EEG entropy analysis in autistic children

Journal of Clinical Neuroscience xxx (xxxx) xxx

Contents lists available at ScienceDirect

Journal of Clinical Neuroscience journal homepage: www.elsevier.com/locate/jocn

Clinical study

EEG entropy analysis in autistic children Jiannan Kang a, Huimin Chen b, Xin Li c,d, Xiaoli Li e,⇑ a

College of Electronic & Information Engineering, Hebei University, Baoding, China Center for Brain Disorders Research, Capital Medical University, Beijing Institute for Brain Disorders, Beijing, China c Institute of Biomedical Engineering, Yanshan University, Qinhuangdao, China d Key Laboratory of Measurement Technology and Instrumentation of Hebei Province, Qinhuangdao, China e State Key Laboratory of Cognitive Neuroscience and Learning and IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China b

a r t i c l e

i n f o

Article history: Received 20 July 2018 Accepted 5 November 2018 Available online xxxx Keywords: Autism Children EEG Neural connectivity Entropy

a b s t r a c t Autism spectrum disorder (ASD) is a complex and heterogeneous neurodevelopmental disorder, which is characterized by impairments of social interaction and communication, and by stereotyped and repetitive behaviors. Extensive evidences demonstrated that the core neurobiological mechanism of autism spectrum disorder is aberrant neural connectivity, so the entropy of EEG can be applied to quantify this aberrant neural connectivity. In this study, we investigated four entropy methods to analyse the resting-state EEG of the autistic children and the typical development (TD) children. Through 43 children diagnosed with autism aged from 4 to 8 years old as compared to 43 normal children matched for age and gender, we found region-specifically and entropy-specifically which were more sensitive with the increase of age. In detail, for 4 years old group, there is significant difference in central by Renyi permutation entropy method; the significant differences are in frontal and central by sample entropy for 5 years old group; the significant difference is in frontal by fuzzy entropy for 6 years old group; the significant difference is in central by Renyi wavelet entropy for 7 years old group and the difference is in occipital by Renyi wavelet entropy for 8 years old group. The results might guide us to make an accurate distinction between ASD and TD children. Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction Autism spectrum disorder (ASD) is a complex and heterogeneous neurodevelopmental disorder, which is characterized by impairments of social interaction and communication, and by stereotyped and repetitive behaviors [1]. The estimated prevalence is about 1 in 45 in the 2014 National Health Interview Survey, there is a significant increase within five years [2]. The amazing number rise of ASD forces us to develop some noninvasive neuroimaging methods, such as electroencephalography (EEG), magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI), to measure the neural activities of children to indicate the abnormal brain development. Functional connectivity refers to the degree of which activity in one region correlates with another or to the synchronization degree of two brain regions during the performance tasks or resting-states [3]. Segregation and integration of information processing in the brain can be mediated by means of functional connectivity [4,5], so the functional connectivity can reflect ⇑ Corresponding author. E-mail address: [email protected] (X. Li).

characteristic changes due to various neurological and psychiatric disorders. At present, the aberrant functional connectivity is a very important concept to study underlying mechanism of the disturbed cognition disorders [6–9]. Some previous studies showed that one core neurobiological mechanism of ASD symptoms involves aberrant functional connectivity (local over-connectivity and global under-connectivity) [10,11]. In this study, the entropy of EEG was applied to quantify this aberrant neural connectivity. It is known that EEG complexity can reflect the irregularity or unpredictability of the brain activity [12]. On the other hand, we found that EEG complexity can be used to indicate functional connectivity [13–15], in general EEG complexity is higher, the stronger of brain functional connectivity. The entropy is an important method to measure EEG complexity for monitoring functional states of brain [16,17], also neurological and neuropsychiatric diseases including epilepsy, Parkinson’s disease, Alzheimer’s disease and schizophrenia [18–21]. In previous studies, some entropy analysis methods were proposed to analyze the EEG recording from ASD as well. A modified multiscale entropy method was used to analyze the EEG data of infants at ages 9–12 months with high risk for autism, EEG entropy-based potential biomarker for ASD was then demonstrated by the support vector machine algorithm

https://doi.org/10.1016/j.jocn.2018.11.027 0967-5868/Ó 2018 Elsevier Ltd. All rights reserved.

