The proper identification of an epileptic focus is a vital diagnostic challenge particularly in non-lesional epilepsy [20]. Although scalp EEG and video-EEG monitoring are the most common methods used for focus localization, the data given by them are limited [21]. Conventional MRI remains the method of choice for identification of the gross structural brain lesions. However, normal MRI findings are observed in about 30% of patients with FIAE of temporal lobe origin [20]. Epilepsy is ultimately a disease of brain function, in which seizures are accompanied by alterations of rCBF and cerebral metabolism [21].
The PASL-MRI is a promising, non-invasive and non-gadolinium-dependent perfusion-based MRI acquisition, which facilitates quantitative estimation of the rCBF and provides serial imaging with greater accessibility than SPECT or PET techniques and without radiation exposure or contrast media [11].
The study in hand goes parallel with the observations of previous studies [1, 22, 23] and revealed that the temporal lobe is the commonest affected lobe with FIAE, while most of patients with extra-temporal lobe FIAE have frontal lobe epilepsy.
The visual-based analysis of PASL-MRI identified the epileptogenic zones with well-lateralization and well-localization in most of the studied patients (21/26; 80.77%). As reported by other previous studies [7, 24,25,26], the epileptogenic zones appeared as ipsilateral regions of abnormal perfusions, either hypo-perfusion with decreased rCBF (19/21; 90.48%) or hyper-perfusion with increased rCBF (2/21; 9.52%), which displayed cool and warm colours, respectively, according to the colour scale of the perfusion maps. Actually, the definite pathological mechanism of the inter-ictal focal hypo-perfusion of the epileptogenic zone with irritative focus is not yet clear. It could be attributed to cortical atrophy, gliosis, reduced synaptic density and/or neural loss in the epileptogenic zone [27]. However, a possible positive linear correlation between the inter-ictal perfusion of the epileptogenic zone and the frequency of the inter-ictal epileptic activity might exist to explain the interictal hyperperfusion in epileptogenic zones [28].
Moreover, the visual-based analysis of PASL-MRI data of the current study demonstrated false localization of the epileptogenic zone in 2 patients (2/26; 7.69%) with temporal non-lesional FIAE (Figs. 3 and 4) and false lateralization in 2 other patients (2/26; 7.68%) with parietal and temporal lobe FIAE (Figs. 5 and 6). In all of them, there was hypoperfusion with cool colour in the focal epileptogenic zones as well as in areas of seizure spread, which can be attributed to the rapid seizure spread or sub-clinical seizure onset [23].
Noteworthy, this visual-based analysis yielded 90.48% sensitivity, 87.10% specificity and 88.46% accuracy in proper identification of the epileptogenic foci. However, Kim et al. [11] reported a diagnostic ability of the inter-ictal visual analysis of PASL-MRI in patients with non-lesional FIAE of 74% sensitivity, 0% specificity and 70% accuracy. These variations might be attributed to differences in the duration of epilepsy, variations in labelling techniques and different post-processing methods. In fact, the visual-based analysis is inefficient in detecting slight changes in the rCBF, which are vulnerable to subjective factors of the observers that significantly influence its diagnostic performance [26].
Therefore, to improve the diagnostic performance of PASL-MRI, we calculated the ASLAI% and found that the significant increase in the ASLAI% (at cut-off value ≥ 5.96% and AUC of 0.968 with 96.01% sensitivity, 91.70% specificity and 95.78% accuracy) and reduced rCBF (at cut-off value ≤ 64.87 and AUC of 0.957 with 98.31% sensitivity, 92.90% specificity and 94.75% accuracy), facilitated proper localization and lateralization of the epileptogenic zone. Parallel to our results, Eryurt et al. [29] utilized a cut-off value 0.89 for reduced rCBF and a cut-off value 5.8% for increased ASLAI% with 71% sensitivity and 78% specificity. On the contrary, Guo et al. [30] used a cut-off value 4.0% for significant hippocampal ASLAI% increase (with 80.8% sensitivity and 72.7% specificity), while they used a cut-off value 4.5% for amygdala (with 65.4% sensitivity and 77.3% specificity). However, they reported significant reduction in the rCBF ipsilateral in the epileptogenic temporal lobe, compared to the contra-lateral non-epileptic temporal lobe and control, with significant increase in the ASLAI%. These differences might be attributed to differences in the region of seizure, post-processing technique and statistical methods.
The 1H-MRS can be considered as an in-vivo metabolic biopsy, which has the potential to provide data about neurochemical changes in a selected volume of the brain that may precede the development of structural lesions particularly in patients with negative MRI. This can be achieved by quantification of the endogenous cerebral metabolites including NAA (a marker for neuronal status and integrity with its reduction is interpreted as mitochondrial metabolism dysfunction or neuronal loss), Cho (a marker for membrane integrity and turnover), Cr (a marker for energy metabolism), MI (a marker for glial cell integrity) and Glx (is an excitatory neurotransmitter, which plays a role in mitochondrial metabolism and can serve as a marker of epileptic networks) [20, 31].
