If the focus of seizures was confirmed to be in the temporal lobe and only in one hemisphere, patients underwent surgery. During the procedure, the temporal cortex was exposed and a 20-electrode grid was positioned covering the anterior temporal neocortex (Figures 2A,B). Also, a 6–8 electrode strip was placed between the tentorium cerebelli and the uncus/parahippocampal gyrus (arrow in Figure 2B). Electrocorticographic recording of spontaneous activity and after pharmacological stimulation with etomidate (to induce seizure-like activity) was carried out (Ortega et al., 2008a,b; Wix et al., 2008; Pastor et al., 2010b; Vega-Zelaya et al., 2014, 2016; Figure 2C). Based on the recording results, a tailored resection of the neocortex and amygdala-hippocampal complex was performed following Spencer’s technique (Spencer et al., 1984b) in two blocks (Figure 2D): (i) a resection from the tip of the temporal lobe sparing the superior temporal gyrus, followed by (ii) removal of approximately 2/3 of the anterior portion of the hippocampus and about 95% of the amygdala. After surgery a control MRI was performed (Figure 3).

Written informed consent was obtained from all participants in accordance with the Helsinki Declaration (World Medical Association, 2013). This study and all protocols received institutional ethics approval by the ethical committee at the “Hospital de la Princesa”. From the cohort of 260 epileptic patients receiving surgical treatment at this hospital, we selected 34 temporal lobe epileptic patients (according to EEG recordings), who suffered from partial, complex and secondarily generalized seizures (Table 1).

Histological analysis of the resected tissue revealed the presence of hippocampal sclerosis in 18 of them, whereas in the remaining 16 cases the hippocampal formation looked normal, not displaying histological alterations on standard histological preparations (e.g., Nissl staining) (Figure 4) or in sections immunocytochemically stained to visualize total neuronal populations (NeuN); GABAergic interneurons (GAD-1, GAT-1, PV, CB, CR, and others); or glial populations such as astroglia (GFAP) or microglia (Iba-1) (Arellano et al., 2004; Andrioli et al., 2007).

We classified hippocampal sclerosis based on ILAE (2013; Blümcke et al., 2013). ILAE Sclerosis type 1 refers to classic hippocampal sclerosis characterized by neuronal loss and gliosis both in CA1 and CA4-hilus, and to a variable degree in the rest of the fields. We distinguished between Sclerosis type 1 classic, when damage was present in CA1 with CA4 and the rest of the fields only presenting mild pathology, and Sclerosis type 1 extensive, when extensive loss of neurons and gliosis was found in all subfields including the DG. Cases showing an intermediate pattern of neuronal loss and gliosis will be referred to as Sclerosis type 1 intermediate. ILAE Sclerosis type 2 refers to neuronal loss and gliosis predominately in CA1, and ILAE Sclerosis type 3 exhibits predominant neuronal cell loss and gliosis in CA4-hilus (endfolium sclerosis).

In the pathological description, we also included the presence of alterations in the distribution of granule cells in the DG (dispersion, bi-layer pattern or patchy alterations), and the presence of corpora amylacea (spherical polyglucosan bodies with a diameter of around 10–15 μm) that are found mostly in the periventricular and subpial regions of the human brain during normal aging. Although the pathological significance of corpora amylacea is not yet clear, they are frequently described in some neurodegenerative diseases and also associated with TLE, particularly with hippocampal sclerosis (e.g., see Rohn, 2015; Riba et al., 2019, 2021).

