1 Introduction

Treatment of mesial temporal lobe epilepsy (MTLE) is predominantly surgical, with the goal of completely removing or disconnecting the epileptogenic zone (EZ) [1]. Anterior temporal lobectomy (ATL) and selective amygdalohippocampectomy (SAH) are proven effective therapies for medically refractory MTLE, with 60%–80% of patients experiencing seizure freedom [2, 3], but they can result in possible side effects such as cognitive impairment, visual field defects (VFDs), intracranial hemorrhage, and neurological damage [4]. Thus, minimally invasive techniques such as stereoelectroencephalography-guided radiofrequency thermocoagulation (SEEG-guided RFTC) and magnetic resonance-guided laser interstitial thermal treatment (MRgLITT) provide promising alternatives to conventional surgery, resulting in less patient discomfort, shorter hospitalizations, better brain function preservation, and fewer surgical complications [5, 6].

Despite the curative nature of MTLE surgery, 30%–40% of patients continue to experience recurrent seizures with loss of awareness [7]. And those who achieve seizure freedom may see a gradual decline in their seizure freedom rate over time following either resective surgery [8, 9], or SEEG-guided RFTC [10]. The reasons for suboptimal surgical outcomes are complicated, and several studies have examined presurgical parameters that may be linked to seizure freedom. Magnetic resonance imaging (MRIs) showing unilateral hippocampal sclerosis (HS) frequently show excellent postoperative outcomes [11, 12], anatomical resection site and extent have been found to be important predictors of good outcomes, with seizure freedom requiring hippocampus removal. [13, 14] However, more recent research has questioned this idea [15, 16] and has shown that additional mesial structures, such as the entorhinal [17] and piriform cortices [18], which may independently cause or sustain TLE seizures [19], may also affect outcomes. Finally, many recent whole-brain neuroimaging investigations have shown that the structural connectivity (SC) and functional connectivity (FC) of the mesial temporal lobe can also promote TLE seizures when compared to a healthy control group [20-23]. However, these aforementioned investigations only included patients who underwent resective surgery and could therefore only study the postoperative brain connectivity of the regions spared during surgery, or the contralateral brain side of the affected regions of interest (ROIs).

During SEEG-guided RFTC, the ablation of the different ROIs is often only partial, leaving certain sections of these regions intact. In the case of MTLE, the analysis of the alterations in SC and FC between the preoperative ROIs and their respective postoperative remaining portions might be clinically relevant in explaining the reasons for suboptimal surgical outcomes in patients undergoing SEEG-guided RFTC. To date, no studies have assessed the association between the FC and SC alterations and postoperative outcome in MTLE patients who underwent SEEG-guided RFTC, using their preoperative and postoperative remaining ROIs.

This retrospective study aims to examine the association between postoperative lesions in distinct ROIs of the brain and the impact that their ablation would have on the structural and functional brain connectivity relative to outcomes. Based on the evidence supporting the importance of various mesial temporal cortical structures in seizure onset in MTLE [13-19], we hypothesized that a larger surgical ablation of specific ROIs within the temporal lobe could have led to the positive postoperative outcome seen in those experiencing seizure freedom. We further hypothesized that the ablation of these ROIs could have altered their brain connectivity, potentially leading to improved surgical outcomes. We employed a pre and postoperative voxel-by-voxel ablation mapping comparison approach, along with a structural and functional connectivity alterations analysis of the ROIs relative to the postoperative outcomes to test these hypotheses.

2 Methods

2.3 SEEG Implantation and RFTC Procedure

SEEG-guided RFTC was recommended as a surgical option to the included patients after their refusal to undergo resective surgery. The electrode placement was based on preoperative hypotheses derived from semiology, scalp EEG, MRI, and PET findings regarding the localization of the EZs. Avascular electrode trajectories were planned using MRI and magnetic resonance angiography (Sinovation Medical, Beijing, China). Since all patients were suspected of MTLE, electrode implantation focused on the temporal lobe, including the hippocampus, amygdala, rhinal cortex, parahippocampal cortex, and temporal gyri (superior, middle, inferior) (Figure 1) (Appendix S1 for more in-depth details).