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

2

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

[22]. Shannon entropy of EEG recordings can be used to differentiate the normal and autistic children via a classifier [23], and the sample entropy could indicate the brain activity characteristics of the children with autism during complex environments [24]. We have also calculated Renyi entropy of EEG recordings with ASD children, and found a significantly diminished global synchronization from ASD to controls [25]. It is noted that the functional connectivity is also associated with the brain development. Therefore, we need to consider the effect of the age on EEG entropy of ASD children [26], so that the EEG entropy can be better to become a potential biomaker for diagnosis of ASD children. Different entropy methods are based on different physical hypothesises, so we need to demonstrate which entropy method is better to indicate the characteristics of EEG recordings from ASD. At the same time, we need to clarify the EEG entropy of which brain area is more sensitive to differentiate ASD children from normal children.

In this study, we intend to explore the EEG entropy for making a better discrimination for ASD and TD children using four entropy methods including sample entropy (SampEn), fuzzy entropy (FuzzyEn), Renyi wavelet entropy (RWE) and Renyi permutation entropy (RPE). Then, we investigate the effect of age to entropy value in ASD and TD. Finally, we demonstrate which brain areas are significant to differentiate the ASD from TD by special entropy method. 2. Materials and methods 2.1. Participants EEG data were collected from 86 children: 43 children with ASD including 31 male and 12 female (mean age = 5.81 years old; age range is from 4 to 8 years old) and 43 typical development children

Fig. 1. The accumulative plot of p value by four entropy methods for ASD and TD. (A) shows the accumulated p value in the brain area of 4 years old group by four entropy; (B) shows the accumulated p value in the brain area of 5 years old group by four entropy; (C) shows the accumulated p value in the brain area of 6 years old group by four entropy; (D) shows the accumulated p value in the brain area of 7 years old group by four entropy; (E) shows the accumulated p value in the brain area of 8 years old group by four entropy.

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

3

matched with gender and age, 31 male and 12 female (mean age = 5.77 years old; age range is from 4 to 8 years old) with no family history of neurodevelopmental disorders. All autistic children were diagnosed by a specialist at Chinese PLA General Hospital. All parental written informed consent was obtained after the parents were informed of the whole experimental procedure before participation and the experiment was approved by the Beijing Normal University ethics committee and was carried out in compliance with the ethical standards established by the Declaration of Helsinki.

impedance was greater than 50 kX or the signal amplitude was more than 200 lV; (2) Independent component analysis (ICA) algorithm was used to remove eye movements, blink, muscle movements and other artifacts; (3) the data was band-pass filtered between 0.5 and 45 Hz; (4) the average reference was done for all channels and down sampling to 256 Hz; (5) the data length was divided into 4 s with no overlap. We chose 19 electrodes (the standard international 10–20 electrode placement: Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T7, T8, P7, P8, Fz, CPz, and Pz) for analysis.

2.2. EEG recordings

2.4. Entropy methods

EEG was recorded in a quiet room, with the participants being awake, seated on a comfortable chair and relaxed in eyes-open state. During the EEG recording, we tried to keep the children quiet and reduce some activities, such as head shaking, teeth gritting and feet moving. Each participant’s EEG was recorded for about 5 min. A 128-channel HydroCel Sensor Net System (Electrical Geodesics, Inc) was used for data collection and the EEG sampling rate was 1000 Hz. We set the central vertex as the referential electrode and the impedances were below 50 kX when the data was collecting.

2.4.1. Sample entropy (SampEn) The approximate entropy was proposed by Pincus in 1991 [27] and it is used for analyzing the finite length signal describing the randomness and unpredictability. The sample entropy is derived from the approximate entropy but different from it by eliminating self-matching. Since the comparison of its self-data segment is not included, the data length is not strictly limited. The additional advantage is the consistency against the change of the embedded dimension and tolerance level [28]. The details on the calculation of sample entropy can be found in [29].