The quantitative analysis of hippocampal histological alterations was based on Arellano et al. (2004) and included tissue from a group of epileptic patients that underwent TLE surgery, compared to tissue obtained by autopsy from three normal males (aged 23, 49, and 63 years) that died in traffic accidents (2–3 h post-mortem). Thickness of the granular layer, proliferation of mossy fibers, number of granule cells per 100 μm-wide column and neuronal densities of hippocampal subfields were quantified. As described in Arellano et al. (2004), neuronal densities were estimated using optical disectors on Nissl-stained sections by counting nucleoli (West and Gundersen, 1990). A Hitachi video camera attached to a Leitz microscope with a motorized stage were used to visualize the field on a computer screen and measure the position in the z axis. For the dentate gyrus, the number of neurons per 100 μm wide segment of the granular cell layer was assessed. The reason for estimating the number of neurons per 100 μm segment instead of neuronal density of granule cells is that most epileptic patients show granular cell dispersion that naturally will decrease the cell density, even if there is no neuronal loss. For this reason, adjacent optical dissectors (100× objective; 48 × 64 μm counting frame; 30 μm thickness) were used to count granular cell nuclei along the entire granular layer including dispersed cells. Results are shown as granule cells per 100 μm wide and 30 μm thick granular cell layer segment. Six segments were counted per patient. For the polymorphic layer of the dentate gyrus (hilus) and CA4, neuronal densities were estimated counting nucleoli in 9 optical dissectors (40× objective; 120 × 160 μm counting frame, 30 μm thickness). For fields CA3, CA2, CA1, and subiculum (this last one not reported in Arellano et al., 2004), neuronal densities were estimated by counting nucleoli in adjacent optical disectors (40× objective; 120 × 160 μm counting frame, 30 μm thickness) covering the entire stratum pyramidale from the stratum oriens border to the stratum radiatum border. Six measures were obtained per patient in CA3, CA2, and subiculum and nine measures were obtained from CA1 in each patient. Aberrant proliferation of mossy fibers was estimated by measuring the extent of the ectopic labeling of Dynorphin A in the DG molecular layer of epileptic patients. All these quantitative analyses were performed at the level of the hippocampal body (Figure 4). In addition, we used Nissl-stained sections to analyze the histopathology of resected extrahippocampal regions, including medial cortical regions adjacent to the hippocampus and lateral temporal cortex.

The psychological parameters for each patient were evaluated through three main blocks: the brief clinical interview, the neuropsychological testing and the study of personality. The specific variables recorded in each study are summarized in Table 2 and described below.

Data analysis and statistical comparisons of psychological variables between the groups were performed using McNemar, the one-factor and Mixed-design ANOVA and Chi-square (χ²) tests to compare the psychological variables for pre-post comparisons. Statistical analysis and graphs were performed using the SPSS statistical package (IBM SPSS Statistics for Windows, v24.0, IBM Corp., NY, USA) and GraphPad Prism 7 statistical package (Prism, San Diego, CA, USA).

We analyzed the patients’ histopathological and psychological data using machine learning by modeling the data as a supervised classification problem where the outcome of epilepsy surgery was the class label (patients with an Engel I versus patients with Engel II or III). The relevance of the clinical descriptors was assessed by a feature subset selection process over 10,000 random restarts (sci-kit-learn ML package). Highly ranked descriptors were discussed based on their predicted statistical influence on the surgery outcome. To this end, a wrapper feature subset selection process was coupled with a logistic regressor (L1 penalized) on a repeated hold-out schema. Features’ coefficients from the regression were filtered to find highly relevant subsets and their classification performance was estimated (Supplementary Table 5).

When reexamining the resected brain tissue from our sample of 34 TLE patients, we focused on patients with hippocampal sclerosis (HS-patients), since patients with normal-looking hippocampus (no neuronal loss and no gliosis) (No-HS-patients) did not show any apparent structural or pathological alterations other than those indicated in Table 1.

We did not find any type 3 sclerosis in our collection, and except for three patients classified as ILAE Sclerosis type 2, most patients (n = 15) were classified as ILAE Sclerosis type 1, of which n = 11 showed classic subtype, n = 3 showed extensive subtype and n = 1 was intermediate (Table 3). Detailed quantitative information for most of the patients is shown in Table 4 and Supplementary Table 1, including the thickness of the granule cell layer, extent of aberrant mossy fibers and percentages of neuronal density in all hippocampal fields and controls. As can be observed in Tables 3, 4, histological alterations were rather heterogeneous between patients. Interestingly, we found a consistent increase in neuronal density in the subiculum (except in one patient) as a consequence of the general atrophy of the hippocampal formation without significant loss of subicular neurons (Table 4). Some examples of the pathological changes found in most of the HS-patients are shown in Supplementary Figure 2.

