Looking for complexity in quantitative semiology of frontal and temporal lobe seizures using neuroethology and graph theory

Epilepsy & Behavior 38 (2014) 81–93

Contents lists available at ScienceDirect

Epilepsy & Behavior journal homepage: www.elsevier.com/locate/yebeh

Looking for complexity in quantitative semiology of frontal and temporal lobe seizures using neuroethology and graph theory Poliana Bertti a,b, Julian Tejada a,c,1, Ana Paula Pinheiro Martins b, Maria Luiza Cleto Dal-Cól a,b, Vera Cristina Terra b,2, José Antônio Cortes de Oliveira a, Tonicarlo Rodrigues Velasco b, Américo Ceiki Sakamoto b, Norberto Garcia-Cairasco a,b,⁎ a b c

Neurophysiology and Experimental Neuroethology Laboratory, Physiology Department, Ribeirão Preto School of Medicine, University of São Paulo, USP, Ribeirão Preto, Brazil Epilepsy Surgery Center, Department of Neuroscience and Behavioral Sciences, Ribeirão Preto School of Medicine, University of São Paulo, USP, Ribeirão Preto, Brazil Physics Department, Ribeirão Preto School of Philosophy, Science and Letters, University of São Paulo, USP, Ribeirão Preto, Brazil

a r t i c l e

i n f o

Article history: Accepted 25 July 2014 Available online 10 September 2014 Keywords: Seizure semiology Clinical manifestations Behavioral sequences Complexity Graph

a b s t r a c t Epileptic syndromes and seizures are the expression of complex brain systems. Because no analysis of complexity has been applied to epileptic seizure semiology, our goal was to apply neuroethology and graph analysis to the study of the complexity of behavioral manifestations of epileptic seizures in human frontal lobe epilepsy (FLE) and temporal lobe epilepsy (TLE). We analyzed the video recordings of 120 seizures of 18 patients with FLE and 28 seizures of 28 patients with TLE. All patients were seizure-free N1 year after surgery (Engel Class I). All patients' behavioral sequences were analyzed by means of a glossary containing all behaviors and analyzed for neuroethology (Ethomatic software). The same series were used for graph analysis (CYTOSCAPE®). Behaviors, displayed as nodes, were connected by edges to other nodes according to their temporal sequence of appearance. Using neuroethology analysis, we confirmed data in the literature such as in FLE: brief/frequent seizures, complex motor behaviors, head and eye version, unilateral/bilateral tonic posturing, speech arrest, vocalization, and rapid postictal recovery and in the case of TLE: presence of epigastric aura, lateralized dystonias, impairment of consciousness/speech during ictal and postictal periods, and development of secondary generalization. Using graph analysis metrics of FLE and TLE confirmed data from flowcharts. However, because of the algorithms we used, they highlighted more powerfully the connectivity and complex associations among behaviors in a quite selective manner, depending on the origin of the seizures. The algorithms we used are commonly employed to track brain connectivity from EEG and MRI sources, which makes our study very promising for future studies of complexity in this field. This article is part of a Special Issue entitled “NEWroscience 2013”. © 2014 Elsevier Inc. All rights reserved.

1. Introduction In patients with TLE, the most common form of focal epilepsy in adulthood, previous semiological studies of epileptic seizures using neuroethological tools have shown great potential to reveal localizing and lateralizing signals, such as the presence of epigastric aura, the lateralization value of dystonias, the impairment of consciousness and ⁎ Corresponding author at: Neurophysiology and Experimental Neuroethology Laboratory, Physiology Department, Ribeirão Preto School of Medicine University of São Paulo, Avenida Bandeirantes, 3900, Ribeirão Preto, São Paulo 14049-900, Brazil. Tel.: +55 16 3602 3330. E-mail address: [email protected] (N. Garcia-Cairasco). 1 Current address: Psychology Department, Center of Education and Human Sciences — CECH, Federal University of Sergipe, UFS, Aracaju, Brazil. 2 Current address: Serviço de Epilepsia e EEG, Hospital de Clínicas, Federal University of Paraná. Epicentro, Hospital Nossa Senhora das Graças, Curitiba, PR, Brazil.

http://dx.doi.org/10.1016/j.yebeh.2014.07.025 1525-5050/© 2014 Elsevier Inc. All rights reserved.

speech during ictal and postictal periods, and the development of secondary generalization [1,2]. The neuroethology–SPECT correlation in TLE was an effective tool to reliably evaluate ictal behavior and the functionally associated brain areas. However, our data did not confirm the association of ipsilateral basal ganglia hyperperfusion with contralateral dystonic posturing, as described in the literature. Nevertheless, we demonstrated that ipsilateral basal ganglia hyperperfusion is associated with contralateral upper limb automatisms and also with the lack of contralateral cephalic version [1]. The second most common form of focal epilepsy, frontal lobe epilepsy (FLE), represents approximately 20% of patients admitted to epilepsy surgery programs [3]. The diverse spectrum of ictal behavioral phenomenology of patients with frontal lobe seizures has received far less attention than that of those with temporal lobe seizures. The typical clinical manifestations includes contralateral clonic movements, unilateral or bilateral tonic motor activity, as well as complex automatisms [4]. In

82

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

order to better define the localization and limits of the epileptogenic zone (EZ), various diagnostic tools such as semiological seizure analyses, electrophysiology, and neuroimaging are used [5]. Altogether, these important clinical, electroencephalographic (EEG), and neuroimaging data may help to define which cerebral circuit or network is responsible for the seizure generation while also evidencing its propagation and guiding potential surgical treatment and outcome. Semiological analysis of clinical seizures has been well documented by many authors [6–13], and the highly heterogeneous behavioral repertoire suggests the activation of variable brain regions and the spreading of their activation to adjacent areas. Some of these studies (mostly about TLE) have intended to standardize the terminology [2,14,15] or to quantify human epileptic seizure behaviors [2,14–17]. A detailed analysis of the ictal semiology can often provide lateralizing or localizing signs that disclose valuable information about the location of the EZ and the pathways potentially involved in seizure propagation [18]. Methodologies based on the description of seizures by the percentages of signs and symptoms may be found in various reports [6,8,9,19]. Wieser [13] pioneered the use of cluster analysis to correlate groups of signs and symptoms and to verify the sequence of behaviors occurring during complex partial seizures in patients with TLE. Subsequently, Kotagal et al. [10,16] also applied cluster analysis to characterize temporal and frontal lobe epileptic seizures. Manford et al. [17] combined cluster analysis and flowchart representation to differentiate TLE from FLE and to correlate the topography of MRI lesion with ictal behaviors. An interesting aspect of the latter report was the seizure representation as flowcharts, with ictal behaviors displayed in a sequential way and the temporal progression of the seizures represented by arrows between behaviors. At some point, this is similar to what we have used already (see below) with seizure analysis in TLE [2]. Neuroethology is a combination of ethology – the comparative study of behavior – and neurophysiology or neurobiology — the study of central nervous system functioning. Neuroethological analysis, from flowcharts built based on frequency, duration, and interaction between behaviors, has been successfully applied and validated in animal models of epilepsy by Garcia-Cairasco and co-workers since 1983 [20–22]. Dal-Cól et al. [2] applied this neuroethological method to a highly selected group of patients with mesial temporal lobe epilepsy (mTLE) for the first time, revealing some localizing and lateralizing signs, such as the presence of epigastric aura, lateralization value of dystonias, and impairment of consciousness and speech during ictal and postictal periods. One advantage of this method is the possibility to analyze all behaviors developed by the patient during the entire seizure, whereas in other methods, the analysis is restricted to only one behavior or to a group of behaviors [8,10,23]. The neuroethological analysis applied to mTLE in humans [2] was further correlated with SPECT findings [1] to evaluate which cerebral areas were or were not recruited during seizures. Although the power of such neuroethological studies is high, we are far from characterizing behavioral sequences with measurements that are nonlinear quantifications of the complexity associated with the expression of epileptogenic brain circuits. This, in fact, has been applied with much more frequency to functional imaging/EEG connectivity [24,25]; however, there is no single study, as far as we know, proposing such evaluations to semiology data, even though this is the final common pathway of brain activity. For that reason, more recently, we have been exploring the use of graph theory [24,25] as another neuroethological analysis method [26] because of the variety of available software, the huge amount of different measurements that can be calculated, and also the widespread applicability of these methods to neuroscience and brain pathologies (including the epilepsies) and their diagnostic tools [27]. Based on the previous findings of Dal-Cól and colleagues [2] and Bertti et al. [1], as well as on the widely described features of TLE and FLE seizures, the main objectives of the present study were to apply and to validate neuroethological methods, flowchart and graph

