基于功能磁共振影像的自闭症分类方法研究

自闭症谱系障碍（ASD）由一系列神经发育障碍组成，由于其较强的异质性，目前所开发出的很多 ASD 辅助筛查模型在由多个采样中心构成大型数据集上还无法达到临床应用中可接受的精度和鲁棒性。本文在现有研究的基础上，从特征选择方法和深度学习模型设计两个方面做出改进，形成从数据预处理到分类模型构建的整套 ASD 分类框架，力求在大型异质数据集上进一步提升 ASD 的预测效果。

本文的主要研究工作包含以下三部分：
研究工作一：基于最大化相关性最小化冗余性（mRMR）和特征批量递归消除的特征降维。我们设计了一种两阶段特征选择方法对高维功能磁共振成像（fMRI）数据进行降维，第一阶段基于 mRMR 进行特征初步筛选，第二阶段采用逻辑斯蒂回归特征批量递归消除（LRbRFE）方法进行二次特征选择。该方法克服了本领域以往研究中特征选择方法计算开销大以及缺乏去冗余能力等局限性，且效果优于以往该领域研究中常用的过滤特征选择方法。
研究工作二：基于双自编码器（dAE）模型和二分支多层感知机（dMLP）的自闭症分类方法。提出了一种 dAE 预训练 + dMLP 微调的深度学习架构，解决了传统自编码器（AE）无法充分利用已有类别信息的局限性。在基于 fMRI 的 ASD分类任务中，该方法的性能优于在以往研究中被广泛应用的 AE 预训练 + 多层感知机（MLP）微调的架构。
研究工作三：基于遗传算法的自闭症分类模型拓扑结构优化。提出了一种有效的神经网络拓扑结构优化方法—Evolutionary connection sparsification（ECS）算法，该算法通过 0-1 矩阵与模型的权重矩阵做 Hardmard 积实现神经网络连接的稀疏化；在遗传算法的基础上，以 0-1 矩阵的集合作为个体，通过矩阵行向量互换和行内元素重组的方式生成新的拓扑结构，并采用逐代缩减种群规模的方法提升算法的收敛速度。实验结果显示，ECS 算法能够有效提升 MLP 和 dMLP 的分类性能，并且降低了模型的泛化误差。

上述三部分内容构成了一整套 ASD 分类框架，所训练出的模型达到 80.26%的平均精度，优于相同数据集上的其他先进方法。

[1] KANG J, HAN X, SONG J, et al. The identification of children with autism spectrum disorderby SVM approach on EEG and eye-tracking data[J]. Computers in biology and medicine, 2020,120: 103722.
[2] KHODATARS M, SHOEIBI A, SADEGHI D, et al. Deep learning for neuroimaging-baseddiagnosis and rehabilitation of autism spectrum disorder: a review[J]. Computers in Biologyand Medicine, 2021, 139: 104949.
[3] HIREMATH C S, SAGAR K J V, YAMINI B, et al. Emerging behavioral and neuroimagingbiomarkers for early and accurate characterization of autism spectrum disorders: a systematicreview[J]. Translational Psychiatry, 2021, 11(1): 42.
[4] BRADSHAWJ,STEINERAM,GENGOUXG,etal. Feasibilityandeffectivenessofveryearlyintervention for infants at-risk for autism spectrum disorder: A systematic review[J]. Journalof autism and developmental disorders, 2015, 45: 778-794.
[5] HA S, SOHN I J, KIM N, et al. Characteristics of brains in autism spectrum disorder: structure,function and connectivity across the lifespan[J]. Experimental neurobiology, 2015, 24(4): 273.
[6] GESCHWIND D H, LEVITT P. Autism spectrum disorders: developmental disconnectionsyndromes[J]. Current opinion in neurobiology, 2007, 17(1): 103-111.
[7] CHLEBOWSKI C, ROBINS D L, BARTON M L, et al. Large-scale use of the modified check-list for autism in low-risk toddlers[J]. Pediatrics, 2013, 131(4): e1121-e1127.
[8] SCHOPLER E, REICHLER R J, RENNER B R. The childhood autism rating scale (CARS)[M]. WPS Los Angeles, 2010.
[9] LORD C, RUTTER M, GOODE S, et al. Austism diagnostic observation schedule: A standard-izedobservationofcommunicativeandsocialbehavior[J]. Journalofautismanddevelopmentaldisorders, 1989, 19(2): 185-212.