2.6 Gray and White Matter Volumetric Analysis

2.6.1 Preoperative Image Processing and Parcellation

Each patient's preoperative T1-weighted (T1W) MRI underwent automatic parcellation with standard anatomic definitions using FreeSurfer version 7.4.0, which is documented and freely available for download online (https://surfer.nmr.mgh.harvard.edu/) [26, 27]. The postoperative MRIs were co-registered to their preoperative MRIs using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). (Appendix S1 for more in-depth details).

2.6.2 Manual Segmentation of Post-RFTC Cavities

Using each patient's 1-year postoperative T1W MRI, two investigators independently performed the manual segmentation of the post-RFTC cavities using 3D Slicer version 5.2.2 (http://www.slicer.org) (Figure 2) (Appendix S1 for more in-depth details).

2.6.3 Choice of ROIs and RFTC Volumetric Analysis

All of the seven ROIs that were implanted were selected: the left or right (according to the RFTC side) hippocampus, amygdala, rhinal cortex, parahippocampal cortex, superior, middle, and inferior temporal gyrus. The “Subtract Scalar Volumes” module from 3D Slicer was used in order to obtain the postoperative ROIs. This module substrates each ablated portions structures from their respective preoperative structures, outputting the postoperative remaining portions of the different ROIs. (Appendix S1 for more in-depth details).

2.6.4 White Matter Ablation Analysis

Differential tractography [28] using the February 8th, 2023, build of DSI Studio was used for the volumetric analysis of the different affected fibers associated with the ROIs (https://dsi-studio.labsolver.org/doc/gui_t3_dt.html) (Appendix S1 for more in-depth details).

3 Statistical Analysis

Ablation volumes of different ROIs were calculated using 3D Slicer. Statistical analyses were performed using SPSS (version 27); two-sided p < 0.05 was considered significant (95% CIs). Mann–Whitney tests and Spearman's rank correlation were used to examine associations between ablation volumes and outcomes. Fisher exact test was used for categorical data, and Wilcoxon Signed-Ranks tests for the analysis of SC and FC alterations in SF and NSF patients.

4 Results

5 Discussion

This study employed a voxel-by-voxel ablation mapping comparison, along with the structural and functional connectivity of the selected ROIs, to examine the impact that the ablation of different temporal lobe ROIs would have on the structural and functional brain connectivity relative to outcomes in a consecutive group of refractory MTLE patients who underwent SEEG-guided RFTC. We found that a better outcome was associated with a larger ablation volume of both the amygdala and rhinal cortex and with an alteration in structural connectivity networks. We also found no statistically significant alterations in functional connectivity network measures in either group.

Previous studies have shown a significant association between the extent of entorhinal and parahippocampal cortex resection and the outcome [17, 33]. Keller et al. studied 87 TLE patients (47 SF, 40 NSF) and found that SF patients tended to have a larger resection volume of the entorhinal cortex and amygdala compared to NSF [34]. Oren Sagher et al. reviewed 96 refractory MTLE patients; standard ATL was conducted in 51 nondominant TLE patients, while 45 language-dominant patients underwent transcortical SAH [35]. Two years post-surgery, 86%–90% of patients with ATL or SAH were seizure-free, with up to 90% of the amygdala, hippocampal, and entorhinal cortex removed, regardless of the technique. They pointed out that their study could not determine if successful seizure control required a specific resection extent as their patients had 90% or more of their mesial structures removed. It was impossible to evaluate the correlation between less extensive resection and a poorer seizure outcome. A recent study of 107 patients with unilateral TLE by Marian Galovic et al. found that removal of at least 50% of the piriform cortex was necessary for achieving seizure freedom after temporal lobe resections [18]. Removing other mesial structures wasn't associated with seizure control. In a recent study of 113 patients with TLE, Ezequiel Gleichgerrcht et al. reported that anterior hippocampus-amygdala-piriform or entorhinal-perirhinal resection and disconnection of the frontal, limbic, and temporal regions through white matter tract loss within the fornix, anterior commissure, and uncinate fasciculus improved the seizure outcome [36]. However, they could not ascertain if removing these sites alone or together improved surgical outcomes. Our results showed that increased ablation of the amygdala and rhinal cortex contributed to a positive outcome. However, as in other studies, we couldn't confirm if ablating both the amygdala and rhinal cortex alone provided surgical freedom or if their combined ablation with the other ROIs improved outcomes. Postoperatively, none of the white matter tracts examined using differential tractography demonstrated a significant correlation with seizure outcomes. The absence of correlation is probably due to the analogous and restricted degree of damage sustained by these tracts during RFTC in both cohorts.