representation of the analysis of preictal, ictal, and postictal signs and symptoms of patients with FLE compared with those with TLE. 2. Methods 2.1. Patients We retrospectively studied a group of patients successfully operated on (Engel Class I; [28]), with previous pharmacoresistant FLE (120 seizures) or TLE (28 seizures). Subjects underwent presurgical evaluation and surgery between 1997 and 2006 at the Epilepsy Surgery Center of the Ribeirão Preto School of Medicine (CIREP/FMRP), University of São Paulo, Brazil. All patients signed an informed consent form, allowing the use of images for research purposes, following recommendations and approval of the Ethics Commission of the Institution (protocol 13528/2010 and 782/1998). The presurgical workup included high-resolution MRI acquired from a 1.5-T Siemens Magneton Vision (Erlangen, Germany) machine; longterm video-EEG monitoring; ictal and interictal SPECT; and neurological, psychiatric, neuropsychological, and socioeconomic evaluations. During video-EEG, medication was either tapered or discontinued. Patients with FLE were selected according to the following criteria: (1) medical history and seizure semiology consistent with FLE, diagnosed by the presurgical workup, (2) presurgical seizures acquired during the video-EEG monitoring at our institution, and (3) more than 6 years of age. Patients with TLE were selected according to the following criteria: (1) medical history and seizure semiology consistent with mTLE, (2) unilateral interictal epileptiform discharges over the anterior and mesial temporal regions, (3) presence of hippocampal sclerosis and no other lesion on MRI, and (4) ictal and interictal SPECT scans. Patients with abnormal neurological and neuropsychological examinations suggesting other brain diseases were excluded. For more details, see [1]. 2.2. Video recording and analysis Video-EEG recordings were performed on a digital EEG system (Vangard System, version 9.1, Cleveland Clinic Foundation) through scalp electrodes placed according to the International 10–20 System of Electrode Placement and through additional frontal intermediate electrodes of the International 10–10 System. Hewlett-Packard workstations (Model 715/64) were used for EEG data acquisition and analysis. Video images were obtained through Panasonic WV-GL704 video cameras and recorded on a Super-VHS Panasonic AG5700 videocassette player. Video editions were performed on a Super-VHS Panasonic AGA96 videocassette player. All videos were captured and digitalized through a video card Pinnacle DC10plus®, Studio 8® software, and observed in an Intel® Core™ i5 computer using Virtualdub 1.4d or Windows Media Player. Only the videotaped seizures with clear image and sound were evaluated and submitted to neuroethological analysis. Behaviors were identified according to a previous TLE semiological dictionary [2]. New behaviors presented by patients with FLE were added to this dictionary (see Supplementary Table 1 with glossary). Seizure onset and termination were defined based on semiological data obtained during seizure analysis. Electroencephalography time marks were considered only when the clinical seizure onset or seizure termination was not clear by semiology. Clinical seizure onset was defined as the time when the patient indicated the occurrence of an aura by pushing the seizure alarm button (or attempting to do so) or when the first unequivocal behavioral change was observed. Clinical seizure termination was defined as the relaxation phase immediately after a secondarily generalized seizure or the recovery of consciousness or ending of tonic or dystonic postures for partial seizures. Each seizure was observed as many times as necessary, including frame-by-frame analysis. Behaviors were mutually exclusive, i.e., each moment corresponded to only one behavior. All seizures were analyzed

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

by one of the authors (P.B.), who was blind to patients' clinical data but aware of their FLE and TLE. Some of the videos were reviewed together with other experienced clinical investigators (V.C.T. and A.P.P.M.) well informed about all patients' clinical data.

83

(frequency and duration), behavioral sequence, and statistical interactions (when they occurred), using a calibration pattern (Fig. 1A) modified from Dal-Cól et al. [2]. We also evaluated, 1 min before the onset of seizures, the preictal period and, 1 min after the end of seizures, the postictal period. In the flowcharts of both FLE (Figs. 1B, C, 2A, B) and TLE (4 A–B) seizures, each behavior is displayed as a rectangle, where the base represents the duration and the height represents the behavioral frequency during the observational period. The width of colored arrows between pairs of behaviors reflects the statistical interaction between them (χ2 analysis). Rectangle and ellipse colors identify groups of behavioral categories (clusters) and have only graphic and qualitative values. All interference by staff was recorded whenever possible, including the patient's answer or lack thereof and incorporated into the glossary of behaviors (Supplementary Table 1) and, obviously, into the behavioral sequences. In previous publications from our laboratory, we described the quantitative semiology of TLE with neuroethological tools [1,2] using flowcharts. Then, using the same dataset used for the construction

2.3. ETHOMATIC software, flowchart construction, and graph theory Behavioral sequences of each of the seizures underwent statistical analysis with ETHOMATIC software [21], which provided the frequency and the mean duration of each behavior and the statistical interaction between behavioral pairs (dyads). This interaction is calculated through a first-order transition matrix by the number of interactions (number of times that a behavior follows another given behavior). The glossary of the whole set of behaviors presented in all the observed videos from both FLE seizures and TLE seizures is presented in Supplementary Table 1. After the analysis of the individual seizures, all data were graphically represented using CorelDRAW® Graphics Suite X5 (Corel Corporation). Flowcharts were built for the sum of seizures, indicating behaviors

A

B

C

1

2

1

VC+ A

SNEZ

TV

G RI M

NOSP

NEFL EDVD

CDL

SMIL CDR

2

NUS

ECLO

EYOP

AL UO

ABUO ARUO

WAKF

NMRU SHLU NMLU

INTO INT C

LOS P FWLU FWRU

LHCH DPBR EDVD

BEAR LYDO

INCT

NUS

BHLU

EDVL

ALL

LOOB

EDVD 2HF

ALUO

CNOM NOSP

OACH

VEU

ALU

DGIN INCO

LOSP

HV

DEGL

BEAR

O ACH

ARUO

ABU ARU

NMRL

INAN

CAQ VC -

LOAR

SEND M OAN

VC+ A BEB NUS

NOSP

ACCO

CDL

CDR

NEFL DGIN

EDVR

SONS

VC+S

SMIL

SPECT

ECLO LOAR

EDVL

3

INAN

VC -

CAQ

BEAR

LHCH NMBU

BEB

CDL

OA LOAR

VC+ S VC+ A

RHCH

EDVL DPBR

CFEX

INAN

3

ARU

ALUO

NMBL

VELU

Fig. 1. (A) Calibration patterns for flowchart construction (modified from Dal-Cól et al. [2]) to study ictal human behaviors in both TLE and FLE. The height of the rectangles corresponds to behavioral frequency, and the base corresponds to behavioral mean duration. Statistical interactions between pairs of behaviors (dyads) are proportional to the width of the arrows that link them (χ2 ≥ 3.841; log χ2 ≥ 0.25; p b 0.05). (B) Flowcharts with the sum of four FLE seizures from the patient ACE and (C) flowcharts with the sum of four FLE seizures from the patient LBS: (1) preictal period, (2) ictal period, and (3) postictal period. For other behavioral details, see the text. For acronym description, see Supplementary Table 1.

84

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

A 1

2

INAN CDR

EDVD EDVR

SONS TDVR PRON

ALU

LOS P NEFL

EYOP

B 1

3 TDVL

INTW

HAH HARU

HARL

INAN VC INCO

LOOB

NUS

LOS P DEGL

TPNE CLRU

CNOM

MOAN

SEND

OACH

ATRU DLU

VC+ F

EDVD

THI R

OA

HI PR ARU

NOSP

TDVR

DPBR

ABU

INCS VC+ A

LOAR CDR

CAQ

BEAR

DLL ABL

INT O

INCS

YELL

LOAR

INTC

INCT INCO

LOOB

AC SLEP

VC -

CDL

RHCH LHCH

OA

ACCO BEAR

SITT

ARUO

TLU

TFBL

WAQ

TRCO

SLHR

O ACH HIPR LYDO

JBHA

ALUO

2

3

Fig. 2. (A) Flowcharts with the sum of four seizures from the patient WFA and (B) flowcharts with the sum of eighteen seizures from the patient BGC: (1) preictal period, (2) ictal period, and (3) postictal period. For other behavioral details, see the text. For acronym description, see Supplementary Table 1.

of flowcharts, we converted the form of the flowcharts into a graph. Therefore, Fig. 3 (FLE) and Fig. 4 (C–F, TLE) are flowchart-like graphs for both types of seizures. The idea was to maintain the spatial representation of the flowcharts in these figures in order to build graphs. We used CYTOSCAPE® [29,30] to create and to analyze the graphs, choosing some of the most used measurements [31,32] for multivariate measurements for nodes such as Degree, Clustering Coefficient, and Betweenness Centrality; for measurements for edges such as Edge Betweenness; and for the networks such as Small-World-Ness Index (SWI) and Hub Identification. For additional information about these measurements and graph construction, see Section 2.4 (below), Supplementary Table 2 (nomenclature), and Supplementary Fig. 1 (rules and patterns). In addition to the expression of the behavioral sequences of the FLE and TLE seizures in flowcharts, we explored some examples of graphs using different criteria: sum of the seizures of a specific patient with FLE (Figs. 1B and C, 2 and 3) and sum of the seizures based on the side of the surgery (left versus right) with TLE (Fig. 4). A flowchart can be considered a graph, because it graphically represents connections between events, with a specific metric to determine the size of rectangles (nodes) and arrows (edges). Our transition started by simplifying the dimensions of the nodes, transforming them from flowchart rectangles to circles, and discarding the temporal information included in the width of the rectangles (Figs. 3A–C). The second step was to introduce graph theory measurements into the flowchart template, with the size of the nodes being proportional to the Cluster Coefficient (Fig. 3B) and the Betweenness Centrality (Fig. 3C), and with the edge width, in all cases, being representative of the Edge Betweenness (Figs. 3B and C). The processes described above were performed with FLE flowcharts (Figs. 3B, C) and also with TLE flowcharts (Figs. 4C–F). In the latter case starting directly from the graph in which the size of the nodes represents the Cluster Coefficient (Figs. 4C–D) and the Betweenness Centrality (Figs. 4E–F). Similar to Fig. 3, the edge width represented the Edge Betweenness in Figs. 4C–F.

2.4. Statistical analysis and characterization of flowcharts, SWI, and hubs For the characterization of behavioral sequences in the flowcharts, the statistical value of each dyadic interaction (pair of behaviors) was calculated by a chi-square test (χ2). Interactions were considered significant if χ2 ≥ 3.84, log χ2 ≥ 0.25; p b 0.05. The ETHOMATIC software allows the analysis of just one seizure or the sum of seizures from an individual patient or a group as a whole. For network dynamics, we made the calculation of the SWI [33], which indicates the amount of formation of Small-Worlds [34] or highly clustered events usually interconnected throughout short paths. We also characterized the presence of hubs by classifying the degree of a node as highly significant and, therefore, defined as a hub, when its value was outside of the 95% one-tailed confidence interval calculated from the degree of all nodes using a nonparametric distribution. 3. Results 3.1. Demographic data All the information related to the demographic data from patients with TLE is detailed in Bertti et al. [1]. Briefly, among the 28 patients studied, 13 patients (8 men and 5 women) had mTLE on the left side and 15 (8 men and 7 women) on the right side. Eight patients had a history of problems in the prenatal period or during delivery, and 13 had a family history of epilepsy (6 with left-sided epilepsy and 7 with right-sided epilepsy). All patients became seizure-free after surgery (Engel 1A). One seizure per patient was scanned, observed, and recorded. For patients with FLE, 13 males and 5 females met the inclusion criteria. The mean postoperative follow-up was 5 years, and nine (50%) adults, three (16.6%) teenagers, and six (33.3%) children older than six years of age were all seizure-free after surgery (Engel 1A). One hundred twenty seizures were scanned, observed, and recorded,

Fig. 3. Different visual representations of the complexity inherent to the semiology of FLE, in which a new graphical layout of Fig. 2B is made — flowcharts with the sum of eighteen seizures from the patient BGC showing the (1) preictal, (2) ictal, and (3) postictal periods, with the same colors and positions, in which the size of the nodes (circle diameter) is proportional to the (A) Frequency of Behaviors, (B) Clustering Coefficient, and (C) Betweenness Centrality. The edge width represents the (A) χ2 log and the (B and C) Edge Betweenness. Calibration patterns (sizes and colors) are at the bottom of the figure. For acronym description, see Supplementary Table 1.