[10] MULLEN E M, et al. Mullen scales of early learning[M]. AGS Circle Pines, MN, 1995.
[11] RUTTER M, BAILEY A, LORD C. The social communication questionnaire: Manual[M].Western Psychological Services, 2003.
[12] GESCHWIND D H. Advances in autism[J]. Annual review of medicine, 2009, 60: 367-380.
[13] OGAWA S, LEE T M, KAY A R, et al. Brain magnetic resonance imaging with contrast de-pendent on blood oxygenation.[J]. proceedings of the National Academy of Sciences, 1990, 87(24): 9868-9872.
[14] HRDLICKA M, SANDA J, URBANEK T, et al. Diffusion tensor imaging and tractography inautistic, dysphasic, and healthy control children[J]. Neuropsychiatric Disease and Treatment,2019: 2843-2852.
[15] MORIDIANP,GHASSEMIN,JAFARIM,etal. Automaticautismspectrumdisorderdetectionusing artificial intelligence methods with MRI neuroimaging: A review[A]. 2022.
[16] CHEN C P, KEOWN C L, JAHEDI A, et al. Diagnostic classification of intrinsic functionalconnectivityhighlightssomatosensory, defaultmode, andvisualregionsinautism[J]. NeuroIm-age: Clinical, 2015, 8: 238-245.
[17] CHEN H, DUAN X, LIU F, et al. Multivariate classification of autism spectrum disorder usingfrequency-specific resting-state functional connectivity—a multi-center study[J]. Progress inNeuro-Psychopharmacology and Biological Psychiatry, 2016, 64: 1-9.
[18] PLITT M, BARNES K A, MARTIN A. Functional connectivity classification of autism iden-tifies highly predictive brain features but falls short of biomarker standards[J]. NeuroImage:Clinical, 2015, 7: 359-366.
[19] HEINSFELD A S, FRANCO A R, CRADDOCK R C, et al. Identification of autism spectrumdisorderusingdeeplearningandtheABIDEdataset[J]. NeuroImage: Clinical, 2018, 17: 16-23.
[20] HUANG Z A, ZHU Z, YAU C H, et al. Identifying autism spectrum disorder from resting-state fMRI using deep belief network[J]. IEEE Transactions on neural networks and learningsystems, 2020, 32(7): 2847-2861.
[21] ZHANG F, WEI Y, LIU J, et al. Identification of Autism spectrum disorder based on a novelfeature selection method and Variational Autoencoder[J]. Computers in Biology and Medicine,2022, 148: 105854.
[22] MOHAN P, PARAMASIVAM I. Feature reduction using SVM-RFE technique to detect autismspectrum disorder[J]. Evolutionary Intelligence, 2021, 14: 989-997.
[23] KONG Y, GAO J, XU Y, et al. Classification of autism spectrum disorder by combining brainconnectivity and deep neural network classifier[J]. Neurocomputing, 2019, 324: 63-68.
[24] GLOVER G H. Overview of functional magnetic resonance imaging[J]. Neurosurgery Clinics,2011, 22(2): 133-139.
[25] DICHTER G S. Functional magnetic resonance imaging of autism spectrum disorders[J]. Dia-logues in clinical neuroscience, 2022.
[26] BISWAL B, ZERRIN YETKIN F, HAUGHTON V M, et al. Functional connectivity in the mo-tor cortex of resting human brain using echo-planar MRI[J]. Magnetic resonance in medicine,1995, 34(4): 537-541.
[27] WARREN S L, MOUSTAFA A A. Functional magnetic resonance imaging, deep learning, andAlzheimer’s disease: A systematic review[J]. Journal of Neuroimaging, 2023, 33(1): 5-18.
[28] POLDRACK R A. Region of interest analysis for fMRI[J]. Social cognitive and affectiveneuroscience, 2007, 2(1): 67-70.
[29] KRIEGESKORTE N, BODURKA J, BANDETTINI P. Artifactual time-course correlations inecho-planar fMRI with implications for studies of brain function[J]. International Journal ofImaging Systems and Technology, 2008, 18(5-6): 345-349.
[30] PAIS-ROLDÁN P, MATEO C, PAN W J, et al. Contribution of animal models toward under-standing resting state functional connectivity[J]. NeuroImage, 2021, 245: 118630.