2.3. Data preprocessing

2.4.2. Fuzzy entropy (FuzzyEn) Fuzzy entropy, which was proposed based on the theory of fuzzy mathematics, is a nonlinear index to evaluate the probability of newly generated modes. In 2007, Chen et al. [30] performed

Data preprocessing was based on Matlab R2016a and EEGLab: (1) the bad channel was replaced by the adjacent channels if the

Fig. 2. (A) Renyi permutation entropy is more effective for 4 years old group to indicate the difference of ASD and TD; (B) The significant difference of topographic map between ASD and TD for 4 years old group; (C) C4 (t = 5.8182, p = 0.0007), C3 (t = 4.7089, p = 0.0022) and Fz (t = 4.7141, p = 0.0022) have significant differences; P3 (t = 3.1445, p = 0.0163) and F3 (t = 2.9182, p = 0.0244) have differences. (D) The distribution channels of the significant difference in the position of topographic map for 4 years old group.

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

4

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

Fig. 3. (A) Sample entropy is more effective for 5 years old group to indicate the difference of ASD and TD; (B) The significant difference of topographic map between ASD and TD for 5 years old group; (C) P4 (t = 3.2902, p = 0.0081), Cz (t = 3.9115, p = 0.0029), F3 (t = 3.5357, p = 0.0054), Fp2 (t = 4.0911, p = 0.0022), F7 (t = 4.1033, p = 0.0021) and Fp1 (t = 4.2488, p = 0.0017) have significant differences; C3 (t = 2.4984, p = 0.0315) and T7 (t = 2.3279, p = 0.0422) have differences. (D) The distribution channels of the significant difference in the position of topographic map for 5 years old group.

modifications of the sample entropy algorithm and proposed a definition for FuzzyEn. The FuzzyEn-based algorithm retains several characteristics of sample entropy, including the relative uniformity and the suitability for the process of short datasets. In addition, by making the similarity measurement formula fuzzy, the FuzzyEnbased algorithm precludes the limitations of the sampEn definition, as FuzzyEn can transit smoothly through varying parameters [31,32]. Parameters selection is similar with the parameters selection in the calculation of sample entropy. The details on the calculation of the FuzzyEn can be found in [32]. 2.4.3. Wavelet entropy (WE) Wavelet entropy is calculated from the signal sequence of wavelet decomposition [33]. The approximation coefficient of signal sequence after wavelet decomposition is cAl , wavelet coefficient of each layer is cDj ; j ¼ 1; 2; :::; l, the sum square of wavelet coefficients consists of wavelet energy Ei ; i ¼ 1; 2; :::; l þ 1 at each scale, so the total energy over all scales is obtained by

Etotal ¼

lþ1 X i¼1

Ei ¼ jcAl j2 þ

l  X  cDj 2 j¼1

Relative wavelet energy at each scale is calculated by

Pi ¼

Ei ; i ¼ 1; 2; :::; l þ 1 Etotal

According to the theory of Shannon entropy [34], WE is defined as

SP ¼ 

lþ1 X

Pi lnPi

i¼1

Based on the definition of Renyi entropy [35], Renyi wavelet entropy (RWE) is defined as

RWE ¼

SaP logNi

2.4.4. Permutation entropy (PE) Permutation entropy was proposed by Bandt in 2002 [36]. It is a measure of the irregularity of signals and based on a comparison of the neighboring order of signal values. It has been shown that PE is unaffected by signal disturbances, and it can be used to analyze time series which is generated by high-dimensional systems with low stationarity. The algorithm is very fast and robust. PeEn can analyze consecutive subvectors of constant length n in the

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

5

Fig. 4. (A) Fuzzy entropy is more effective for 6 years old group to indicate the difference of ASD and TD; (B) The significant difference of topographic map between ASD and TD for 6 years old group; (C) F3 (t = 3.7056, p = 0.0060) have significant differences; Pz (t = 2.4241, p = 0.0416), C4 (t = 2.4607, p = 0.0393), Cz (t = 2.3359, p = 0.0477) and Fz (t = 2.36.4, p = 0.0459) have differences. (D) The distribution channels of the significant difference in the position of topographic map for 6 years old group.