A

1

2

B

1

2

3

3

TRSE

EYOP LOSP

VC+S

ATRU

MOAN

LOSP

SITT ATC

SONS

BEAR ARL

2

EYOP

LOSP

ECLO

3

TAAR

TARU

VEU

MCCL

10-19 0-9

Arrow Width 2 4 6 8

Cluster Coefficient (CC)

20-29

Frequency

30-39

1.0 0.7

TBU

SITT

RELX

GTCC

GTC

VC NUS VC+A

EDVD HV

INAN

LOSP

NOSP OA

ACCO

INCO

STUP HIPR

ALUO

TFRU

VOCA

ETUL

CRHF CLTH

ETLL

TFLL

OMT

0.5 0.4 0.3

0.5

DEGL

VC+S FWLU

ABL

CLHF

TCLH

CLVR CLF

Chi-Square

INCT

TRCO

SEND

CRES

CLVL

CDR

PIMO

TALU

SONS

40-50

NOIS

ATRU

ATC

ARL

CAQ

INAN

RICO

BEAR

HIPR

TFLL

ETLL

0.2

0.2

0.1

0.0

0.0

Betweenness Centrality (BC)

LYDO

CLTH

VC -

MOAN

WAKF

STUP

ABL

TRSE

INAN

VC+S

NOSP ACCO

OMT

LOSP SLEP

OA

ETUL

CRHF TCLH CLF

1

LOSP

NUS

INCO

HV ALUO

GTC

TFRU

VOCA

CRES

CLVR

VC -

CLHF

VEU

MCCL CLVL

EDVD

INAN

FWLU VC+A

RELX

GTCC

TBU

TAAR

TARU

TRCO

CDR DEGL

SEND

RICO

WAKF

ECLO

EYOP

INCT

TALU

LYDO

C

NOIS

VC+S

CAQ PIMO

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

SLEP

EYOP

INAN

INAN

VC -

Edges Betweenness (EB) Arrow Width 1 2 3

85

86

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

A

B

C ALAR

D

INAN

VC+A

BEAR

WAKE LOAR

INAN

LOAR

ECLO

CDR

LOSP

SONS

HIPR

ABL

DLU

DEGL OA

OA

HIPR

OACH AGLO

OACL

SEND

ARUO

BABB HCRU

INCO

OASC

ABUO

AGLO

NEFL

ALUO

EBLB

CDL ALU CDR

MOAN

ABU

SPECT

OACL

LOSP

DEGL SMIL ARL

ALU

VOCA

VC -

NEFL

SPECT VC -

MEDI

INCT

INCT

JBHA

CDL

VC+S WAKE

ALAR

ARU

SONS

BEAR

LOOB

RLU

ARUO OAPT

STAR

EDVL

ABL

RELX

NUS ARU

MOAN

ABUO

GTCC ATRU

E

F INAN

BEAR WAKE LOAR

ECLO CDR

CDL

LOOB

INCT

ARU

JBHA

SONS

VC+A

BEAR

ALAR

ABU

SPECT

DEGL SMIL ARL MOAN NEFL HIPR

BABB

OACH HCRU

NEFL

CDR ALU

OA

VOCA SPECT

EBLB ABUO

DEGL INCO

DLU

OASC OA

AGLO

SEND

ARUO

OAPT ARUO

OACL

INCT

VC -

LOSP

SONS

MEDI

INAN

VC+S

OACL

CDL

ALUO ABL

LOAR

LOSP VC -

ALU

WAKE

ALAR

HIPR

ABUO

EDVL

ABL RLU AGLO

MOAN

ATRU

ARU

STAR

RELX

NUS

0.5

0.05

0.4

0.04

0.3

0.03

0.2

0.02

0.1

0.01

0.0

0.0

Betweenness Centrality (BC)

Cluster Coefficient (CC)

GTCC

Edges Betweenness (EB) Arrow Width 1 2 3

Fig. 4. Different visual representations of the complexity inherent to the semiology of TLE, in which is made a comparison between flowchart (A–B) and graph (C–F) building with the sum of all TLE seizures (30) evaluated in this study taking into account the side of the seizure focus — left (A, C, and E) or right (B, D and F). The size of the nodes represents the (A–B) Frequency of Behaviors, (C–D) Clustering Coefficient, or (E–F) Betweenness Centrality. The edge width represents the (A–B) χ2 log or the (C–F) Edge Betweenness. The rectangle calibration and statistical interaction calibration (arrow width) are shown in A and B as in Fig. 1A. (C–F) Other calibration patterns (sizes and colors) are at the bottom of the figure. For acronym description, see Supplementary Table 1.

with an average of 6.6 per patient (minimum of two and maximum of 18 seizures per patient). The average seizure duration in all patients was 48 ± 31 s. The age at onset of epilepsy was 6.9 years on average.

Ten (55.6%) patients presented with no focal neurological abnormalities, while the remaining 8 had some kind of neurological deficits. Six (33.3%) patients had no significant identified risk factor or cause for

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

epilepsy, 3 (16%) had brain tumor, 3 (16%) had a family history of epilepsy, 2 (11.1%) had febrile seizures, 1 (5.5%) had prior traumatic brain injury, 1 (5.5%) had tuberous sclerosis, 1 (5.5%) had infection of the central nervous system, and 1 (5.5%) had neurocysticercosis. Twelve (66.7%) patients reported lack of precipitating factors, while 6 (33.3%) patients reported sleep as a seizure precipitant. Two (11.1%) patients reported aura, and 16 (88.9%) did not report any signs or symptoms characteristic of the presence of aura. The MRI was normal in 3 (16.6%) patients, 1 (5.5%) had tuberous sclerosis, 1 (5.5%) had cavernous angioma, 1 (5.5%) had vascular injury, 5 (27.7%) had focal cortical dysplasia, 5 (27.7%) had tumor, and 2 (16.6%) had gliosis. The background EEG activity was normal in 15 (83.3%) and abnormal in 3 (16.7%) patients. Two (11.1%) patients had normal interictal EEG, and 3 (16.6%) had interictal discharges seen parasaggitally. The interictal abnormalities were localized in the frontal region in 7 (38.9%) patients, in the rolandic region in 3 (16.6%), in the frontotemporal region in 2 (11.1%), and in the frontal and extrafrontal regions in one (5.55%) patient. The interictal EEG was lateralized to the right hemisphere in 7 (38.9%), to the left in 2 (11.1%), and bilaterally in 4 (22.2%) patients; normal in 2 (11.1%) patients; and parasaggitally line in 3 (16.6%) patients. The localization of the ictal abnormalities was relatively different among the patients regarding its location. It was frontal in 5 (27.7%), non-localizable in 1 (5.5%), frontotemporal in 1 (5.5%), parasagittal in 1 (5.5%), with movement artifacts in 1 (5.5%), rolandic in 1 (5.5%), rolandic and diffuse in 1 (5.5%), extrafrontal in 1 (5.5%), frontal and rolandic in 2 (11.1%), diffuse in 2 (11.1%), and extrafrontal in 2 (11.1%). The ictal changes were lateralized to the right in 7 (38.9%), to the left in 3 (16.6%), and bilaterally in 6 (33.3%) patients; normal in 1 (5.5%) patient; and parasaggitally in 1 (5.5%) patient. The patients were submitted to the following surgeries: right (2 patients, 11.1%) and left (2 patients, 11.1%) frontal lobectomy; right (5 patients, 27.7%) and left (6 patients, 33.3%) frontal lesionectomy; right corticectomy (2 patients, 11.1%); and right hemispherotomy (1 patient, 5.5%). Pathological examination showed the presence of focal cortical dysplasia (7 patients, 38.8%), tuberous sclerosis (1 patient, 5.5%), gliosis (5 patients, 27.7%), tumor (4 patients, 22.2%), and cavernous angioma (1 patient, 5.55%). Therefore, the study was conducted in 10 (55.5%) patients with right frontal lobe epilepsy and in 8 (44.4%) patients with left frontal lobe epilepsy. Based on information provided by the statistical program ETHOMATIC, flowcharts were built for each sum of seizures per patient (see below Figs. 1 and 2). A series of behaviors were detected, such as pushing the seizure alarm button, automatisms, cephalic deviation, behavioral arrest, dystonic postures, cephalic versions, and tonic and clonic postures, among others. The presence of sleep preceding the seizure onset was identified in 34 of the 120 seizures (28.33%; p b 0.00001). Half of the patients had at least one seizure out of sleep. Six (33.3%) patients, four adults, and two teenagers had the majority of their seizures (50–100%) out of sleep. Behavioral arrest before the seizure onset was identified in 39 of the 120 seizures (32.5%; p b 0.00001). Thirteen (72.2%) of the eighteen patients had behavioral arrest before the seizure onset in at least one seizure. Five (27.7%) patients, four children and one adult, had the majority of their seizures (N 50%) preceded by behavioral arrest. Fourteen (77.8%) patients had no aura during EEG recording and denied any aura on questioning. Four (22.2%) patients reported in the postictal phase some type of subjective sensation before their seizures started. Early in the seizures, three of these 4 patients triggered the alarm, which may suggest the presence of an aura. Tonic postures were the most common findings among the patients, with unilateral or bilateral expressions, involving the extension or flexion or even abduction of the upper, lower, or both upper and lower limbs. On the other hand, dystonic postures were less frequent, detected only in four patients but predominantly contralateral to the seizure focus.