A major GABAergic input to the soma (and proximal dendrites) and to the axon initial segment of principal cells originates from large basket cells and chandelier cells, respectively. These GABAergic subpopulations have been proposed to be critically involved in epilepsy (reviewed in DeFelipe, 1999). Thus, unless otherwise specified, in this section we will focus mainly on axon terminals innervating the soma and proximal dendrites (basket formations) and, in particular, on the axon terminals of chandelier cells (Ch-terminals) that innervate the axon initial segment. In the normal hippocampal formation, GABA transporter 1 (GAT-1)- and parvalbumin (PV)-immunoreactive (-ir) Ch-terminals are identified in the granular layer of the dentate gyrus, in the stratum pyramidale of the CA fields, and in the pyramidal layer of the subicular complex.

Data on the distribution and the neurochemical characteristics of Ch-terminals and basket formations in the DG and CA fields were available from Arellano et al. (2004) for 14 patients. These terminals were labeled using immunocytochemistry for GAT-1 and for the calcium binding proteins PV and calbindin D-28k (CB). As described in Arellano et al. (2004), two subtypes of Ch-terminals were observed with regard to their size and the density of terminals that were named simple and complex. Simple Ch-terminals were made up of one or two rows of labeled boutons and were the standard type found in controls. In contrast, some epileptic patients exhibited complex (or hypertrophic) Ch-terminals, consisting of tight cylinder-like structures made up of multiple rows of boutons. The distribution and neurochemical characteristics of the Ch-terminals and basket formations are shown in Supplementary Tables 2–4.

These observations reveal a conspicuous heterogeneity in the alterations of inhibitory circuits both between patients and within the same patient. The GABAergic alterations are apparently not associated with any particular cytoarchitectonic characteristic. Within the regions of severe neuronal loss, and despite the presence of some parvalbumin-immunoreactive cell bodies and some axonal processes, parvalbumin-immunoreactive Ch-terminals and basket formations were rarely found; but parvalbumin-immunoreactive dense Ch-terminals and/or basket formations were commonly found associated with some surviving neurons at the border of regions adjacent to the areas of neuronal loss.

We created a supervised classification dataset that used the Engel scale for surgical outcome as the classification class (Engel et al., 1993). Patients with an Engel I were encoded as the negative class, or class 0, and patients with Engel II or III were grouped together as the positive class, or class 1. All other clinical information including histopathology and psychological data was translated into 56 features for a total of 29 patients for whom we had enough information to perform this analysis. Features, average coefficients, and SHAP (SHapley Additive exPlanations)⁷ values for the highest ranked are included in the (Supplementary Table 5). In Figure 11, we sorted the features by the sum of their SHAP value magnitudes for all 29 patients. This figure presents the distribution of the influence each feature has on the model output, in this case the surgery output reflected by the Engel scale. The color represents the feature value (red high, blue low). This reveals, for example, that a high value for Hypervigilance Index (HVI) after surgery strongly influences the model towards an Engel I. Conversely, a high value for Suicide Constellation (S-CON) before surgery positively influences the model to an Engel II or III outcome.

There is a very extensive bibliography dealing with TLE, surgery, and psychology. What follows is not intended to be complete, and the unavoidable limitations in the selection of data and their interpretation reveal, in many cases, the personal views and interests of the authors.