analyzed signal interval. The samples order in every subvector according to their amplitudes is computed and is defined as permutations of order n. The parameter value is given by the entropy of the distribution of the obtained permutations and quantifies the monotone behavior of adjacent signal amplitudes. The symbol sequence of each row in the matrix which is reconstructed by a time series is

SðlÞ ¼ ðj1 ; j2 ; :::; jm Þ where l ¼ 1; 2; :::; k and k 6 m!. The probability of each symbol sequence is P1 ; P2 ; :::; P k . Based on the definition of Renyi entropy, the Renyi permutation Pk a log P j¼1 j . entropy (RPE) can be defined by RPE ¼ ð1aÞlnm! 3. Results 3.1. Entropy in different brain areas with different ages in autistic children First we tried to investigate which specific brain areas have significant difference by different entropy methods at different age groups. Fig. 1 shows the results that each color indicates an entropy and its height represents p value of entropy. For 4 years old group, the significant differences concentrate on central area between ASD and TD children; for 5 years old group, at the frontal

and temporal left areas the entropies are differences between ASD and TD children; for 6 years old group, at the frontal and central areas the entropies are differences between ASD and TD children; for 7 years old group, entropy in the central area has difference and no significant difference is found for 8 years old group.

3.2. Different entropy with different ages in autistic children Then we further reveal which entropy is more effective at each age, and which channels are more significant between ASD and TD children. For each age we calculated four entropies and found the significant differences between ASD and TD for certain entropy and brain regions. After multiple correction by false discovery rate (FDR) for comparing the difference significance p value of four entropies, the optimal entropy algorithm and corresponding channels to distinguish the difference between ASD and TD were obtained. The results were shown in Figs. 2–6. We have conclusions that the differences in brain regions using certain entropy method at each age group can be proposed to distinguish ASD and TD children. In detail, for 4 years old group, there is significant difference in central by RPE method; the significant differences are in frontal and central by SampEn for 5 years old group; the significant difference is in frontal by FuzzyEn for 6 years old group; the significant differences are in central by RWE for 7 years old group and the difference is in occipital by RWE for 8 years old group. We measure the differences of the autistic and

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

6

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

Fig. 5. (A) Renyi wavelet entropy is more effective for 7 years old group to indicate the difference of ASD and TD; (B) The significant difference of topographic map between ASD and TD for 7 years old group; (C) F8 (t = 4.9856, p = 0.0076) have significant differences and C4 (t = 4.5296, p = 0.0106) have differences. (D) The distribution channels of the significant difference in the position of topographic map for 7 years old group.

TD children by two dimensions, thus, the measurement might be more accurate. 4. Discussions In the present study, we investigated the performance of 4 entropy algorithms to compare the EEG complexity of ASD and TD children, including SampEn, FuzzyEn, RWE and RPE at different age groups. Our results show that all entropy values of the ASD are almost lower than TD children in frontal, left temporal, central and right temporal, which shows the reduced complexity of ASD children. The results aligns with the previous research by Bosl et al. [37], which shows a reduced complexity for infants at a high-risk of ASD. The authors considered EEG complexity might reveal functional endophenotypes of ASD at these early stages and suggested it as a potential biomarker for risk of ASD at very early ages. Our present results could be a complementary to the paper. There are some evidences of frontal, left temporal and central dysfunction in ASD children in resting-state [38,39]. In this paper, the results are consistent with the work and also we found the differences of brain regions have some changes with the increase of age, because increased brain signal variability might reflect enhanced brain activity in a certain brain region in ASD [26].

Each entropy index has its advantages and disadvantages in estimating EEG of the children. However, the effective methods might change with age increase and brain development. Agerelated EEG complexity is widely studied for ASD and normal children and the results are not entirely consistent. One research examined EEG complexity for 7–17 years TD subjects at rest and during the performance of verbal and spatial cognitive tasks and the results indicated an overall increase of EEG complexity with age both in a resting state and during the performance of cognitive tasks [40]. Both increased and decreased signal variability occurring both region-specifically and frequency-specifically were found by Ghanbari et al. [13]. Previous research have also proposed to make a distinction between ASD and TD by one entropy method and in this paper we considered that brain activities might have changes with the increase of age, so the more effective entropy methods might change. In general, the higher entropy value is, the more complicated the system is. SampEn, FuzzyEn, RWE and RPE show their good performance in this study. Sample entropy value has related with parameters of embedded dimension and similar tolerance, but does not contain its own data segment, therefore does not depend on the length of the data. If one signal has a higher entropy value than the other, the relationship remained unchanged when calculating with other embedded

