Magnetoencephalographic spike analysis in patients with focal cortical dysplasia: What defines a 'dipole cluster'?

Abstract

Background

The purpose of this study is to clarify the source distribution patterns of magnetoencephalographic (MEG) spikes correlated with postsurgical seizure-free outcome in pediatric patients with focal cortical dysplasia (FCD).

Methods

Thirty-two patients with pathologically-confirmed FCD were divided into seizure-free and -persistent groups according to their surgical outcomes based on Engel's classification. In each patient, presurgical MEG was retrospectively reviewed. Dipole sources of MEG spikes were calculated according to a single dipole model. We obtained the following quantitative indices for evaluating dipole distribution: maximum distance over all pairs of dipoles, standard deviation of the distances between each dipole and the mean coordinate of all dipoles, average nearest neighbor distance, the rate of dipoles located within 10 mm, 20 mm, 30 mm from the mean coordinate, and the rate of dipoles included in the resection. These indices were compared between the two patient groups.

Results

Average nearest neighbor distance was significantly smaller in the seizure-free group compared to the seizure-persistent group (p=0.008). The rate of dipoles located within 10 mm, 20 mm, 30 mm from the mean coordinate were significantly higher in the seizure-free group (p=0.001, 0.001, 0.005, respectively). The maximum distance, standard deviation and the resection rate of dipoles did not show a significant difference between the two groups.

Conclusions

A spatially-restricted dipole distribution of MEG spikes is correlated with postsurgical seizure-free outcomes in patients with FCD. The distribution can be assessed by quantitative indices that are clinically useful in the presurgical evaluation of these patients.

Introduction

Focal cortical dysplasia (FCD) is a major cause of medically-intractable epilepsy in the pediatric population but may be amenable to surgical treatment^{1, 2}. Previous studies have shown that complete removal of the anatomical/electrophysiological abnormality is an important prognostic factor of postsurgical seizure freedom^{3, 4}. Therefore, investigation of epileptic discharges, including interictal spikes, is critical for planning epilepsy surgery in pediatric patients with FCD.

Magnetoencephalography (MEG) is a non-invasive tool that records neuromagnetic fields from the brain, and is useful for localizing epileptic discharges in presurgical evaluation. Previous studies have investigated interictal MEG spikes by using a single dipole model in FCD patients, showing the intrinsic epileptogenicity of the lesion as compared with MRI and intracranial EEG^5–9. Several researchers observed dipole distribution in a restricted cortical area, which is called a 'dipole cluster'^{10, 11}. They suggested that removal of the cluster is correlated with a favorable surgical outcome^{10, 11}, however, postsurgical seizure-freedom is not always achieved. In the past studies, the dipole cluster was subjectively detected by visual inspection of dipole distribution maps^{8, 9}, or by applying predefined criteria, such as 'six or more spike sources with 1 cm or less between adjacent sources'^{10, 11}. The basis of these criteria is still unclear: no studies have revealed how closely the dipoles should be located for determining the dipole cluster that is clinically relevant. Revisiting the concept of dipole cluster beyond the subjective, predefined criteria is necessary for better planning of epilepsy surgery.

The purpose of this study is to objectively and quantitatively reveal the dipole distribution that is useful for estimating postsurgical outcomes. We investigate 1) the spatial patterns of dipole distribution by using numerical indices, and 2) the relation of these indices with the surgical outcome in patients with FCD. We hypothesize that these indices showing a spatially-restricted dipole distribution are correlated with postsurgical seizure freedom and provide the basis of clinically-relevant dipole cluster.

Patients and Methods

Patients

We retrospectively studied 32 pediatric patients (18 males, 14 females, age 1–18, mean 11 years old) who underwent an MEG as a part of clinical evaluation and subsequent epilepsy surgery in 2003–2016. All patients had a histopathological diagnosis of FCD. There were 11 patients diagnosed according to the classification proposed by Palmini et al.¹² (Patients 4–7, 9, 18, 21–25), and 11 patients with the International League Against Epilepsy classification¹³ (Patients 8, 10–13, 15–16, 26, 28–29, 32). The pathology reports of 10 patients (Patients 1–3, 14, 17, 19–20, 27, 30–31) only indicated FCD without further stratification. We included patients with isolated FCD and with associated principal lesions, such as encephalomalacia, infarction and cyst, therefore, the patients were not characterized by a single pathology. Six patients had previous surgery, which showed cortical dysplasia/malformation without further classification (Patients 15, 17, 25, 27 and 32) and ganglion cell tumor (Patient 20) in pathology.