Due to the significant importance of the piriform cortex [18, 37, 38], we examined the potential unintentional ablation of some parts during the thermocoagulation of the amygdala and rhinal cortex due to its closeness to these structures. The piriform cortex lies at the junction of the frontal and temporal lobes, medial to the temporal stem, and lines the superior and inferior banks of the rhinal sulcus in humans. It features a U-shaped cross section that curves around the middle cerebral artery in coronal slices. We manually carried out its segmentation using a protocol based on anatomical landmarks and correlation with histology (Appendix S1). We observed that none of its portions were affected by thermocoagulation, likely due to our reluctance to place electrodes close to the middle cerebral artery during preoperative planning.

Most studies have reported on the structural brain networks connectivity of temporal lobe epilepsy patients relative to healthy control groups using graph theory analysis, MRI data, and machine-learning algorithm [20, 23, 39-42]. But few have analyzed the SC alterations relative to outcomes [20, 23, 42, 43]. Taylor et al. reported on 53 TLE patients who underwent temporal lobe epilepsy surgery, and using graph theory and machine learning, they investigated the properties of change between the preoperative and predicted postoperative networks [20]. They found that seizure-free patients had greater reductions of connectivity. Bernhardt et al. examined 122 patients with drug-resistant TLE and 47 healthy controls, using MRI-based cortical thickness measurements [42]. Their study showed that an increase in clustering coefficient and path length was associated with decreased postoperative seizure freedom. Bonilha et al. reviewed 20 patients with refractory MTLE and a matched 18 healthy control group [23]. Their findings revealed that when compared to controls, patients had a decrease in ipsilateral thalamocortical connection and a pathological increase in medial temporal, insular, and frontal connectivity. Ko et al. reported on the seizure outcomes of 24 MTLE patients who underwent LITT and whether structural connectomes could predict surgical outcomes [43]. They employed preoperative diffusion tensor imaging (DTI) to simulate changes in structural connectivity following laser ablation. They found that favorable SC changes in connection strength were associated with an ablation volume involving relatively mesial, anterior, and inferior locations.

Our results are in line with these findings despite our differences in methodology. Both Taylor and Ko et al. used machine learning and preoperative DWI to generate the structural white matter network, along with the postoperative T1W to infer the impact of surgical resection on this network. In this study, we used both the pre and postoperative DWI along with their respective pre and post T1W to more accurately analyze the SC alterations following surgery; we also employed diffusion spectrum imaging (DSI) tractography instead of DTI due to its reported generation of fewer false connections compared with DTI [44]. We found a significant decrease in SC in the SF group, whereas the NSF showed no alterations. We theorize that despite the similar extent of white matter injury incurred by the fiber tracts in both groups after RFTC, the ablation of the different gray matter ROIs in the SF patients, notably the amygdala and rhinal cortex, which showed higher ablation volume in the SF compared to the NSF group, might have been significantly sufficient to positively disrupt the SC to achieve seizure freedom, as opposed to the NSF patients. The smaller ablation volume of the amygdala and rhinal cortex might have been the major contributor to the non-significant alteration in SC seen in the NSF group.