87

The postictal facial wiping behavior was observed in 11 (61.1%) patients, and except in two patients, it was not executed exclusively ipsilateral or contralateral to the seizure focus. 3.2. Flowcharts and individual findings 3.2.1. Sum of seizures per patient To assess the applicability of neuroethological analysis in individual patients, we applied the method to the sum of all FLE seizures presented by each patient. In FLE, we also evaluated, 1 min before the onset of seizures, the preictal period and, 1 min after the end of seizures, the postictal period. The TLE seizures were grouped according to the side of seizure focus, one seizure per patient. For details, see [1]. 3.2.1.1. Descriptions from FLE seizures: patient ACE. The flowcharts from Fig. 1(B) represent the sum of four seizures from the patient ACE. In the preictal period (Fig. 1B, 1), some behaviors had longest duration (rectangles with larger base), such as cleaning the hands (CHAN), behavioral arrest (BEAR), smiling (SMIL), normal speech (NOSP), eating (EAT), and staff interferences that do not demand answer or conversation (INCO, in pink). We also observed the presence of some statistical interactions (arrows) between the following: wakefulness (WAKF) and lying down (LYDO, p b 0.0001); cephalic deviation to the right (CDR) and looking to the speaker (LOSP, p b 0.001); INCO and nonictal movements with both hands (NMBU, p b 0.001); staff interference that demands answer (INAN) and positive verbal command with action (VC + A, p b 0.05); NOSP and SMIL (p b 0.05); BEAR and SMIL (p b 0.05); and BEAR and seizure onset (SONS; p b 0.05), typical behaviors of a person that is fully awake. During the seizures (Fig. 1B, 2), only BEAR and clonus with the right upper limb (CLRU, in red) had longer duration. Moaning (MOAN), tonic flexion of both upper limbs (TFBU, red), and tonic posturing of the right hand (TRU, also in red) had both higher frequency and longest duration. Tonic extension of the right lower limb (TERL) also happened but similar to other behaviors, with low frequency and duration. Some interactions between these items also happened: MOAN and TFBU, TFBU and MOAN, p b 0.0001; TRU and MOAN, p b 0.001; and MOAN and TRU, p b 0.05. The patient was unable to respond to staff commands (INAN) with speech or action (negative verbal command, VC−, p b 0.0001) and had BEAR before bilateral eye blinking (BEB, p b 0.0001). Looking around (LOAR) preceded the end of seizures (SEND, p b 0.001). The seizures were then characterized by tonic postures and moaning. In the sum of the postictal periods (Fig. 1B, 3), there was a variety of postictal behaviors, including the presence of certain automatisms (green). In this flowchart, the main findings are the behaviors that had longer duration and also a higher frequency, as INAN, LOSP, LOAR, BEAR, and nonictal movements with the right upper limb (NMRU) and the statistically significant relationships (arrows). The patient answered staff questions (INAN) with action (VC + A, p b 0.01), MOAN (p b 0.001), LOSP (p b 0.001), VC− (p b 0.001), and rarely with speech (VC + S). We also observed three behaviors in orange, two of them directed to the region of the face and executed exclusively with the left hand: eye wiping with the left upper limb (WELU) and facial wiping with the left hand (FWLU). The neuroethological approach presumed the left frontal impairment. The left rolandic region was affected according to the interictal EEG recordings, and the ictal EEG showed left frontal–rolandic discharges. The MRI evidenced a benign tumor in the left frontal region, and the patient had a left brain lesionectomy involving the supplementary motor area. The histopathology confirmed MRI findings. 3.2.1.2. Descriptions from FLE seizures: patient LBS. In Fig. 1C, 1–3, on the other hand, we can see four seizures of this patient characterized by version. The preictal period (Fig. 1C, 1) also shows typical behaviors of a person that is fully awake (WAKF), such as NOSP, watching television (TV), scratching the head with the left upper limb (SHLU), changes in

88

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

facial expression (CFEX), automatisms with both upper limbs with an object (ABUO), automatisms with the left upper limb with an object (ALUO), automatisms with the right upper limb with an object (ARUO), and also oral automatisms (OA). Some statistical interactions between WAKF and LYDO (p b 0.0001); LYDO and TV (p b 0.001); and TV and left upper automatisms with an object (ALUO, p b 0.05) and between BEAR and SONS (p b 0.0001) were detected. The sum of the four ictal periods (Fig. 1C, 2) is particularly interesting. Seizure onset (SONS) was statistically related to LOAR (p b 0.0001). This behavior was then followed by eyes closed (ECLO, p b 0.001). The other sign extremely important was the version of the eyes to the left and upwards (VELU), with expressive duration and frequency. This behavior had statistical interactions with VC− (p b 0.001) and automatisms with the left upper limb (ALU, p b 0.001). The VC− was prevalent, but some interactions between staff interference (INAN) and answers (VC−, p b 0.00001; VC + A, p b 0.0001; and VC + S, p b 0.05) and also the presence of incoherent speech (INCS), NOSP, and correct answer to questioning (CAQ) suggest the preservation of consciousness and speech at least in part. In the postictal period (Fig. 1C, 3), the patient had higher frequency and longest duration of BEAR and LOAR, while some automatisms and other behaviors were also detected. The interactions between the staff interference (INAN) and answers (VC−, VC + S, CAQ, and look to the object — LOOB; p b 0.05) and also between LOOB and correct nomination (CNOM, p b 0.00001) are evidence for the postictal rapid recovery. The main manifestation identified in this patient was version of the eyes to the left and upwards (VELU), suggesting the involvement of the frontal lobe visual field of the right hemisphere. Interictal EEG discharges were bilateral frontal, while the ictal recordings were rolandic and diffuse to the right hemisphere. The MRI finding was suggestive of right frontal lobe tumor. The patient then underwent right lesionectomy for resection of the tumor in the right frontal convexity, anterior to the motor area, and the histopathology confirmed the MRI diagnosis. 3.2.1.3. Descriptions from FLE seizures: patient WFA. The patient WFA (Fig. 2A, 1–3) had a different seizure pattern from that described above, with a relative variety of items during the seizures. Before the seizure onset, the patient was always sleeping (SLEP, Fig. 2A, 1). The sum of the four seizures (Fig. 2A, 2) showed a variety of items developed by the patient, but only yelling (YELL) and ALUO had a longer duration. Hyperkinetic automatisms with the right upper limb (HARU) had a higher frequency and a longer duration when compared with the other behaviors. Axial contractions (AC, in red), CLRU, tonic flexion of both upper limbs (TFBL), tonic flexion of the left upper limb (TLU), dystonic posture of the left upper limb (salmon, DLU), and dystonic posture of the left lower limb (DLL) were also present. Other automatisms (green) also occurred: oral (OA), chewing (OACH), trunk (ATRU), both lower limbs (ABL), both upper limbs (ABU), the left upper limb (ALU), the right upper limb (ARU), and the right lower limb (ARL) and hyperkinetic automatisms with the head (HAH) and the right lower limb (HARL). There were also statistically significant interactions between SONS and AC (p b 0.001), SONS and eyes open (EYOP, p b 0.001), HAH and YELL (p b 0.05), YELL and HAH (p b 0.05), YELL and ATRU (p b 0.001), ATRU and YELL (p b 0.05), HARU and OA (p b 0.001) and also between tachypnea (TPNE) and SEND (p b 0.0001). Those interactions illustrate that the seizures usually began with AC or EYOP. Moreover, during the seizures, the most repetitive behaviors were yelling, hyperkinetic, oral, and also trunk automatisms. Tachypnea preceded the end of the seizures. In the postictal period (Fig. 2A, 3), trunk deviation to the left (TDL), LOOB, and BEAR had the longest duration, and LOSP, OACH, and INAN had the higher frequency and the longest duration. The patient answered staff questions (INAN) with VC + A (p b 0.001), CAQ (p b 0.001), VC− (p b 0.001), or LOOB (p b 0.05). Other frequent interactions included INCS and LOAR (p b 0.001), VC + A and OACH (p b 0.001), VC + S and OACH (p b 0.05), and INCO and LOSP (p b 0.001).

The sum of seizures suggests involvement of the right frontal regions. The interictal EEG demonstrated right frontal discharges, while ictal recordings were abnormal over the bilateral frontal regions. The MRI findings were suggestive of right focal cortical dysplasia localized in the frontal operculum. The right lesionectomy included the resection of the frontal operculum and also a part anterior to the motor area. The histopathology confirmed MRI findings.

3.2.1.4. Descriptions from FLE seizures: the outstanding case of patient BGC with 18 summed seizures. In the preictal phase (Fig. 2B, 1), some behaviors had the longest duration, such as WALK, TV, ACCO, INAN, VC + F, NOSP, and WAKF. The items with the higher frequency and duration were SLEP and BEAR. The significant statistical interactions were between EYOP and ECLO (p N 0.0001), ECLO and BEAR (p N 0.001), BEAR and SONS (p b 0.05), and SLEP and SONS (p b 0.00001). Sleep and BEAR were the main behaviors related to SONS. The patient was in the bed, awake and interacting with the staff: WAKF and LYDO (p b 0.0001), LOSP and INAN (p b 0.05), INAN and VC + F (p b 0.0001), and VC + F and INAN (p b 0.05). The preictal period showed a variety of other behaviors, including some upper limb (ARUO and ALUO) and oral automatisms (OAPT and OAOC), movements of the eyes (EYOP, ECLO, LOOB, LOSP, and LOAR), scratching behaviors (FWLU, WFLU, WELU, WNLU, and SHLU), staff interferences (INAN INCO, NOIS and INCT), and patient answers (VC + A, VC + F, VC−, NUS and NOSP), among others. The sum of the ictal periods (Fig. 2B, 2) showed a vast amount of tonic and clonic behaviors (in red). However, some limb (ALUO, ABU and ARL), oral (OAPT and OACH), and trunk automatisms (ATRU), as well as right leg hyperkinetic automatism (HARL) were also detected. The staff interferences were also present (INAN, INCO, INTW, NOIS, INTC, MEDI, INTO, and SPECT) without patient answer (VC−) or LOSP. Some behaviors had higher duration: RICO, BEAR, ARL, ABL, CLRU, CLF, TFLU, TFLL, VEU, TBU, and TAAR, and others had higher frequency and duration: EYOP, MOAN, INAN, VC −, INCT, ATC, GTC, GTCC, and CRES. Axial tonic contraction and CRES received and sent arrows to a variety of items. The statistical interactions between SONS and EYOP (p b 0.00001), EYOP and ATC (p b 0.00001), and EYOP and TALU (p b 0.001) clearly show that ATC and also TALU were the main items directly related to the seizure onset: the seizures then evolved to TAAR (p b 0.0001), TALU (p b 0.001), and also TARU (p b 0.001), characterizing the axial tonic seizures. Moaning was then recorded after TALU (p b 0.0001), TARU (p b 0.001), TAAR (p b 0.001), and ARL (p b 0.05) and preceded RICO (p b 0.001) and NOIS (p b 0.05), while RICO anticipated VEU (p b 0.0001). Unilateral clonic and also tonic flexion was present only in the right arm (CLRU and TFRU). Moreover, right hemiface clonus (CRHF) had statistical interactions with other items more than the left hemiface clonus (CLHF). Those interactions included TFRU and CRHF (p b 0.0001), CRHF and CLTH (p b 0.0001), CRHF and CRES (p b 0.001), CRES and INCT (p b 0.001), and INCT and CLHF (p b 0.001). The cephalic version, just like CLRU and TFRU, was also recorded toward the right side (VCED), without any statistical interaction. The clonic cephalic version was directed to the left (CLVL) and to the right (CLVR), both with significant statistical interactions with CRES: CLVL and CRES (p b 0.001), CLVR and CRES (p b 0.0001), and also CRES and CLVR (p b 0.001). Other statistically significant interactions were observed: CRES and MCCL and vice-versa (p b 0.001), CRES and VOCA (p b 0.05), VOCA and CRES (p b 0.0001), CLF and CRES (p b 0.001), OMT and CLTH (p b 0.0001), and OMT and VEU (p b 0.0001). The main statistically significant interactions during the secondary generalization were between GTC and ETLL and viceversa (p b 0.00001), GTC and GTCC (p b 0.0001), GTCC and TFLL (p b 0.0001), TFLL and ETUL (p b 0.001), ETUL and GTCC (p b 0.0001), GTCC and RELX (p b 0.00001), and RELX and SEND (p b 0.00001). SITT and SEND (p b 0.05). Both SITT and RELX were closely related to