Patients were post-operatively followed to 12–92 months (mean 32 months). Surgical outcomes were evaluated by Engel's classification at the time of last follow-up¹⁴, and we divided the patients into two groups: Seizure-free (Class Ia outcome, 16 patients) and seizure-persistent (other outcomes, 16 patients). All aspects of the study were approved by the institutional review board and were performed in accordance with the Declaration of Helsinki. Informed consent was obtained from all patients and their guardians. Table 1 gives an overview of the clinical profiles of the patients.

Table 1.

Patient profile

Patient	Age/Sex	Diagnosis	Scalp EEG (Interictal/Ictal)	MEG	MRI	Resection	Pathology	Surgical outcome
1	13/M	L TLE	LT/LT	LT	Arachnoid cyst+CD in LT(ant)	LT(ant)	CD	Free
2	9/F	L TLE	LT/L hemi	LT	Abnormal T2 signal in LT(inf)	LT(inf)	Low-grade glioma+CD	Free
3	11/F	R FLE	BiF/RF	RF	CD in RF(mes)	RF(mes)	CD	Free
4	11/M	R FLE	RF/Diffuse	RF	CD in RF(mes)	RF(mes)	Palmini IIB	Free
5	18/F	R TLE	RT/RT	RT	CD in RT(mes)	RT(ant+mes)	Palmini IA	Free
6	11/M	L TLE	LT/LF,LT	LT	L MTS, CD in LT	LT(ant)	MTS+Palmini IIA	Free
7	10/M	L TLE	LT/LT,LC,LP	LT	L MTS, L hemi atrophy	LT(ant)	MTS+Palmini IIA	Free
8	13/M	Multifocal	RF,RP/RF	RF,RT	Dysmorphic cortex in RT(pos), PMG in R hemi	R functional hemispherectomy	ILAE IIId	Free
9	16/F	L TLE	LT/LT	LT	Heterotopia, dysmorphic T2 hyperintense in LT,LO	LT(ant+pos)	Palmini IIA	Free
10	13/F	L TLE	LT/LT	LT	Blurring in LT(ant)	LT(ant)	ILAE IIa	Free
11	9/M	R OLE	RO/RO	RO	Encephalomalacia in RO	RO	ILAE IIId	Free
12	7/F	R TLE	RF,RT/RF,RT	RT	CD in RT(inf)	RT(inf)	Ganglioglioma+ILAE IIIb	Free
13	7/M	R FLE	RF,RT/RF,RT	RF	CD in RF(preF)	RF(preF)	ILAE IIb	Free
14	5/M	R FLE	RF,RC,RT/RF,RC,RT	RF,RP	normal	RF+RP	CD	Free
15	2/F	L TLE	LF,LT/LF,LT	LF,LT	Previous resection, residual CD in LT(sup)	LT(sup)	ILAE IIa	Free
16	3/F	L FLE	LF/Diffuse	LF	CD in LF(preF)	LF(preF)	ILAE IIb	Free
17	12/M	L TLE	LT,LP/LT,LC,LP	LT	Previous resection	LT+LP	CD	Persistent
18	14/M	R FLE	RF,RC/BiF,RC	RF,RT	CD in RF(preC)	RF(preC)	Palmini IA	Persistent
19	16/F	L TLE	LT/LT	LT	L MTS	LT(ant)	MTS+CD	Persistent
20	15/M	L TLE	LT/Diffuse	LT,LP	Previous resection in LF(preC)	LT(ant)	CD	Persistent
21	15/M	R FLE	RF/RF	RF	CD in RF(sup)	RF(sup)	Palmini IIB	Persistent
22	15/F	R TLE	RT/RF,RT	RT	normal	LT(ant)	Palmini IA	Persistent
23	12/M	R TLE	RT/RF,RC,RT	RT	T2 hyperintense in RT(ant)	RT(ant)	Palmini IIA	Persistent
24	15/M	L TLE	LF,LT/LF,LT	LT	normal	LT(ant)	Palmini IA	Persistent
25	17/F	L TLE	LT,LP/LT,LP	LT,LP	Previous resection in LT	LT(pos)	Palmini IA	Persistent
26	8/F	R FLE	RF,RC/BiF,BiC,BiP	RF	Lesion in RF(mes)	RF(mes)	ILAE IIb	Persistent
27	11/M	L TLE	BiF,BiT/LC	BiF,BiT	Previous resection in LT(pos),LO	LT(ant)	CD	Persistent
28	12/M	L TLE	LT/LT	LF,LT,LP	Previous infarction in LF,LT	LF+LT(ant)	ILAE IIId	Persistent
29	6/F	R FLE	RF,RT/Diffuse	RF,RT,RP	CD in RF(inf)	RF(inf)	ILAE IIb	Persistent
30	1/F	R FLE	R hemi/R hemi	R hemi	Cortical malformation in RF	R functional hemi spherectomy	CD	Persistent
31	17/M	L TLE	LT/LT	LT	Small cystic foci in L hippocampal region	LT(inf)	Glioneuronal tumor+CD	Persistent
32	12/M	L FLE	BiF,LT/LF,LT	BiF,LT	Previous resection in LT(ant), blurring in LT(inf)	LF (pre)+LT (inf)	ILAE IIa	Persistent