The psychological studies performed by Scoville and Milner (1957) in patients with schizophrenia or epilepsy submitted to bilateral medial temporal lobe resection (including the well-known case patient H. M.) are of great interest not only from the surgical point of view but also for better understanding memory and cognitive changes that occur after surgery. In turn, these studies helped to provide important clues of how these parts of the brain are involved in memory and cognition (Squire, 2009b). The surgical procedure consisted in either a limited removal of the uncus and amygdala (uncotomy) or a larger resection which, in general, also included the anterior two thirds of the hippocampus and the parahippocampal gyrus and, in some cases, the temporal pole (the anterior end of the temporal lobe). When the hippocampus and parahippocampal gyrus were removed bilaterally, Scoville and Milner found a persistent impairment of recent memory, which seemed to be more pronounced depending on the extent of hippocampal removal. Memory loss included both anterograde and some retrograde amnesia, but left early memories and motor skills intact, and no deterioration in personality or general intelligence was observed. Milner and Penfield (1955) reported lack of permanent memory defects after unilateral removal of hippocampal and parahippocampal tissue in more than 90 cases of temporal lobe epilepsy. However, they also discussed two cases of unilateral partial temporal lobe resections in the dominant hemisphere (left) which resulted in an unusual severe loss of recent memory. After surgery, the general intelligence of these two patients was unimpaired and they continued with their jobs, but memory of daily events was very seriously damaged. That is, the surgical outcome was similar to those patients that had bilateral medial temporal lobe resections. Milner and Penfield concluded that in these two cases there was a severe lesion in the hippocampal region in the opposite hemisphere (right) that went unnoticed preoperatively. Thus, the unilateral removal of the hippocampal region in these two patients resulted in effect in loss of bilateral hippocampal function leading to impaired memory.

In general, deficient verbal memory is a common finding in left TLE patients indicating the involvement of the left hippocampus. In contrast, correlation between low visual memory scores and right hippocampal neuron loss or right temporal lobe surgery is not consistent between studies (e.g., Rausch and Babb, 1993; Wyler et al., 1995; Jones-Gotman et al., 1997; Shin et al., 2009). A particularly relevant study in this regard is the research performed by Jones-Gotman et al. (1997). These authors examined the general differential psychological outcome after the resection of the language-dominant left temporal lobe or the right temporal lobe, and compared the outcomes with different surgical procedures. They distinguished three groups of patients from different institutions that had undergone different types of surgical procedures: (1) anterior temporal-lobe resection including amygdala and the anterior part of the hippocampus; (2) temporal neocorticectomy, removing the neocortex and leaving intact the amygdala and hippocampus; and (3) selective resection of the medial temporal region, leaving intact the temporal neocortex. This comparison represents a good opportunity to study the involvement of the amygdala, hippocampus and temporal neocortex in learning and memory for verbal and visuospatial material. These authors found that learning and retention of words were impaired regardless of the structures removed from the left temporal lobe. However, they were unimpaired in patients submitted to right temporal resections. Learning for abstract designs was impaired in the right temporal lobe resections, but some less accentuated deficits were also observed on some trials in left-resection groups. A possible explanation given by Jones-Gotman et al. of these unexpected results is that the resection of either the hippocampal region or the temporal neocortex produces a disconnection of a circuit that is critical for the normal storage and retrieval of information. As shown in Boxes 4, 5, this conclusion is supported by the complex network of connections involved in the dialogue of the hippocampal formation with neocortical areas.

The main findings are six-fold when comparing TLE patients with or without hippocampal sclerosis (HS-patients and NoHS-patients, respectively). First, a significantly higher percentage of psychopathological alterations appears in HS-patients before surgery (HS-patients 94.4%; NoHS-patients 43.8%). After surgery, no statistically significant differences were found between groups but the proportion of NoHS-patients with psychopathological alterations increased. Second, the evaluation of the WAIS showed no differences between HS-patients and NoHS-patients; both groups had similar FSIQ and Digit Span values, except HS-patients that showed a significant lower performance on digit backward test. After surgery, the results were rather variable since some patients improved and others worsened in both the IQ and in Digit Span values or they showed changes in only one of these two tests. Third, the evaluation of the ROCF showed no significant differences between HS-patients and NoHS-patients when comparing either pre- or post-surgery. Most of the epileptic patients displayed high or very high values in the direct copy scores, and low or very low in the memory scores. Differences were not found to be statistically significant when comparing pre- and post-surgical conditions, although type II (poor planning) was more common in the HS-patients after surgery. From a clinical point of view, some HS-patients and NoHS-patients showed large drawing distortion and clinical or idiosyncratic casuistic differences before and/or after surgery. Fourth, although all patients showed deficits in episodic memory, neuropsychological testing revealed significant differences between NoHS-patients and HS-patients, the latter being worse in the logical and visual memory test (WMS). However, no differences were observed when comparing before and after surgery conditions. Fifth, the Rorschach test revealed significant differences between HS-patients and NoHS-patients with regard to some of the 10 variables analyzed. Before surgery, NoHS-patients displayed a higher suppression and constraint of emotion (SumC’). In all patients, the surgery intervention produced certain psychological changes: an increase in obsessive (OBS) and hypervigilant style (HVI) and a reduction in negative introspection (SumV). After surgery, NoHS-patients showed a significant decrease in the perceptual and thought disorder (PTI), as well as in the negative, ineffective and/or non-adaptive responses related to human relationships (PHR > GHR). Changes in personality style were observed after surgery in some HS-patients and NoHS-patients, with some of these changes being positive while others were negative. Sixth, the House-Tree-Person projective analysis revealed that most patients (both HS- and NoHS patients) showed affective, cognitive and motor difficulties when completing the drawings, both before and after surgery.