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

7

Fig. 6. (A) Renyi wavelet entropy is more effective for 8 years old group to indicate the difference of ASD and TD; (B) The significant difference of topographic map between ASD and TD for 8 years old group; (C) O1 (t = 2.7207, p = 0.0297) have differences. (D) The distribution channels of the significant difference in the position of topographic map for 8 years old group.

dimensions and similar tolerances. Parameter adjustments have little influence for SampEn, which makes it more suitable in biophysiological signals analysis. FuzzyEn is based on SampEn and contains the advantages of SampEn. WE is a method combining wavelet transform and entropy, where wavelet analysis has the advantages of suitable in non-stationary signal analysis. Thus, WE not only has the ability to analyze the signal on the local frequency scale, but also has the information entropy to signal energy distribution statistics, which can reflect the uncertainty and complexity of the signal. PE is a signal mutation detection method which aims at the space characteristics of a time series. Due to its simplicity, PE has a low computational complexity and is robust against noisy data. The results showed SampEn, FuzzyEn, RWE and RPE showed good performance at different age groups to distinguish ASD and TD children. There are some limitations in this study. First, resting-state EEG of ASD children can be influenced by many factors, including EOG and other movements, although we have made a data preprocessing. Second, our sample number is not enough to make a more detailed analysis for the consideration of gender, which is an important factor for brain development.

Acknowledgments This research was founded in part by the National Natural Science Foundation of China (grant No. 61761166003) and Ai You Foundation. Interest disclosure The authors declare no conflict of interest. References [1] Kocsis RN. Diagnostic and statistical manual of mental disorders: fifth edition (DSM-5). Int J Offend Ther Comp Criminol 2013;57(12):1546–8. [2] Zablotsky B et al. Estimated prevalence of autism and other developmental disabilities following questionnaire changes in the 2014 national health interview survey. Natl Health Stat Report 2015;87:1–20. [3] Horwitz B. The elusive concept of brain connectivity. Neuroimage 2003;19 (2):466–70. [4] . [5] O’Reilly C, Lewis JD, Elsabbagh M. Is functional brain connectivity atypical in autism? A systematic review of EEG and MEG studies. PLoS ONE 2017;12(5): e0175870.

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027

8

J. Kang et al. / Journal of Clinical Neuroscience xxx (xxxx) xxx

[6] Han YL et al. Changes of EEG spectra and functional connectivity during an objectlocation memory task in Alzheimer’s disease. Front Behav Neurosci 2017;11. [7] Blinowska KJ et al. Functional and effective brain connectivity for discrimination between Alzheimer’s patients and healthy individuals: a study on resting state EEG rhythms. Clin Neurophysiol 2017;128(4):667–80. [8] Sunwoo JS et al. Altered functional connectivity in idiopathic rapid eye movement sleep behavior disorder: a resting-state EEG study. Sleep 2017;40(6). [9] Lie OV, van Mierlo P. Seizure-onset mapping based on time-variant multivariate functional connectivity analysis of high-dimensional intracranial EEG: a Kalman filter approach. Brain Topogr 2017;30(1):46–59. [10] Wass S. Distortions and disconnections: disrupted brain connectivity in autism. Brain Cogn 2011;75(1):18–28. [11] Just MA et al. Cortical activation and synchronization during sentence comprehension in high-functioning autism: evidence of underconnectivity. Brain 2004;127(Pt 8):1811–21. [12] Li X, Ouyang G, Richards DA. Predictability analysis of absence seizures with permutation entropy. Epilepsy Res 2007;77(1):70–4. [13] Ghanbari Y et al. Joint analysis of band-specific functional connectivity and signal complexity in autism. J Autism Dev Disord 2015;45(2):444–60. [14] McDonough IM, Nashiro K. Network complexity as a measure of information processing across resting-state networks: evidence from the Human Connectome Project. Front Hum Neurosci 2014;8:409. [15] Misic B et al. Functional embedding predicts the variability of neural activity. Front Syst Neurosci 2011;5:90. [16] Li X, Cui S, Voss LJ. Using permutation entropy to measure the electroencephalographic effects of sevoflurane. Anesthesiology 2008;109 (3):448–56. [17] Liang Z et al. EEG entropy measures in anesthesia. Front Comput Neurosci 2015;9:16. [18] Deng B et al. Multivariate multi-scale weighted permutation entropy analysis of EEG complexity for Alzheimer’s disease. Cogn Neurodyn 2017;11 (3):217–31. [19] Tan O et al. EEG complexity and frequency in chronic residual schizophrenia. Anadolu Psikiyatri Dergisi 2016;17(5):385–92. [20] Artan NS. EEG analysis via multiscale Lempel-Ziv complexity for seizure detection. 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (Embc), 2016. p. 4535–538. [21] Piryatinska A, Darkhovsky B, Kaplan A. Binary classification of multichannelEEG records based on the is an element of-complexity of continuous vector functions. Comput Methods Programs Biomed 2017;152:131–9.