M: Male, F: Female, L: Left, R: Right, Bi: Bilateral, FLE: Frontal lobe epilepsy, TLE: Temporal lobe epilepsy, L(R)F: Left(Right) frontal, L(R)T: Left(Right) temporal, L(R)C: Left(Right) central, L(R)P: Left(Right) parietal, L(R)O: Left(Right) occipital, hemi: hemisphere, CD: Cortical dysplasia, MTS: Mesial temporal sclerosis, PMG: Polymicrogyria, ant: anterior, pos: posterior, sup: superior, inf: inferior, mes: mesial, preF: prefrontal, preC: precentral, Free: Seizure-free, Persistent: Seizure-persistent, also shown by the gray background.

MEG recording

MEG was recorded with a 306-channel, whole-head MEG system (Elekta-Neuromag, Helsinki, Finland). The sampling frequency was 600 Hz (Patients 1–8, 17–27) or 1000 Hz (Patients 9–16, 28–32) with a band-pass filter of 0.1–200 Hz. We recorded spontaneous activity for 50–60 min in each patient. Patients were recorded in supine position and instructed to rest or sleep. Antiepileptic medications were maintained without tapering and no sedation was performed at the time of MEG study. We collected scalp EEG simultaneously with MEG by using a 70-channel electrode cap. The EEG findings are shown in Table 1. The data were low-pass filtered at 40 Hz for the analysis. The details of the MEG recording have previously been described¹⁵.

In all patients, anatomical MRI data were acquired with magnetization-prepared rapid acquisition gradient-echo sequences (MPRAGE; TE=3.37 ms, TR=2000 ms, voxel size=1×1×1 mm) with a high-resolution 3T scanner (TIM TRIO, Siemens AG, Erlangen, Germany). Post-surgical MRI was also obtained with MPRAGE, T1- or T2-axial/coronal/sagittal sequences.

MEG analysis

We visually examined MEG data and identified interictal spikes. Equivalent current dipoles (ECDs) were calculated at the peak of each spike using a single-dipole model without selecting a region of interest (i.e., all 306 sensors were used for the analysis). ECDs with goodness of fit >70 % and dipole moment <500 nAm were considered adequate as spike sources. The mean coordinate of all ECDs was obtained in each patient. Then we calculated the following indices for evaluating the spatial distribution of ECDs: (1) maximum distance: the largest distance over all pairs of ECDs; (2) SDD: standard deviation of the distances between each ECD and the mean coordinate; (3) average nearest neighbor distance: mean distance between each ECD and its nearest ECD; (4)–(6) within 10 mm, within 20 mm and within 30 mm: the rate of ECDs located within 10 mm, 20 mm and 30 mm from the mean coordinate, respectively; (7) resection rate: the rate of ECDs located in the resection. We coregistered the postsurgical MRI to the presurgical images and visually determined whether each ECD was removed or not.