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

the end of the seizures, showing that not all the seizures evolved to secondary generalization. The lack of verbal answers while TBU and LOSP were present (see the arrows between TBU and INAN, INAN and LOSP, INAN and VC−, and VC − and TBU) suggests consciousness impairment or dominant hemisphere involvement, as well as the development of generalized tonic-clonic seizures. Axial tonic contractions (ATC, TAAR, TARU, and TALU) were the most relevant items of the patient BGC. Other tonic items were present, reflecting seizure spread to adjacent frontal areas. However, the interactions between tonic and clonic behaviors were recorded with right hemibody, indicating left hemisphere involvement. The postictal periods had a variety of behaviors (Fig. 2B, 3): automatisms (limb — ABU, ARUO, ALUO, and ABL; oroalimentary — OACH, OA, and OAPT; and trunk — ATRU); staff interferences (INAN, INCO, INTC, INTW, INCT, MEDI, and SPECT); answers to staff (VC + S, VC + A, VC−, CNOM, and CAQ); and also some scratching behaviors (FWRU, FWLU, SRLL, and SLLU). The higher durations were found in ABL, NUS, VC + A, CAQ, TRCO, INCT, and OA, while PIMO, HV, ALUO, LOAR, LOSP, BEAR, ACCO, HIPR, INCO, INAN, VC + S, and VC − had higher frequencies and durations. The interactions between SEND and HV (p b 0.0001), HV and PIMO (p b 0.00001), and PIMO and HV (p b 0.00001) show that the main items were related to the end of the seizures. Other interactions point to the variety of the postictal answers: INAN and VC + F (p b 0.00001), INAN and VC− (p b 0.00001), and INAN and VC + A (p b 0.001). The VC + S was followed by CAQ (p b 0.0001) and TRCO (p b 0.05), which, in turn, precedes NOSP (p b 0.0001). In addition, the lack of answers (VC −) was associated with ABL (p b 0.001), followed by ALUO (p b 0.05) and then HIPR (p b 0.001). Other interactions were also detected: STUP and INCO (p b 0.00001), INCT and OACH (p b 0.001), OA and TRCO (p b 0.001), ACCO and HIPR (p b 0.001), HIPR and ABL (p b 0.001), ABL and HIPR (p b 0,05), LOSP and EDVD (p b 0,0001), and EDVD and DEGL (p b 0.05). The postictal scratching behaviors were not restricted to one hand, nor exclusively directed to the face, but were prevalent with the left hand (FWLU, SRLL, and SLLU), and only FWLU had statistical interactions with INAN (p b 0.05). In this patient, the neuroethological analysis suggests the potential involvement of the left frontal regions. The interictal and ictal EEGs evidenced left frontal region involvement, while the MRI showed focal cortical dysplasia in the left frontal pole. The patient underwent left lesionectomy, and the histopathology confirmed the MRI diagnosis.

3.3. Graph theory as a tool to describe FLE and TLE seizure semiology 3.3.1. FLE seizures Regarding the use of graph theory methods, the new measurements offer reliable information about the intrinsic dynamic of the behavioral seizure sequence and complexity in addition to characterization of the behaviors during the preictal and postictal periods. For instance, it is possible to observe that both preictal and postictal periods of the FLE only showed lower values of Clustering Coefficient (Fig. 3B, 1 and 3), highlighting isolated behaviors with few connections. On the other hand, when we observed the ictal period (Fig. 3B, 2), it is possible to notice that the tonic–clonic behaviors have higher Clustering Coefficient values, indicating that these behaviors have cohesive structure with several connections between behaviors. When the Betweenness Centrality is observed, a similar pattern emerges for the ictal period of the FLE (Fig. 3C, 2), in which the tonic–clonic behaviors have again the higher values, indicating that these behaviors are not only cohesive but also, in most of the cases, part of the connectivity paths between behaviors. It is interesting to verify also, for example, the cases of ATC, CRES, GTC, and MOAN behaviors, which have higher values of Clustering Coefficient, Betweenness Centrality, and Edge Betweenness (Figs. 3B

89

and C), in addition to significantly higher degree, showing that they are important hubs of the ictal period of the patient BGC with FLE. Despite the lower values of the Clustering Coefficient in the preictal and postictal FLE, it is possible to find some behavior with higher values of the Betweenness Centrality, which can be explained because these are intermediary behaviors between the beginning of the recording and the seizure onset. 3.3.2. TLE seizures For the TLE seizures, on the other hand, the predominant behaviors, in terms of higher values of Clustering Coefficient, Betweenness Centrality, and Edge Betweenness, are automatisms, which shows that these behaviors are strongly cohesive and, at the same time, intermediaries between other behaviors, such as ATRU, ALU, ARU, and OACL that also could be considered as a hubs by their significantly higher degree. Other behaviors that also appear to be hubs are INCT, INAN, and SPECT in the “Staff Interactions” (pink color) category and behaviors such as ALAR, DEGL, LOAR, RELX, SONS, and VC + A, which belong to the “Other Behaviors” (cyan color) category. It is important to notice that the right TLE showed the lowest values of Clustering Coefficient (Fig. 4D), evidencing a less structured pattern than in the left TLE, but in which the automatisms still have the role of intermediaries between other behaviors, as shown by the higher Betweenness Centrality and Edge Betweenness, highlighting ARU and OCAL behaviors. 3.3.3. SWI calculations in FLE and TLE seizures Finally, we made the calculation of the SWI using as practical examples the sum of 18 FLE seizures of patient BGC or the sum of seizures from patients with left TLE versus the sum of seizures from patients with right TLE. In both examples, the sum of 18 seizures in patient BGC (Fig. 3B) and the sum of 15 seizures (one per patient) in right TLE (Fig. 4D), we found remarkably high SWI, which is illustrated in Table 1. 4. Discussion 4.1. Neuroethology as a reliable quantitative tool to characterize FLE and TLE semiology Semiological analysis of FLE seizures has been well studied by many authors [6–13], and the variable behavioral repertoire suggests the activation of certain brain regions and the spreading to adjacent areas. To our knowledge, this is the first time that neuroethology, which previously developed and already validated both experimental models of epilepsy [20,35–37] and human mesial TLE seizures [1,2], has been applied the study of FLE seizures. The extent of the frontal lobes with their multiple heterogeneous subregions and local connectivity, in addition to the expressive amount of different network connections involving the frontal lobes and other brain areas, can generate a wide variety of seizures [7,10,38]. The main goal of this research was to know, from the neuroethological point of view, how the frontal lobe seizures would be expressed in patients who become seizure-free. For that purpose, we included all

Table 1 Small-World-Ness Index (SWI) estimated for the graphs indicated in Figs. 3B and 4C and D. The SWI was calculated using the corrected ratio between the Cluster Coefficient and the Shortest Path Length. The correction was made, adjusting the Cluster Coefficient and the Shortest Path Length by their corresponding values estimated in equivalent random graphs (for details, see [33]). Figure/graph/groups

Small-World-Ness Index

Fig. 3B: FLE, total of 18 seizures, ictal period Fig. 4C: left TLE Fig. 4D: right TLE

⁎1.4313 0.5774 ⁎1.3976

⁎ Observation: values greater than 1 (bold numbers) indicate that the graph presented a small-world pattern following [33].