As for SC analysis, few studies have investigated the FC in patients with TLE both before and after surgery, specifically focusing on the differences in connectivity in SF and NSF patients, with most conducting comparative analyses of patients relative to healthy control groups [45-49]. Maccotta et al. reported on the FC of 17 patients with refractory TLE who underwent resective surgery with an Engel I outcome and a matching group of 17 healthy controls [45]. Their findings revealed that, following epilepsy surgery, the seizure-free state of their patients did not result in substantial alterations in functional network connectivity. The functional network anomalies observed after the surgery closely resembled the preoperative condition. They theorized that epilepsy has a chronic effect on brain connectivity that is “burned in” by the time a patient with intractable TLE has surgery, which usually occurs years after diagnosis. Morgan et al. reported on the FC of 20 patients (9 SF, 11 NSF) with refractory TLE who underwent resective surgery and a matching 44 healthy controls [46]. Their findings revealed that, when examining the impact of FC at the regional level, no significant differences in FC were seen postoperatively in the individual regional connections among the groups. The SF group showed no significant difference in both the pre and postoperative FC when compared to the control group, whereas the NSF group, which preoperatively exhibited no discernible distinction from the control group, demonstrated a lower FC compared to the control group postoperatively. Conversely, a study by Simula et al. has shown that thermocoagulation by LITT elicited FC alterations in the electrical brain activity of individuals with DRE for a duration of at least 15 min [49]. Functional connectivity decreased in responders (delta band only) and increased for the majority of frequency bands in nonresponders following RFTC.

Our findings align with those of Maccotta and Morgan et al., although they differ from the conclusions presented by Simula et al. In their analysis, Maccotta et al. employed graph theory analysis and network measures using both the pre- and postoperative T1W and BOLD from patients with a follow-up of 2 years. However, Simula et al. employed local power spectral density (PSD) to assess FC alterations in a 3-min segments recorded immediately prior to (baseline), immediately following, and 15 min after RFTC. Their analysis only assessed short-term alterations in FC in different types of DRE patients (not specifically in TLE patients). These different findings at a first glance seem to be contradictory; however, they may in fact suggest that while a short-term alteration in FC after thermocoagulation might be observed in DRE patients, in the case of TLE patients specifically, this alteration seems to revert back to the prior preoperative brain network state in the long term, probably due to the “burned in” effect theorized by Maccotta et al. (Appendix S1 for more in-depth discussion).

Ablation of most of the entire mesial temporal lobe complex may not be needed to positively alter the SC in MTLE. Minimally invasive treatments on key therapeutic ROIs through surgical destruction may interrupt the occurrence or propagation of epilepsy, thereby achieving cessation of epileptic seizures [50].

6 Limitations

The current findings should be taken in light of the study's limitations. We didn't include a healthy control group, which could have been of significant interest to describe a graph theory measure's baseline. Our sample size was relatively small, resulting in the combination of left and right MTLE patients into one group. Other limitations may include the potential biases of DSI, such as challenges related to spatial resolution, partial volume effects, crossing fibers that may not be fully resolved, and the reliability of tracts that are very small. Finally, the generalizability of these findings may not be applicable to other epilepsy populations due to our sole inclusion of MTLE patients.

7 Conclusion

Our study provides insight into some essential elements of brain connectivity networks in MTLE and might contribute to the generation of novel evidence that could potentially facilitate progress in surgical procedures in the future and improve SEEG-guided RFTC interventions in MTLE patients who choose this treatment option. Further research is required to substantiate the validity of our findings via a more extensive set of data.

Author Contributions

Stéphane Jean: conceptualization, methodology, software, formal analysis, writing – original draft, writing – review and editing, investigation, data curation, validation. Rifeng Jiang: data curation, supervision, software, formal analysis. Yihai Dai: project administration, resources, supervision. Weitao Chen: project administration, resources, data curation. Weihong Liu: supervision, data curation, project administration. Donghuo Deng: data curation, resources. Panashe Tevin Tagu: writing – review and editing. Xiaoqiang Wei: data curation. Shan Chen: data curation. Xinrong Fang: data curation. Shiwei Song: conceptualization, funding acquisition, validation, visualization, supervision, resources, data curation, project administration, investigation.

Ethics Statement

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.

Consent

This study was conducted under the ethical standards of the Declaration of Helsinki. It was approved by the Ethics Committee Board of Fujian Medical University Union Hospital (2024KY152), and informed consent was obtained from all participants and the parents or guardians.

Conflicts of Interest

The authors declare no conflicts of interest.