Statistics

We compared all seven indices as defined above between the seizure-free and - persistent groups by using Mann-Whitney tests. Since the numbers of ECDs are different according to the number of spikes recorded in each patient, we tested the correlations between the numbers of ECDs and indices (1)–(7) by means of Spearman's correlation coefficient for understanding its effect on these indices. We also investigated the correlations between (1)–(6) and the resection rate to test whether the ECD distribution affects the removal of ECDs. P value<0.01 was considered significant.

Results

Table 2 summarizes the results of each patient. The average nearest neighbor distance was significantly smaller in seizure-free group than in seizure-persistent group (p=0.008). The seizure-free group also showed larger values of within 10 mm, within 20 mm, within 30 mm (p=0.001, 0.001, 0.005). We did not find any significant difference in other indices between these patients groups.

Table 2.

Results

Patient	Number of ECDs	Mean coordinates (x, y, z)	Maximum distance	SD	Nearest	Within 10 mm	Within 20 mm	Within 30 mm	Resection rate
1	44	−44.1, 17.6, 19.2	40.1	6.2	3.1	0.75	0.98	1.00	0.95
2	28	−47.4, −6.5, 38.3	60.1	12.2	7.7	0.32	0.61	0.86	0.61
3	59	22.7, 63.5, 60.7	34.4	5.4	2.9	0.75	0.98	1.00	0.88
4	47	35.9, 34.1, 62.6	111.6	20.3	7.7	0.32	0.74	0.89	0.00
5	40	31.8, 10.5, 21.4	82.8	15.1	7.7	0.23	0.50	0.80	0.70
6	16	−44.7, 2.1, 47.8	68.8	15.1	10.7	0.19	0.81	0.94	0.50
7	21	−38.3, − 11.1, 39.1	95.6	19.3	10.9	0.00	0.14	0.52	0.43
8	18	31.5, −18.1, 51.2	109.9	33.2	10.6	0.00	0.50	0.78	0.17
9	108	−38.8, 3.2, 40.6	120.7	19.7	5.3	0.06	0.47	0.84	0.49
10	57	−43.6, 13.0, 21.6	49.4	8.5	4.4	0.49	0.88	0.98	0.77
11	34	19.7, −53.9, 64.7	54.6	11.0	6.8	0.18	0.71	0.94	0.88
12	23	42.7, −10.1, 44.0	41.0	7.7	6.8	0.39	0.91	1.00	0.35
13	65	16.5, 50.8, 52.6	95.1	21.8	6.0	0.11	0.52	0.83	0.49
14	33	52.5, −9.5, 77.1	80.7	13.3	7.1	0.36	0.82	0.88	0.79
15	59	−38.7, 6.1, 46.5	64.6	11.1	4.6	0.32	0.83	0.92	0.32
16	207	−38.4, 45.3, 63.1	44.1	6.4	2.3	0.65	0.97	1.00	0.79
17	21	−28.0, −1.8, 69.2	102.6	28.4	12.4	0.05	0.10	0.33	0.19
18	19	26.8, 29.8, 91.9	94.8	22.3	8.0	0.05	0.47	0.84	0.32
19	14	−41.2, 9.4, 33.3	70.8	16.4	11.5	0.00	0.21	0.64	0.50
20	26	−40.6, 0.5, 50.7	79.4	17.6	12.0	0.04	0.38	0.69	0.46
21	23	15.8, 34.3, 103.4	59.9	11.8	6.7	0.17	0.61	0.96	0.48
22	29	45.4, 5.1, 32.8	51.1	9.7	7.4	0.14	0.72	0.97	0.55
23	82	49.6, 2.8, 44.7	86.8	11.4	5.6	0.26	0.78	0.95	0.23
24	28	−36.7, −3.6, 39.6	79.0	14.9	8.3	0.29	0.75	0.93	0.43
25	21	−28.8, − 15.1, 74.1	98.7	21.5	15.2	0.05	0.48	0.86	0.57
26	13	26.2, 36.7, 63.1	81.6	16.3	14.5	0.08	0.46	0.62	0.15
27	123	−12.8, 18.3, 60.9	125.7	47.0	3.7	0.00	0.00	0.02	0.03
28	75	−39.6, −9.0, 64.9	84.1	15.1	6.5	0.11	0.49	0.73	0.44
29	14	38.1, 19.7, 58.5	68.4	15.1	12.2	0.00	0.36	0.71	0.00
30	112	36.6 −19.9, 52.9	87.5	21.7	4.9	0.02	0.22	0.63	0.34
31	5	−45.5, − 23.0, 38.5	50.7	15.7	17.5	0.20	0.80	0.80	0.80
32	8	−8.7, 21.9, 68.0	99.3	32.2	13.5	0.00	0.00	0.00	0.25