Despite the limited number of patients examined, the multidisciplinary combination of pathological condition (sclerotic condition), clinical interview, psychological test and machine learning tools together provide an important improvement in the criteria for selecting candidates for epilepsy surgery and to predict the outcome after surgery. We found machine learning approaches a significant aid in predicting the outcome of epilepsy surgery. Based on supervised classification data mining (as discussed in Armañanzas et al., 2013) machine learning protocols can easily take into account clinical variables as well as pathological and psychological evaluations. In the present study, the SHAP influence coefficients (see Figure 11) indicate that Hypervigilance Index (HVI) after surgery, Suicide Constellation (S-CON) before surgery, Detection of sclerotic tissue (Sclerosis), and Depression (DEP) from the clinical interview before surgery were the more relevant features to predict the outcome. We found that a high value for HVI after surgery as well as the presence of hippocampal sclerosis (Sclerosis) strongly influences the model towards an Engel I. DEPI item also influenced the model towards an Engel I, but with less impact than HVI and Sclerosis indexes. Conversely, a high value for S-CON before surgery deviated the model to an Engel II or III outcome.

The improvement of certain cognitive functions after surgical excision of epileptogenic regions in some patients is well-known since the early epilepsy surgeries. This is most probably explained by the dense inter-connectivity of the structures involved that allows the spread of epileptogenic activity to otherwise normal functioning areas, producing a general depression of function. This explanation was already drafted in the early 19th century theories about remote lesion effects and recovery of function. For example, regarding epileptic patients, Hughlings Jackson wrote in 1884: “destruction of a small part of the highest centers produces very little effects, whilst sudden and excessive discharges of such small part produces, although indirectly, an enormous effect…. It is notorious that a small part of the brain—a small part of the highest centers— may be destroyed without any striking symptoms; compensation is practically perfect.” (Jackson, 1884). Later, at the beginning of the 20th century, Constantin von Monakow developed a theory he called diaschisis to describe that injury in one brain area may induce significant decrease in activity in functionally connected brain areas, and how if the disruption eventually disappear, give rise to some recovery of function (Finger et al., 2004). In 1914, Monakow explained this theory as follows: “…an ‘interruption of function’ appearing in most cases quite suddenly… and concerning widely ramified fields of function, which originates from a local lesion but has its points of impact not in the whole cortex (corona radiata, etc.) like apoplectic shock but only at points where the fibers coming from the injured area enter into primarily intact grey matter of the whole central nervous system” (text taken from Finger et al., 2004). Similarly, Wilder Penfield proposed in the 1940s that these improvement of cognitive functions after surgery might be explained by a reduction or elimination of the propagation of the epileptiform discharges from the epileptogenic region or “nociferous cortex” as named by Penfield, to normal regions (see Hebb, 1977; Hermann and Seidenberg, 1995 and below sections).

