI− TLE

Abbreviations

aAI

absolute asymmetry indices

ASL

arterial spin labeling

aLTL

anterior lateral temporal lobe

CBF

cerebral blood flow

FDG-PET

fluoro-D-glucose-positron emission tomography

HVs

healthy volunteers

LTL

lateral temporal lobe

MNI

Montreal Neurological Institute

MRI+

magnetic resonance imaging positive

MRI−

magnetic resonance imaging negative

MTL

mesial temporal lobe

nCBF

normalized cerebral blood flow

PCASL

Pseudo-Continuous ASL

PhG

parahippocampal gyrus

pLTL

posterior lateral temporal lobe

ROI

region of interest

SPECT

single-photon emission computed tomography

T1wI

T1 weighted image

TLE

temporal lobe epilepsy