ECDs: Equivalent current dipoles, SD: Standard deviation of the distances between each ECD and the mean coordinate, Nearest: Average nearest neighbor distance of ECDs, Within 10 mm, 20 mm, 30 mm: Rate of ECDs within 10 mm, 20 mm, 30 mm from the mean coordinate. Patients in the seizure-persistent group are shown by the gray background

The average nearest neighbor distance was correlated with the number of ECDs (Rs=−0.88, p<0.001). There were also correlations seen in the maximum distance (Rs=− 0.62, p<0.001), SDD (Rs=−0.59, p<0.001), within 10 mm (Rs=0.54, p<0.001), within 20 mm (Rs=0.55, p=0.001) and within 30 mm (Rs=0.53, p=0.002) with the resection rate.

Fig. 1 shows the typical patterns of dipole distribution in seizure-free and - persistent groups. Fig. 2 plots the rate of seizure-free patients to within 10 mm, within 20 mm, within 30 mm (Fig. 2-upper) and the average nearest neighbor distance (Fig. 2-lower). The trend shows more dipoles within 10 mm, 20 mm, 30 mm from the mean coordinate result in a higher rate of seizure-free patients, and larger average nearest neighbor distances result in a lower seizure-free rate.

Figure 1.

(A) Dipole distribution of a patient in the seizure-free group (Patient 16), projected on the presurgical (Left) and postsurgical (Right) MRI. Most dipoles are tightly clustered (average nearest neighbor distance=2.3 mm, within 10 mm=0.65) and located within the resection. (B) Dipole distribution of a patient in the seizure-persistent group (Patient 23), projected on the presurgical (Left) and postsurgical (Right) MRI. Most dipoles are loosely clustered (average nearest neighbor distance=5.6 mm, within 10 mm=0.26) and located outside the resection.

Figure 2.

(Upper) The rate of postsurgical seizure-free patients (%) is plotted corresponding to the threshold of within 10 mm (circle), within 20 mm (square), within 30 mm (triangle). The trend shows that larger values of within 10 mm, within 20 mm and within 30 mm result in a higher rate of seizure-free patients. (Lower) The rate of postsurgical seizure-free patients (%) is plotted corresponding to the threshold of average nearest neighbor distance. The trend shows that larger average nearest neighbor distances result in a lower seizure-free rate.

Discussion

In this study, we investigated the spatial patterns of dipole source distribution obtained from interictal epileptiform discharges recorded using MEG in patients with FCD. The results suggest that a spatially-restricted dipole distribution is correlated with seizure-free surgical outcome, as shown by the significantly smaller average nearest neighbor distance and the higher rate of dipoles located within 10 mm, 20 mm and 30 mm from the mean coordinate in the seizure-free group compared to the seizure-persistent group of patients.