Neuropsychological studies performed by some authors of the present work have shown improvement of intellectual functions after surgery in the medial temporal lobe (Martín et al., 2002). In the present study we report a similar tendency but with higher variability. The working memory differences in HS-patients found in this study before and after surgery could be indicating that the relief from epileptogenic activity induces an improvement of distant brain circuits, where the spread of the epileptogenic activity (hippocampal-prefrontal circuits) was inducing malfunctioning previously. It is important to indicate that there is a relationship between the performance on IQ test and executive functions tests in epileptic patients (Helmstaedter et al., 2019). As shown below, Hermann and Seidenberg explored the interesting example of the poor preoperative performance of the Wisconsin Card Sorting Test (WCST) that occurs in some patients with TLE. This test measures abstraction ability and cognitive flexibility that is the ability to shift from one mode of solution to another on a sorting task. Milner (1963) reported greater deficits in patients with frontal injuries than in patients with posterior brain damage and since then this test is widely used to assess possible deficits in prefrontal or executive function in humans. Hermann et al. (1988) observed in some TLE patients a poor performance on the WCST that improved (or at least did not worsen) after surgery, indicating that this was not due to temporal lobe dysfunction. Instead, they proposed that the poor preoperative WCST was probably due to the propagation of “neural noise” from the temporal lobe/hippocampal epileptic focus “via pathways which link the anterior temporal lobe and hippocampus with frontal areas.” Other researchers confirmed later the poor performance of the WCST in TLE patients (see Hermann and Seidenberg, 1995). However, Corcoran and Upton (1993) proposed that the hippocampus is directly involved in some of the executive functions that are measured with the WCST⁸. Finally, Hermann and Seidenberg (1995) compared the nociferous cortex hypothesis with the hippocampal hypothesis to explain poor WCST in TLE patients⁹ and concluded that impaired preoperative WCST performance that is seen in some TLE patients is not directly caused by hippocampal damage, but by the functional disruption of extratemporal regions, supporting the nociferous cortex hypothesis.

There are multiple sources of interindividual variability in normal brains, and even more so in those with pathological conditions. In this regard, we should first consider that there is high interindividual variability between epileptic patients not only due to obvious factors such as differences in medical treatment, onset of the disease and pathological substrates, but also because the normal human brain show large interindividual variability in many of its structural and functional attributes. For example, there is pronounced inter-subject variability with respect to brain size, shape and cytoarchitecture (e.g., Amunts et al., 1999, 2005; Uylings et al., 2005). At the microanatomical level, there are variations in the structure of cortical neurons (e.g., Jacobs et al., 1993, 1997; Benavides-Piccione et al., 2021) and in the number and density of cortical synapses (e.g., Alonso-Nanclares et al., 2008; Montero-Crespo et al., 2020; Cano-Astorga et al., 2021; Domínguez-Álvaro et al., 2021). Some of these variations have been attributed to differences in age, sex or education. There are also variations in terms of functional connectivity which has been associated to certain anatomical features such as cortical folding, brain size, amount of gray or white matter and size of different brain regions, as well as with behavioral and demographic variations and other factors (e.g., Mueller et al., 2013; Caspers et al., 2014; Demirtaş et al., 2019; Llera et al., 2019; Bethlehem et al., 2022; Marek et al., 2022). Second, “hippocampal sclerosis” is not a homogeneous entity. This pathology encompasses multiple patterns of neuron loss and gliosis as well as a variety of changes in the expression of numerous molecules in the surviving cells and axonal reorganization including both excitatory and inhibitory axons (see section “GABAergic circuits are altered in temporal lobe epilepsy” and Box 3). Thus, hippocampal sclerosis is associated with a variety of functional alterations, which contributes to interindividual variability. As we will further discuss in the next section, these features made challenging the interpretation of the neuropsychological effects based on surgical procedures and pathological findings.