In accordance with previous studies [1, 32, 33], we observed a significant reduction in the mean peak ratios of NAA/Cho, NAA/Cr and NAA/(Cho + Cr) (P < 0.02, P = 0.01 and P < 0.001, respectively), in the proposed zone of focal epileptogenic activity when compared to the homologous region of the contra-lateral non-epileptogenic region. However, Zhang et al. [20] and Davis et al. [34] stated that, in patients with temporal lobe epilepsy, NAA/(Cho + Cr) value ipsilateral to the EEG epileptogenic focus was not significantly lower than that of the corresponding contralateral regions. Moreover, Davis et al. [34] observed elevated NAA/Cr in the epileptogenic focus when compared to the corresponding contralateral region. However, Simister et al. [35] concluded that temporal lobe epilepsy was associated with reduction in NAA/Cr ratio in both ipsilateral and contra-lateral temporal lobes. Also, Ercan et al. [36] observed a non-significant reduction in both NAA/Cr and NAA/(Cho + Cr) ratios in patients with left temporal lobe epilepsy compared to normal controls. These dissimilarities may be referred to differences in the study design, selection criteria, the time of estimation of these metabolites after seizure onset and the disease chronicity. Many explanations were provided to elucidate the mechanism of the NAA reduction in the proposed epileptogenic zone, including neuronal loss, seizure duration, frequency and severity [23] in addition to reduction in synaptic density due to metabolic impairment [36].
The Glx complex is highly involved in the excitatory neuronal activity, and its increased level can be considered as a marker of epileptogenic processes particularly in non-lesional epilepsy [37]. Many previous studies that support the hypothesis of increased glutamate level in the epileptogenic foci assume that elevated glutamate level within the glial-neuronal unit is a key sign of both mitochondrial and metabolic injury induced by the hyper-excitable state that characterizes seizures [37]. Also, the elevated extracellular Glx concentration in those patients can be explained by reduction in glutamate reuptake [32]. Additionally, Savic et al. [38] stated that the alterations in Glx/NAA and Glx/Cr peak ratios could play a crucial role in evaluating patients of temporal lobe epilepsy with negative MRI.
In agreement with the results obtained by previous studies [34, 35, 38], the present study revealed a significant increase in the mean peak ratios of Glx/NAA and Glx/Cr (P < 0.001 and P < 0.018, respectively) in the suggested zone of focal epileptogenic activity when compared to the corresponding region in the contralateral hemisphere. In contrast, Hammena et al. [39] reported an increase in the Glx concentration contra-lateral to the epileptogenic focus in patients with MRI negative results and postulated a possible neuronal damage contra-lateral to epileptogenic focus.
The peak value of MI concentration could be an essential marker of the biologic effects of overt and/or occult seizure activity, which can provide an early key sign of secondary brain damage [40]. The study in hand demonstrated a significant increase in the mean MI/NAA peak ratio in the suggested zone of focal epileptogenic activity when compared to the homologous region on the contralateral hemisphere (P = 0.044). On the contrary, Doelken et al. [32] noted a non-significant change in cerebral MI concentration in both temporal lobes of MRI-negative patients with temporal lobe epilepsy. However, Wellard et al. [40] observed an elevation of MI concentration ipsilateral to the epileptogenic temporal lobe in patients with hippocampal sclerosis compared to the corresponding contra-lateral temporal lobe and normal controls. They attributed this change to the associated astrocytosis [40].
Temporal lobes have rich functional interconnectivity with other brain regions, in which the posterior lateral temporal region serves as the spot of a “functional hub” which is a key for the propagation of ictal activity to other brain regions, causing the clinical phenomenon of secondary generalization [41]. Obviously, patients with temporal lobe epilepsy show an additional hypometabolism of the frontal, parietal and occipital lobes, which represents the inhibitory phenomena initiated by the epileptogenic focus [42, 43].
In the current study, despite the significant increase of the MI/NAA ratio in the epileptogenic zones in patients of false localization (Figs. 3 and 4) and false lateralization (Figs. 5 and 6) on PASL-MRI, there was a significant reduction of this metabolite ratio in the regions of seizure spread (P < 0.01), which might be attributed to the osmolyte changes caused by seizures [40]. These areas with diminished MI/NAA ratios were considered as cerebral areas of propagation of epileptogenic activity. Wellard et al. [40] advocated the use of this finding in discrimination between the seizure focus with increased MI concentration and the areas of seizure spread with reduced MI concentration. The hypometabolism observed in other brain lobes ipsilateral and/or contra-lateral to the epileptogenic temporal lobe can be attributed to the seizure spread and propagation to these brain regions [23, 31, 43].