90

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

recorded seizures with clear sound and image quality. In general, in our series, there was a prevalence of tonic seizures, whereas clonic, hypermotor, or versive seizures were less frequently detected. As previously described for TLE seizures [1,2], we observed that neuroethology allows the quantification of a variety of behavioral sequences in FLE seizures. In the latter, the most frequent items were unilateral or bilateral tonic behaviors, symmetrical or not, versions, clonic jerks, and also hyperkinetic (complex motor behaviors) or oroalimentary automatisms. It also identified ictal speech arrest and different forms of ictal vocalization. One major highlight of neuroethology is the observation and recording of all behaviors developed by the patient, while in other studies, the analysis is restricted to only one new behavior or to a group of behaviors already summarized in the literature [8,10,23]. The latter approach may reduce the possibility of finding new behaviors that may have significant localization, lateralization, and prognostic value. This was taken into account when we included 1 min of preictal and 1 min of postictal time window for each analyzed seizure. We found that sleep (p b 0.00001) and behavioral arrest (p b 0.00001) frequently preceded the seizure onset. More than 50% of the seizures started out of sleep in the adults plus adolescent populations. The sleep before the seizure onset was always detected, for example, in the preictal period of the patient WFA (Fig. 2A, 1). In accordance with literature data, in FLE, there is a prevalence of nocturnal events related to sleep [9,38,39]. Jobst and colleagues [9], from their analysis of 449 seizures in 26 adult patients with drug-resistant FLE, described that fourteen (54%) patients had the majority (N 50%) of their seizures out of sleep. Our group of patients was smaller, but our results are in agreement with the literature [38,40]. Behavioral arrest related to seizure onset was identified in 39 (32.5%) of the 120 seizures. The arrows between BEAR (p b 0.0001) and the seizure onset (SONS), in the preictal periods of ACE (Fig. 1B, 1), LBS (Fig. 1C, 1), and BGC (Fig. 2B, 1), demonstrate that behavioral arrest frequently preceded the beginning of the seizures. The sudden seizure onset is usually described in FLE [4,9,38,39]. However, the preictal period has not previously been analyzed in a systematic way, with the statistical approach of neuroethology. Even with tonic prevalence, the sum of seizures per patient allowed the observation of the more relevant semiologic items, the ones which had higher frequency and duration, as well as the ones which had higher values of Clustering Coefficient and Betweenness Centrality (see below). Moreover, the different patterns of behavioral expression, found by the statistical interactions, contributed to detailed analysis and discussion, on each patient, of the possible cerebral areas involved in seizure propagation. In such methods, it is possible to discuss, based on the literature descriptions, the lateralizing and localizing values of the behaviors and, most importantly, to infer about the possible pathways involved with this specific seizure propagation. For example, the tonic posturing in ACE seizures (Fig. 1B, 2) was unilateral (TRU) or bilateral (TFBU) and affected mainly the upper limbs. The tonic extension of the right lower limb (TERL) was also identified. Moreover, clonic contractions happened only within the right hand. Tonic postures are the most common finding in FLE seizures [6,8,9,17]. According to the literature, the involvement of frontal regions in FLE seizures could cause such tonic and clonic behaviors. The lateralizing value of tonic posturing is controversial [18], tending to be seen contralateral to the focus [41] especially in extratemporal seizures. According to Loddenkemper and Kotagal [18], the tonic posturing is generated not only by the activation of the supplementary motor area (SMA) but also by the premotor area, anterior cingulate, and subcortical structures such as the basal ganglia. Clonic contractions present contralateral lateralizing value and can be generated by stimulating the primary motor area, Brodmann's area 4 [18]. The absence of the patient's verbal response to staff questioning suggests the involvement of the dominant cerebral hemisphere and/or impaired consciousness. The VC− can

be considered a localizing sign of impairment of Broca's area (area 44 and part of Brodmann's area 45). The involvement of the left cerebral hemisphere was then predicted by neuroethology. Those remarkable semiologic findings and their predictions of ictal brain activation were later confirmed by the patient's history and the seizure freedom after the excision of a left rolandic tumor. Different from the patient ACE, those with LBS seizures (Fig. 1C, 2) were versive with preserved consciousness. According to Lüders and colleagues [15] versive seizures are defined as sustained and extreme conjugate eye movement to one side or the movement of the head and, occasionally, the whole body to one side. The lateral movement of the eyes frequently consists of a combination of a smooth tonic lateral movement on which small saccades are superimposed that progressively move the eye out to an extreme position. On other occasions, a smooth lateral movement without any saccades may be observed. The version of body parts has a similar character, but the saccades are replaced by small clonic lateral movements of the head or body. Electrical stimulation of the frontal eye field elicits mainly saccadic, contralateral, conjugate eye movement, frequently followed by head version [4]. This patient had only tonic conjugate eye movements and no other sign of tonic posturing or head version. This behavior suggests the involvement of the right frontal eye and motor areas anterior to the precentral gyrus (Brodmann areas 6 and 8; [18]). The seizure freedom after the lesionectomy in the right frontal convexity, anterior to the motor area, confirmed the neuroethological predictions. On the other hand, the WFA seizures (Fig. 2A, 2) were always out of sleep with the prevalence of hyperkinetic automatisms, but tonic, clonic, and also dystonic behaviors were present. Frontal lobe bizarre hyperactive seizures are stereotypic, brief, and occur in clusters, can be exclusively nocturnal, and disrupt sleep [38]. The hyperkinetic seizures have been associated mostly with seizure onset in the orbitofrontal cortex and the anterior medial frontal regions, including the cingulate gyrus [17]. Ictal behavior can be the result of propagation of the electrical seizure discharge and may not necessarily correlate with the site of seizure origin [9]. Moreover, the patient also had yelling during seizures, a behavior that also can occur in hyperkinetic seizures [40]. The unilateral tonic posture of the left hand (TLU) and also the dystonic postures of the left upper (DLU) or lower (DLL) limbs were not frequent or long-lasting items in WFA seizures, but their unilateral occurrence can suggest the impairment of the right hemisphere. As mentioned above, the tonic posturing is generated not only by the activation of the SMA but also by the premotor area, anterior cingulate, and subcortical structures such as the basal ganglia [18] and tends to be seen contralateral to the focus especially in extratemporal seizures [41]. Dystonic posturing affects mainly the distal part of the superior limb and consists of tonus alteration associated with a rotational component, typically including the flexion of the wrist and metacarpophalangeal joints, finger extension, and forearm rotation [42]. Dystonic posturing is one among various ictal behaviors that usually occurs during TLE seizures as a reliable contralateral sign [23,43–45]. Our group previously described a significant negative correlation between contralateral dystonia and secondary generalization (bilateral tonic or clonic behaviors) in patients with TLE [46]. In FLE seizures, dystonic postures not only were frequently observed but also have a lateralizing value contralateral to the seizureonset zone [6,19,47]. Studies using SPECT showed an increase in blood flow of the ipsilateral basal ganglia, particularly the putamen, in TLE-related dystonia [44,48]. Bonelli and co-workers [6] found unilateral dystonic posturing in 26% of patients and 20% of seizures with a significant lateralization value for seizures but not for patients. As the authors said, this can be explained by the low number of patients with this symptom. The patient WFA was one of the four patients who had dystonic postures, and, as in the others, it was contralateral to the seizure onset. Data obtained from the sum of seizures in this patient suggest the impairment of the right frontal regions. The occurrence of

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

tonic and dystonic behaviors in addition to the hyperkinetic automatisms can represent the spread to other regions in the brain. So, the hyperkinetic automatisms were the main finding in this patient's seizure, suggesting a frontal lobe focus. The tonic and dystonic postures were findings that contributed to the lateralization and also to the study of the potential spread pathways involved. The patient became seizurefree after the lesionectomy in the right frontal operculum and the excision of the focal cortical dysplasia, supporting the neuroethological approach. The postictal periods were also investigated, and some behaviors were suggestive of cognitive and motor rapid recovery. The ACE postictal periods (Fig. 1B, 3) included the following: CNOM, NOSP, CAQ, VC + A, and VC + S. The patient LBS (Fig. 1C, 3) had CAQ, VC + A, VC + S, NOSP, INCS, and CNOM. This rapid recovery has been described by many authors as typical of FLE seizures [7,9,39]. The facial wiping (FW), defined as the directed movement of the hand to the face, where the patient rubs, scratches, and runs his hand on the nose (“nose wiping”), eye, cheek, mouth, chin, forehead, ear, or face [49], was another postictal finding. In mTLE, FW is frequently observed in the postictal period and has an ipsilateral lateralizing value [49–52]. The involvement of the contralateral limb in dystonia, in tonic postures, or due to the occurrence of paresis or neglect could explain the ipsilateral predominance of the facial wiping in TLE [18,51]. In extratemporal epilepsy, however, it is less frequent and without lateralizing value in FLE seizures [6,53]. The postictal facial wiping behaviors in our group were observed in 11 patients and were exclusively unilateral only in two patients. One of them was the patient ACE who had the postictal facial wiping behaviors developed exclusively with the left hand (Fig. 1B, 3), suggesting the involvement of the left hemisphere. The other was hemiplegic and had contralateral FW. All the other patients had, sometimes, ipsilateral or, sometimes, contralateral postictal FW, so without lateralizing value, in accordance with literature data [6,53]. Only four (22.2%) patients reported the presence of an aura. Patients with FLE seizures often describe a nonspecific feeling that is nonlocalized or localized to the head [40]. In a series of 26 patients with frontal lobe seizures, 31% of patients had no aura, whereas the remaining patients reported nonlocalized head feelings, fear, autonomic symptoms, somatosensory symptoms, and sensations localized to the chest [40]. Tonic seizures can be preceded by an aura of unilateral or bilateral sensory symptoms [38]. Quesney and colleagues [54] have associated somatosensory auras with parasagittal seizure onset, but all of the other frontal lobe auras were not considered localizing. Our group of patients was relatively small and also composed of different ages, including children, which could explain the reduced description of auras found in our series. Semiology can reflect only the symptomatogenic zone and, therefore, can give indirect information about the seizure-onset zone or the epileptogenic zone, as the epileptic activity may have spread from a “silent” cortical area into a different cortical area that actually produces symptoms [5,18]. The demonstration of these circuitries requires coupling with neuroimaging (SPECT and functional MRI) and electrophysiology. This kind of approach was applied recently by our group in mTLE seizures [1]. This study investigated the pathophysiological correlations between ictal behavioral expressions and ictal SPECT findings in a homogeneous group of patients with TLE. A detailed analysis of individual seizures and their associated brain substrates by coupling neuroethology and SPECT findings reliably evaluated ictal behavior and functionality of associated brain areas in patients with TLE. For the current analysis, we believe that there are specific circuits involved with FLE manifestation that include a complex network connecting the focus with other cerebral regions whose activation can also generate the signals and signs of this particular type of epilepsy. In all patients studied, these parameters were recorded and, although not considered in the current work, they will be correlated in future studies. Because the quantification of movement trajectory in patients with TLE [55] based on the method developed by Li and colleagues [56]

91

serves as another contemporary and appropriate tool for characterizing the details of behavioral manifestations of seizures, the technical developments and rationale behind movement trajectory quantification can be coupled to the current neuroethology tools to further improve semiological evaluation in FLE (see a comment in [57]).