There are numerous examples in the literature that report either similar or different conclusions regarding the impact of surgery in the neuropsychology of the TLE patients and the presence or not of hippocampal sclerosis. The differences between studies may be explained by multiple factors, including different clinical characteristics of the patients, extent of brain tissue surgically removed and types and protocols of neuropsychological measures. From the structural point of view, it seems obvious that the neuropsychological differences found between TLE patients may be related to their different pathological conditions. Namely, the type of pathology (for example, tumors, cortical atrophy, hippocampal sclerosis) and whether it is bilateral or unilateral and whether more than one pathology exist. Based on hippocampal connectivity (Boxes 4, 5), it seems obvious that even in TLE cases with unilateral hippocampal sclerosis, differences in neuronal loss in one or various hippocampal regions (DG, CA1-CA3 fields and the subicular complex) between TLE patients might have important functional consequences. The multiplicity of pathways involved could reflect the multiplicity of cognitive functions. It also indicates the various direct and indirect pathways that might contribute to these functions, and thus be differentially affected in pathological conditions, which a priori seems to be more directly relevant when comparing NoHS-patients with HS-patients. For example, the entorhinal-hippocampal system together with the perirhinal cortex and parahippocampal cortex represent key brain circuits involved in the process of learning and memory (e.g., Squire and Zola-Morgan, 1991; Milner et al., 1998; Brown and Aggleton, 2001; Amaral and Lavenex, 2006; Squire, 2009a; van Strien et al., 2009; Strange et al., 2014; Squire and Dede, 2015; Kitamura, 2017; Nilssen et al., 2019; Marks et al., 2021; Witter and Amaral, 2021). These brain regions have been proposed to be differentially involved in multiple memory systems (for reviews, see Tulving and Schacter, 1990; Milner et al., 1998; Tulving and Markowitsch, 1998; Squire, 2009a; Squire and Dede, 2015). Since TLE patients may have different pathological characteristics, this would explain the wide range of differences between patients with regard to memory attributes¹⁰.

Another critical aspect regarding hippocampal connectivity, which is often overlooked in studies of TLE, is concerning the connections of the hippocampus with the frontal cortex. It is important to point out the strong direct projections of CA1 to orbito- and medial prefrontal cortex (in macaque: Barbas and Blatt, 1995; Cavada et al., 2000; Ichinohe and Rockland, 2005; Zhong et al., 2006; review: Sigurdsson and Duvarci, 2015). Since neuronal loss in CA1 is the most typical pathological condition in patients with hippocampal sclerosis, all these patients might show deficits in frontal lobe functions.

Connections of the hippocampus with the frontal cortex might be particularly important regarding the digit backwards, H-T-P and ROCF tests. It is pertinent to emphasize the major differences between both tests. In the H-T-P test, the patient is asked to draw a full-body person and write a name and age. Thereafter, they were asked to do the same for a person of the opposite sex (Figures 8, 10). Thus, the patient is projecting a distorted mental image of the human body. However, in the ROCF test, the patient is asked to copy the figure (for the evaluation of visuospatial constructional ability) and three minutes later (in the present study) they are asked to draw what they remembered (for the evaluation of nonverbal visuospatial memory) (Shin et al., 2006; Figures 5, 6A,B). Consequently, patients with lesions in the hippocampal formation in the right temporal lobe are expected to do worse on the visuospatial memory test than patients with lesions on the left hippocampal formation. Furthermore, as pointed out by Shin et al. (2006), copying the ROCF card is a very complex cognitive task “which involves the ability to organize the figure into a meaningful perceptual unit. Therefore, the ROCF is considered to be a useful tool for the evaluation of frontal lobe function, which is required for strategic planning and organizing.” This is in line with the well know functional interplay between the hippocampus and the prefrontal cortex (for reviews see Milner et al., 1985; Buckner et al., 1999; Laroche et al., 2000; Shu et al., 2003; Sigurdsson and Duvarci, 2015).

Finally, it is also important to consider the amygdala in the interpretation of psychological assessment. The necessary surgical removal of the amygdala in several of the surgical protocols is likely to have an impact on the cognitive performance of the epileptic patients. Among the most significant cognitive functions associated with the amygdala are those linked to the strong interactions with the prefrontal cortex. Particularly relevant to the present results, the amygdala has been implemented in memory-related functions concerned with the emotional significance of external and visceral stimuli (for recent reviews; Palomero-Gallagher and Amunts, 2022; see Dixon and Dweck, 2021). These facts should also be taken into consideration when evaluating the post-surgical psychopathological aspects of these epileptic patients.