In humans, the thalamus displays reciprocal widespread connections with subcortical structures and other cortical regions [44] and is thought to be responsible for the initiation and expression of partial and generalized seizures. Therefore, the role of the thalamus in localization-related epilepsy has currently been receiving attention [45]. The present study reported a case with right temporal lobe FIAE that appeared normal by PASL-MRI and considered a false negative result, while the 1H-MRS showed increased Glx/NAA, decreased NAA/(Cho + Cr) and decreased MI/NAA ratios of the right temporal lobe and ipsilateral thalamus (Fig. 1). This goes in harmony with the results obtained by Jain et al. [46], who stated that ipsilateral thalamic hypometabolism played an important role in localization of the epileptogenic focus in patients with temporal lobe epilepsy.
The present study demonstrated significant thalamic changes detected ipsilateral to the epileptogenic zone in two patients [one with frontal lobe FIAE (Fig. 2), and one with temporal lobe FIAE (Fig. 3)], and in the contralateral thalamus in one patient with right temporal lobe FIAE (Fig. 5). These thalamic changes included significant reduction in the rCBF (P = 0.023) by PASL-MRI, in addition to a significant reduction in the NAA/(Cho + Cr) ratio (P < 0.001) and a significant increase in Glx/NAA ratio (P = 0.02) by 1H-MRS. This is supported by the results of previous studies [44,45,46,47,48], which recommended the utilization of interictal thalamic hypoperfusion and/or metabolic alterations as a complementary sign for proper localization and lateralization of epileptogenic foci in patients with temporal and frontal lobe epilepsy.
A mirror focus is a special pattern of secondary epileptogenesis in which activity is provoked through dense inter-hemispheric callosal or commissural connection networks and considered as one of the signs of epilepsy progression [49, 50]. In a similar way, the current study showed a patient with left parietal non-lesional FIAE, with false lateralization of the epileptogenic zone based on the PASL-MRI results. The 1H-MRS of this patient detected properly the lateralization of the epileptogenic zone in the left parietal region, wherein there was reduced NAA/(Cho + Cr) and increased MI/NAA ratios in the epileptogenic zone (left parietal), while the MI/NAA ratio was decreased in the corresponding area in the right parietal region that represented a mirror focus of seizure spread (Fig. 6).
To the best of our knowledge, limited studies discussed the optimal cut-off value of PASLAI% for proper identification of the epileptogenic zone in patients with non-lesional FIAE. Additionally, the vast majority of prior spectroscopic studies addressed the metabolic alterations of the hippocampi and/or temporal lobe, while diminutive attention was paid to other brain regions with a wide argument [29, 30, 46].
The statistical data analysis of the MV 1H-MRS results yielded a cut-off value ≥ 9.98% for the increased mean %AF with AUC 0.981 (100% sensitivity, 97.65% specificity and 98.14% accuracy), and a cut-off value ≤ 0.59 for detection of significant decrease in the NAA/(Cho + Cr) ratio with AUC of 0.977 (100% sensitivity, 96.9% specificity and 97.74% accuracy), between the proposed epileptogenic zone and homologues contralateral region. In contrast, Eryurt et al. [29] suggested a cut-off value 8.7% for increased %AF between the epileptogenic temporal lobe and contralateral non-epileptic lobe with 86% sensitivity and 79% specificity. On the other hand, Elnekidy et al. [33] used a cut-off value 14% and 9% to the epileptogenic zones of the anterior and posterior portions of temporal lobe, respectively, while Sharma [51] used a cut-off value 12% to the total temporal lobe. Moreover, Elnekidy et al. [33] used a cut-off value 0.32 and 0.65 for detection of significant reduction in the NAA/(Cho + Cr) concentration ratio in the epileptogenic zones of the anterior and posterior portions of temporal lobe, respectively, while Burtscher and Holtas [52] utilized a cut-off value < 0.71% for identification of a significant reduction in the NAA/(Cho + Cr) ratio ipsilateral to the epileptogenic zone.
Moreover, in the zone of assumed epileptogenic focus, we recorded a mean of 1.03 ± 0.21 for significantly increased MI/NAA concentration ratio with an optimal cut-off value ≥ 0.98 and AUC 0.989 (100% sensitivity, 96.34% specificity and 97.86% accuracy), and a mean of 1.56 ± 1.25 for significantly increased Glx/NAA ratio with an optimal cut-off value ≥ 1.55 and AUC 0.972 (100% sensitivity, 96.35% specificity and 97.24% accuracy). On contrast, Aydin et al. [33] recorded a mean of 1.15 ± 1.19 and 1.16 ± 1.07 in the right and left hippocampi, respectively, with an AUC 0.450 ± 0.091, for significantly increased peak of Glx in the epileptogenic focus in patients with temporal lobe epilepsy, while they reported a mean of 0.76 ± 0.61 and 0.60 ± 0.42 in the right and left hippocampi, respectively, for significantly increased peak of MI with an AUC 0.434 ± 0.060. However, they could not display the optimal cut-off values for the significantly increased peaks of MI and Glx in the epileptogenic focus.
Furthermore, the current work demonstrated that in the interictal state, the combined use of PASL-MRI and MV 1H-MRS enabled accurate localization and lateralization of the epileptogenic zone in patients with FIAE with the highest sensitivity, specificity and accuracy (100%, 98.45% and 98.86%, respectively).