4.2. Graph theory as applied to FLE and TLE quantitative semiology: a network and complexity view of epilepsy studies In the last decade, complexity measurements [58–61] and graph theory [25,34,62,63] have been frequently applied to neuroscience and particularly to connectivity in brain networks, using either fMRI or EEG data (see below). Incredibly, although the behavioral expression of epileptic seizures are the end point of the activity of those brain networks, there are practically no publications using graph theory as applied to this complex semiological data. Indeed, there are some studies using graphs with MRI [64,65] or EEG data [65,66] in patients or even in animal models of epilepsy [67], but none with ictal semiology of epileptic seizures in patients. A quite nice example of this approach as applied to MRI and EEG connectivity analysis was recently published [68]. In the epilepsy semiological arena, as far as we know, our laboratory has been the first to follow up on previous neuroethological studies [2] in which we describe either TLE seizure semiology alone or its coupling to ictal SPECT [1]. We, therefore, reviewed recently the international literature and described how quantitative semiology analysis would be improved by means of complexity measurements, which includes graph theory [26,69]. The current coupling of neuroethological tools with graph analysis adds computational and prediction power to our studies. Although the current initial graph analysis was used only for the 18 seizures from the patient BGC with FLE (Fig. 3) and from the interhemispheric comparison in TLE (Fig. 4), the results are compelling. First of all, we were able to determine, using the same dataset used for flowcharts (Figs. 1 and 2), that specific behaviors and their sequential interactions are essential for the dynamics and connectivity of a complex behavioral network, obviously with the advantage of measuring these interactions with graph metrics (Figs. 3 and 4; Table 1). We, in fact, were able to characterize network complexity of FLE and TLE seizures semiology based upon Node Metrics such as Clustering Coefficient, Betweenness Centrality, and Network Metrics such as Edge Betweenness, SmallWorld-Ness Index, and Hub Detection. Strong and selective connections, hubs, and network topology already shown with MRI and EEG data and typical of complex networks were, by analogy, found in our FLE and TLE data. Last but not the least, it has been said that usually small word networks are efficient ones [70], in fact meaning that those networks are seen as systems that are both globally and locally efficient. In the current study, the values we found positive for SWI in graphs derived from sequences of behaviors of both FLE and TLE seizures (Table 1) strongly support the view that the ability to produce synchronous or coherent behavioral events, in other words, repeated sequences with specific network topology and metrics, can be dependent on seizures arriving from left or right hemispheres as well as from patients with TLE or FLE. In that direction, recently, Gong et al. [71] have shown that, during epileptiform discharges in the in vitro Mg2+-free epilepsy, the hippocampal networks formed more local connections compared with the control. In other words, they found an alteration from a random architecture in the normal network to an organized architecture in the pathological one. The authors compare their data with other data from studies using human EEG data where increased small-world organization or even more regular network configuration in epileptic versus normal or less pathological control brains was found [72,73], which is in contrast with other studies with disrupted small-world organization in epileptic brains [74–76].

92

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93

Because, based upon the current studies, we were able to detect crucial differences between FLE and TLE seizures, from right and left TLE seizures, future studies should evaluate for the so-called vulnerability or resilience (resistance to attack) of such networks by analogy to studies made with neuroanatomical and electrophysiogical substrates. Finally, future research using simultaneous measurements and recordings of neuroanatomical and EEG connectivity, with behavioral studies such as those illustrated here, with both neuroethology and graph theory tools and others with more sophisticated approaches such as 3D image capturing and reconstruction of the whole body, will be needed in order to compose a more realistic landscape for the integrated complexities present in the behavioral expression of epileptic brains. Other experiments involving larger samples, probably multicenter studies, and also the association of neuroethology and graph theory with other examinations such as EEG, SPECT, and MRI should be performed to verify and confirm the activation of the circuits responsible for the generation of behavior and propagation of seizures and the eventual coherence between the data collected from this multidisciplinary approach. This will be of particular importance for the evaluation of epilepsy and neuropsychiatry comorbidities, a clinical challenge with even greater complexity where not only behavior but also memory, emotions, and cognition are at risk. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.yebeh.2014.07.025. Acknowledgments Thanks are due to FAPESP (07/50261-3), Cinapce-FAPESP (05/ 56447-7), CNPq (05/2007), PROAP-CAPES (0241/09), PROEX, and FAEPA for their financial support. We thank the CIREP staff and technicians for their kind help with patients, video-taping, and training during the development of this study. We also thank all the Neurophysiology and Experimental Neuroethology Laboratory (LNNE) members. PB holds a PhD FAPESP Post-Doctoral Fellowship, and NGC holds a CNPq Research Fellowship. Disclosure PB, JT, APPM, MLCDC, VCT, JACO, TRV, ACS, and NGC declare that they have no conflict of interest with respect to the work submitted in this article. References [1] Bertti P, Dal-Cól MLC, Wichert-Ana L, Kato M, Terra VC, de Oliveira JAC, et al. The neurobiological substrates of behavioral manifestations during temporal lobe seizures: a neuroethological and ictal SPECT correlation study. Epilepsy Behav 2010;17:344–53. [2] Dal-Cól MLC, Terra-Bustamante VC, Velasco TR, Oliveira JAC, Sakamoto AC, GarciaCairasco N. Neuroethology application for the study of human temporal lobe epilepsy: from basic to applied sciences. Epilepsy Behav 2006;8:149–60. [3] Manford M, Hart YM, Sander JW, Shorvon SD. National General Practice Study of Epilepsy (NGPSE): partial seizure patterns in a general population. Neurology 1992;42:1911–7. [4] Kellinghaus C, Lüders HO. Frontal lobe epilepsy. Epileptic Disord 2004;6:223–39. [5] Rosenow F, Lüders H. Presurgical evaluation of epilepsy. Brain J Neurol 2001;124: 1683–700. [6] Bonelli SB, Lurger S, Zimprich F, Stogmann E, Assem-Hilger E, Baumgartner C. Clinical seizure lateralization in frontal lobe epilepsy. Epilepsia 2007;48:517–23. [7] Chauvel P, Kliemann F, Vignal JP, Chodkiewicz JP, Talairach J, Bancaud J. The clinical signs and symptoms of frontal lobe seizures. Phenomenology and classification. Adv Neurol 1995;66:115–25 [discussion 125–126]. [8] Janszky J, Fogarasi A, Jokeit H, Ebner A. Lateralizing value of unilateral motor and somatosensory manifestations in frontal lobe seizures. Epilepsy Res 2001;43:125–33. [9] Jobst BC, Siegel AM, Thadani VM, Roberts DW, Rhodes HC, Williamson PD. Intractable seizures of frontal lobe origin: clinical characteristics, localizing signs, and results of surgery. Epilepsia 2000;41:1139–52. [10] Kotagal P, Arunkumar G, Hammel J, Mascha E. Complex partial seizures of frontal lobe onset statistical analysis of ictal semiology. Seizure 2003;12:268–81. [11] Kramer U, Riviello JJJ, Carmant L, Black PM, Madsen J, Holmes GL. Clinical characteristics of complex partial seizures: a temporal versus a frontal lobe onset. Seizure 1997;6:57–61.