[22] Bosl W et al. EEG complexity as a biomarker for autism spectrum disorder risk. BMC Med 2011;9:18. [23] Djemal R et al. EEG-based computer aided diagnosis of autism spectrum disorder using wavelet, entropy, and ANN. Biomed Res Int 2017. [24] Lei M et al. Sample entropy of electroencephalogram for children with autism based on virtual driving game. Acta Phys Sin 2016;65. [25] Han JX et al. Global synchronization of multichannel EEG based on Renyi entropy in children with autism spectrum disorder. Appl Sci Basel 2017;7. [26] Clarke AR et al. Age and sex effects in the EEG: development of the normal child. Clin Neurophysiol 2001;112(5):806–14. [27] Pincus SM. Approximate entropy as a measure of system complexity. Proc Natl Acad Sci USA 1991;88(6):2297–301. [28] Richman JS, Moorman JR. Physiological time-series analysis using approximate entropy and sample entropy. Am J Physiol Heart Circ Physiol 2000;278(6): H2039–49. [29] Richman JS, Lake DE, Moorman JR. Sample entropy. Methods Enzymol 2004;384:172–84. [30] Chen WT et al. Characterization of surface EMG signal based on fuzzy entropy. IEEE Trans Neural Syst Rehabil Eng 2007;15(2):266–72. [31] Kosko B. Fuzzy entropy and conditioning. Inf Sci 1986;40(2):165–74. [32] Cheng HD, Chen YH, Jiang XH. Thresholding using two-dimensional histogram and fuzzy entropy principle. IEEE Trans Image Process 2000;9(4):732–5. [33] Barchiesi D, Gharbi T. Local spectral information in the near field with wavelet analysis and entropy. Appl Opt 1999;38(31):6587–96. [34] Bollini CG, Oxman LE. Shannon entropy and the eigenstates of the single-mode squeeze operator. Phys Rev A 1993;47(3):2339–43. [35] Bashkirov AG. Maximum Renyi entropy principle for systems with power-law Hamiltonians. Phys Rev Lett 2004;93(13):130601. [36] Bandt C. Ordinal time series analysis. Ecol Model 2005;182(3–4):229–38. [37] Bosl W et al. EEG complexity as a biomarker for autism spectrum disorder risk. BMC Med 2011;9. [38] Coben R et al. EEG power and coherence in autistic spectrum disorder. Clin Neurophysiol 2008;119(5):1002–9. [39] Murias M et al. Resting state cortical connectivity reflected in EEG coherence in individuals with autism. Biol Psychiatry 2007;62(3):270–3. [40] Anokhin AP et al. Complexity of electrocortical dynamics in children: developmental aspects. Dev Psychobiol 2000;36(1):9–22.

Please cite this article as: J. Kang, H. Chen, X. Li et al., EEG entropy analysis in autistic children, Journal of Clinical Neuroscience, https://doi.org/10.1016/j. jocn.2018.11.027