Previous MEG studies have observed a pattern of dipole distribution in a restricted cortical area, which is described as a 'dipole cluster' in patients with epilepsy^16–20. Dipole clusters are considered to guide a resection volume and predict a good outcome in epilepsy surgery²¹. However, these studies predefined the criteria to determine a dipole cluster variably, as described by 'six or more spike sources with ≤1 cm between adjacent sources'^{10, 11, 16, 18, 22–28}, '10 or more ECDs located contiguously within neighboring gyri'²⁹, '10 or more spike sources with 15 mm or less between adjacent sources'¹⁹, or 'at least 5 dipoles within a 1-cm² region'^{21, 30}. No studies have demonstrated objective and quantitative basis of these criteria. Moreover, there is evidence suggesting that FCD represents specific features in MEG spikes and their source distribution, as compared with astrocytic inclusions²⁷ or other lesions³¹. The patterns of dipole clusters may be useful for diagnosing FCD distinctive from other etiologies when they are defined based on quantitative observation. Recent studies classified the dipole cluster into 'tight' and 'loose' clusters at a lobar/sublobar level and suggested that tight clusters are characteristic in FCD patients^{20, 32}. Our results may be informative for quantitatively determining the criteria of a dipole cluster that is clinically relevant in patients with FCD by using numerical indices of dipole distribution. More specifically, the mean values of average nearest neighbor distance were 6.5 mm and 10.0 mm in seizure-free and -persistent groups, respectively. The criteria of dipole clusters, such as '≤1 cm between adjacent sources' and '15 mm or less between adjacent sources', include most of the dipole distributions seen in both of our patient groups, and may not be useful for estimating postsurgical outcomes. In fact, the rate of seizure-free patients is only 62 % and 55 % at the threshold of 10 mm and 15 mm for average nearest neighbor distance in our patients (Fig. 2-lower). Considering the average nearest neighbor distance was correlated with the number of dipoles, the distance between dipoles may not be appropriate for the criteria of dipole cluster. Alternatively, all patients who had 30 %, 90 % and 100 % of dipoles within 10 mm, 20 mm and 30 mm from the mean coordinate became seizure-free after surgery (Fig. 2), suggesting that these indices provide a strong basis of clinically-relevant dipole cluster and a good indicator for estimating postsurgical outcome.

Several researchers have demonstrated that surgical removal of dipole cluster may predict favorable surgical outcomes in patients with epilepsy, by investigating whether the dipole cluster was completely or partially removed^{17, 21, 33–36}. A few studies suggested a higher rate of dipole removal is correlated with favorable surgical outcomes^{37, 38}, while Kim et al.³⁹ reported no statistical relationship between these two measurements. For FCD patients, Widjaja et al.¹⁰ found that the complete removal of a dipole cluster had a higher rate of achieving Engel's Class I outcome than partial removal; however Wilenius et al.¹¹ did not find a significant correlation between the resection rate of dipoles and seizure-free outcomes in patients with dipole clusters.

Our results failed to show a relationship between resection rate and postsurgical seizure-freedom. Mislocalization of MEG spike sources may occur due to spike propagation⁴⁰, leading to a low rate of dipole resection in some patients of our seizure-free group. On the other hand, the resection rate was correlated with the dipole distribution pattern as indicated by maximum distance, SDD, within 10 mm, within 20 mm and within 30 mm, suggesting that a restricted dipole distribution correlates with a higher resection rate. The results also differ depending on other factors, such as the existence of MRI-visible lesions³⁹; thus, the significance of dipole removal is still unclear.

We did not consider the size of the patient's brain in our analysis, although it is different depending on the patient's age. The brain size affects the correlation between the dipole distribution and anatomical regions. For example, a certain distance, such as 30 mm, may cross multiple gyri in younger patients, while it may be included in a single gyrus in older patients. Applying an anatomical atlas to the source space would be useful for counting the numbers of dipoles at a lobar or sublobar level and evaluating the dipole distribution correlated with the anatomical regions. Similarly, the brain size may be a factor affecting the resection rate, however, our patients did not show a significant difference in the resection rate between seizure-free and -persistent groups even without patients under 5 years of age (Patients 15, 16 and 30, P>0.01). There are only a small number of young pediatric patients who show considerably small size of the brain in our study. Further investigations of such patients will clarify the effect of brain size on the clinical relevance of dipole distribution.

This study has several limitations. First, the study design is retrospective, and a prospective study is necessary for controlling the confounding factors. Second, we did not consider the pathological classification of FCD, since the pathology findings were obtained by different criteria. Moreover, our study included patients with different types of FCD pathology, such as isolated and associated with principal lesions. Previous studies have suggested an impact of FCD type in the clinical outcome^{26, 32}. Third, we investigated the dipole distribution regardless of its location. Bilateral or distant dipole locations may reflect multiple dipole clusters²⁹, in which a single mean coordinate is not feasible for analysis. Nakajima et al.²⁸ reported the different patterns of dipole distribution in FCD patients depending on the lesion location at the bottom of the sulcus or gyral surface. Further studies are necessary for addressing these issues.

In conclusion, investigating the distribution patterns of spike dipole sources is informative for understanding the clinical usefulness of MEG in patients with FCD. Dipole source distribution in a restricted area suggests favorable outcomes of epilepsy surgery.