[12] Salanova V, Morris HH, Van Ness P, Kotagal P, Wyllie E, Lüders H. Frontal lobe seizures: electroclinical syndromes. Epilepsia 1995;36:16–24. [13] Wieser HG. Electroclinical features of the psychomotor seizure: a stereoelectroencephalographic study of ictal symptoms and chronotopographical seizure patterns including clinical effects of intracerebral stimulation. Gustav Fischer; 1983. [14] Blume WT, Lüders HO, Mizrahi E, Tassinari C, van Emde Boas W, Engel JJ. Glossary of descriptive terminology for ictal semiology: report of the ILAE task force on classification and terminology. Epilepsia 2001;42:1212–8. [15] Lüders H, Acharya J, Baumgartner C, Benbadis S, Bleasel A, Burgess R, et al. Semiological seizure classification. Epilepsia 1998;39:1006–13. [16] Kotagal P, Lüders HO, Williams G, Nichols TR, McPherson J. Psychomotor seizures of temporal lobe onset: analysis of symptom clusters and sequences. Epilepsy Res 1995;20:49–67. [17] Manford M, Fish DR, Shorvon SD. An analysis of clinical seizure patterns and their localizing value in frontal and temporal lobe epilepsies. Brain 1996;119:17–40. [18] Loddenkemper T, Kotagal P. Lateralizing signs during seizures in focal epilepsy. Epilepsy Behav 2005;7:1–17. [19] Bleasel A, Kotagal P, Kankirawatana P, Rybicki L. Lateralizing value and semiology of ictal limb posturing and version in temporal lobe and extratemporal epilepsy. Epilepsia 1997;38:168–74. [20] Garcia-Cairasco N, Wakamatsu H, Oliveira JA, Gomes EL, Del Bel EA, Mello LE. Neuroethological and morphological (Neo–Timm staining) correlates of limbic recruitment during the development of audiogenic kindling in seizure susceptible Wistar rats. Epilepsy Res 1996;26:177–92. [21] Garcia-Cairasco N, Doretto MC, Prado RP, Jorge BP, Terra VC, Oliveira JA. New insights into behavioral evaluation of audiogenic seizures. A comparison of two ethological methods. Behav Brain Res 1992;48:49–56. [22] Garcia-Cairasco N, Sabbatini RM. Role of the substantia nigra in audiogenic seizures: a neuroethological analysis in the rat. Braz J Med Biol Res 1983;16:171–83. [23] Williamson PD, Thadani VM, French JA, Darcey TM, Mattson RH, Spencer SS, et al. Medial temporal lobe epilepsy: videotape analysis of objective clinical seizure characteristics. Epilepsia 1998;39:1182–8. [24] Rubinov M, Sporns O. Complex network measures of brain connectivity: uses and interpretations. NeuroImage 2010;52:1059–69. [25] Sporns O. Networks of the brain. 1st ed. The MIT Press; 2010. [26] Tejada J, Costa KM, Bertti P, Garcia-Cairasco N. The epilepsies: complex challenges needing complex solutions. Epilepsy Behav 2013;26:212–28. [27] Onias H, Viol A, Palhano-Fontes F, Andrade KC, Sturzbecher M, Viswanathan G, et al. Brain complex network analysis by means of resting state fMRI and graph analysis: will it be helpful in clinical epilepsy? Epilepsy Behav 2013 Dec 27. http://dx.doi.org/ 10.1016/j.yebeh.2013.11.019 pii: S1525-5050(13)00617-3 [Epub ahead of print]. [28] Engel J. Surgical treatment of the epilepsies. 2 sub. Lippincott Williams & Wilkins; 1993. [29] Cline MS, Smoot M, Cerami E, Kuchinsky A, Landys N, Workman C, et al. Integration of biological networks and gene expression data using cytoscape. Nat Protoc 2007;2: 2366–82. [30] Lopes CT, Franz M, Kazi F, Donaldson SL, Morris Q, Bader GD. Cytoscape web: an interactive web-based network browser. Bioinforma Oxf Engl 2010;26:2347–8. [31] Bianconi G, Coolen ACC, Vicente CJP. Entropies of complex networks with hierarchically constrained topologies. Phys Rev E 2008;78. [32] Shannon CE. A mathematical theory of communication. Bell Syst Tech J 1948;27: 379–423. [33] Humphries MD, Gurney K. Network “small-world-ness”: a quantitative method for determining canonical network equivalence. PLoS ONE 2008;3:e0002051. [34] Watts DJ, Strogatz SH. Collective dynamics of /‘small-world/’ networks. Nature 1998; 393:440–2. [35] Etholm L, Arabadzisz D, Lipp H-P, Heggelund P. Seizure logging: a new approach to synchronized cable-free EEG and video recordings of seizure activity in mice. J Neurosci Methods 2010;192:254–60. [36] Furtado MA, Castro OW, Del Vecchio F, de Oliveira JAC, Garcia-Cairasco N. Study of spontaneous recurrent seizures and morphological alterations after status epilepticus induced by intrahippocampal injection of pilocarpine. Epilepsy Behav 2011;20: 257–66. [37] Rossetti F, Rodrigues MCA, de Oliveira JAC, Garcia-Cairasco N. Behavioral and EEG effects of GABAergic manipulation of the nigrotectal pathway in the Wistar audiogenic rat strain. Epilepsy Behav 2011;22:191–9. [38] Jobst BC, Williamson PD. Frontal lobe seizures. Psychiatr Clin N Am 2005;28:635–51. [39] McGonigal A, Chauvel P. Frontal lobe epilepsy: seizure semiology and presurgical evaluation. Pract Neurol 2004;4:260–73. [40] Williamson PD, Jobst BC. Frontal lobe epilepsy. Adv Neurol 2000;84:215–42. [41] Kotagal P, Bleasel A, Geller E, Kankirawatana P, Moorjani BI, Rybicki L. Lateralizing value of asymmetric tonic limb posturing observed in secondarily generalized tonic–clonic seizures. Epilepsia 2000;41:457–62. [42] Kotagal P. Significance of dystonic posturing with unilateral automatisms. Arch Neurol 1999;56:912–3. [43] Marks WJJ, Laxer KD. Semiology of temporal lobe seizures: value in lateralizing the seizure focus. Epilepsia 1998;39:721–6. [44] Newton MR, Berkovic SF, Austin MC, Reutens DC, McKay WJ, Bladin PF. Dystonia, clinical lateralization, and regional blood flow changes in temporal lobe seizures. Neurology 1992;42:371–7. [45] Olbrich A, Urak L, Gröppel G, Serles W, Novak K, Porsche B, et al. Semiology of temporal lobe epilepsy in children and adolescents. Value in lateralizing the seizure onset zone [corrected]. Epilepsy Res 2002;48:103–10. [46] Cleto Dal-Cól ML, Bertti P, Terra-Bustamante VC, Velasco TR, Araujo Rodrigues MC, Wichert-Ana L, et al. Is dystonic posturing during temporal lobe epileptic seizures the expression of an endogenous anticonvulsant system? Epilepsy Behav 2008;12:39–48.

P. Bertti et al. / Epilepsy & Behavior 38 (2014) 81–93 [47] Chee MW, Kotagal P, Van Ness PC, Gragg L, Murphy D, Lüders HO. Lateralizing signs in intractable partial epilepsy: blinded multiple-observer analysis. Neurology 1993; 43:2519–25. [48] Mizobuchi M, Matsuda K, Inoue Y, Sako K, Sumi Y, Chitoku S, et al. Dystonic posturing associated with putaminal hyperperfusion depicted on subtraction SPECT. Epilepsia 2004;45:948–53. [49] Meletti S, Cantalupo G, Stanzani-Maserati M, Rubboli G, Alberto Tassinari C. The expression of interictal, preictal, and postictal facial-wiping behavior in temporal lobe epilepsy: a neuro-ethological analysis and interpretation. Epilepsy Behav 2003;4: 635–43. [50] Bustamante VCT, Sakamoto AC, Yacubian EMT, Garzon E. Epilepsias do lobo temporal: características clínicas e semiologia crítica ao vídeo-EEG. Vídeo-Eletrencefalografia Fundam. E Apl. Nas Epilepsias. Lemos; 1999. [51] Hirsch LJ, Lain AH, Walczak TS. Postictal nosewiping lateralizes and localizes to the ipsilateral temporal lobe. Epilepsia 1998;39:991–7. [52] Leung LS, Ma J, McLachlan RS. Behaviors induced or disrupted by complex partial seizures. Neurosci Biobehav Rev 2000;24:763–75. [53] Geyer JD, Payne TA, Faught E, Drury I. Postictal nose-rubbing in the diagnosis, lateralization, and localization of seizures. Neurology 1999;52:743–5. [54] Quesney LF, Constain M, Fish DR, Rasmussen T. The clinical differentiation of seizures arising in the parasagittal and anterolaterodorsal frontal convexities. Arch Neurol 1990;47:677–9. [55] Chen L, Yang X, Liu Y, Zeng D, Tang Y, Yan B, et al. Quantitative and trajectory analysis of movement trajectories in supplementary motor area seizures of frontal lobe epilepsy. Epilepsy Behav 2009;14:344–53. [56] Li Z, Martins da Silva A, Cunha JPS. Movement quantification in epileptic seizures: a new approach to video-EEG analysis. IEEE Trans Biomed Eng 2002;49:565–73. [57] Garcia-Cairasco N, Dal-Cól ML, Bertti P. Quantitative movement trajectory analysis and neuroethology in clinical epileptology. Epilepsy Behav 2009;15: 266–7. [58] Paramanathan P, Uthayakumar R. Application of fractal theory in analysis of human electroencephalographic signals. Comput Biol Med 2008;38:372–8. [59] Raiesdana S, Hashemi Golpayegani MR, Nasrabadi AM, Nasrabadi AM. Complexity evolution in epileptic seizure. Conf Proc Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Conf 2008:4110–3. [60] Stam CJ, Jones BF, Nolte G, Breakspear M, Scheltens P. Small-world networks and functional connectivity in Alzheimer's disease. Cereb Cortex 2007;17:92–9. [61] Wagenmakers E, van der Maas HLJ, Farrell S. Abstract concepts require concrete models: why cognitive scientists have not yet embraced nonlinearly coupled, dynamical, self‐organized critical, synergistic, scale‐free, exquisitely context‐sensitive, interaction‐dominant, multifractal, interdependent brain–body–niche systems. Top Cogn Sci 2012;4:87–93.

93

[62] Bollobás B. Modern graph theory. Springer; 1998. [63] Bullmore E, Sporns O. The economy of brain network organization. Nat Rev Neurosci 2012;13:336–49. [64] Bernhardt BC, Hong S, Bernasconi A, Bernasconi N. Imaging structural and functional brain networks in temporal lobe epilepsy. Front Hum Neurosci 2013;7:624. [65] Xue K, Luo C, Zhang D, Yang T, Li J, Gong D, et al. Diffusion tensor tractography reveals disrupted structural connectivity in childhood absence epilepsy. Epilepsy Res 2014;108:125–38. [66] Kim J-Y, Kang H-C, Kim K, Kim HD, Im C-H. Localization of epileptogenic zones in Lennox–Gastaut syndrome (LGS) using graph theoretical analysis of ictal intracranial EEG: a preliminary investigation. Brain Dev 2014. http://dx.doi.org/10.1016/ j.braindev.2014.02.006. [67] Srinivas KV, Jain R, Saurav S, Sikdar SK. Small-world network topology of hippocampal neuronal network is lost, in an in vitro glutamate injury model of epilepsy. Eur J Neurosci 2007;25:3276–86. [68] Hagmann P, Cammoun L, Gigandet X, Meuli R, Honey CJ, Wedeen VJ, et al. Mapping the structural core of human cerebral cortex. PLoS Biol 2008;6:e159. [69] Garcia-Cairasco N. Puzzling challenges in contemporary neuroscience: insights from complexity and emergence in epileptogenic circuits. Epilepsy Behav 2009; 14(Suppl. 1):54–63. [70] Latora V, Marchiori M. Efficient behavior of small-world networks. Phys Rev Lett 2001;87:198701. [71] Gong X-W, Li J-B, Lu Q-C, Liang P-J, Zhang P-M. Effective connectivity of hippocampal neural network and its alteration in Mg2+-free epilepsy model. PLoS One 2014; 9:e92961. [72] Bartolomei F, Bettus G, Stam CJ, Guye M. Interictal network properties in mesial temporal lobe epilepsy: a graph theoretical study from intracerebral recordings. Clin Neurophysiol 2013;124:2345–53. [73] Kramer MA, Kolaczyk ED, Kirsch HE. Emergent network topology at seizure onset in humans. Epilepsy Res 2008;79:173–86. [74] Achard S, Salvador R, Whitcher B, Suckling J, Bullmore E. A resilient, low-frequency, small-world human brain functional network with highly connected association cortical hubs. J Neurosci Off J Soc Neurosci 2006;26:63–72. [75] Bernhardt BC, Chen Z, He Y, Evans AC, Bernasconi N. Graph-theoretical analysis reveals disrupted small-world organization of cortical thickness correlation networks in temporal lobe epilepsy. Cereb Cortex 2011;21:2147–57. [76] Zhang Z, Liao W, Chen H, Mantini D, Ding J-R, Xu Q, et al. Altered functional– structural coupling of large-scale brain networks in idiopathic generalized epilepsy. Brain J Neurol 2011;134:2912–28.