[31] DOYLE-THOMAS K A, LEE W, FOSTER N E, et al. Atypical functional brain connectivityduring rest in autism spectrum disorders[J]. Annals of neurology, 2015, 77(5): 866-876.
[32] EILAM-STOCK T, XU P, CAO M, et al. Abnormal autonomic and associated brain activitiesduring rest in autism spectrum disorder[J]. Brain, 2014, 137(1): 153-171.
[33] TRAYNOR J, DOYLE-THOMAS K, HANFORD L, et al. Indices of repetitive behaviour arecorrelated with patterns of intrinsic functional connectivity in youth with autism spectrum dis-order[J]. Brain research, 2018, 1685: 79-90.
[34] KUMAR V, GARG R. Resting state functional connectivity alterations in individuals withautism spectrum disorders: a systematic review[J]. medRxiv, 2021: 2021-07.
[35] SHOU X J, XU X J, ZENG X Z, et al. A volumetric and functional connectivity MRI study ofbrain arginine-vasopressin pathways in autistic children[J]. Neuroscience Bulletin, 2017, 33:130-142.
[36] SHEN M D, LI D D, KEOWN C L, et al. Functional connectivity of the amygdala is disruptedinpreschool-agedchildrenwithautismspectrumdisorder[J]. JournaloftheAmericanAcademyof Child & Adolescent Psychiatry, 2016, 55(9): 817-824.
[37] GUO X, DUAN X, LONG Z, et al. Decreased amygdala functional connectivity in adolescentswith autism: A resting-state fMRI study[J]. Psychiatry Research: Neuroimaging, 2016, 257:47-56.
[38] GREEN S A, HERNANDEZ L, BOOKHEIMER S Y, et al. Salience network connectivity inautism is related to brain and behavioral markers of sensory overresponsivity[J]. Journal of theAmerican Academy of Child & Adolescent Psychiatry, 2016, 55(7): 618-626.
[39] TOMASI D, VOLKOW N D. Reduced local and increased long-range functional connectivityof the thalamus in autism spectrum disorder[J]. Cerebral Cortex, 2019, 29(2): 573-585.
[40] ALAERTS K, NAYAR K, KELLY C, et al. Age-related changes in intrinsic function of thesuperior temporal sulcus in autism spectrum disorders[J]. Social cognitive and affective neuro-science, 2015, 10(10): 1413-1423.
[41] CHIENHY,LINHY,LAIMC,etal. Hyperconnectivityoftherightposteriortemporo-parietaljunction predicts social difficulties in boys with autism spectrum disorder[J]. Autism Research,2015, 8(4): 427-441.
[42] HOFFMANN F, KOEHNE S, STEINBEIS N, et al. Preserved self-other distinction duringempathy in autism is linked to network integrity of right supramarginal gyrus[J]. Journal ofautism and developmental disorders, 2016, 46: 637-648.
[43] BELLMAN R. Dynamic programming, princeton univ[J]. Press Princeton, New Jersey, 1957.
[44] TRUNKGV. Aproblemofdimensionality: Asimpleexample[J]. IEEETransactionsonpatternanalysis and machine intelligence, 1979(3): 306-307.
[45] MCLACHLAN G J. Discriminant analysis and statistical pattern recognition[M]. John Wiley& Sons, 2005.
[46] WANG Y, WANG J, PAL N R. Supervised Feature Selection via Collaborative NeurodynamicOptimization[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022.
[47] VENKATESHB,ANURADHAJ. Areviewoffeatureselectionanditsmethods[J]. Cyberneticsand information technologies, 2019, 19(1): 3-26.
[48] WANG Y, LI X, WANG J. A neurodynamic optimization approach to supervised feature selec-tion via fractional programming[J]. Neural Networks, 2021, 136: 194-206.
[49] JEE G, CHOUHAN S, GOURISARIA M K, et al. Detection of Autism Spectrum Disor-der Through Orthogonal Decomposition and Pearson Correlation for Feature Selection[C]//2021 4th International Conference on Recent Trends in Computer Science and Technology(ICRTCST). IEEE, 2022: 103-109.
[50] RAKIĆ M, CABEZAS M, KUSHIBAR K, et al. Improving the detection of autism spectrumdisorder by combining structural and functional MRI information[J]. NeuroImage: Clinical,2020, 25: 102181.
[51] SALEH A I, RABIE A H. A new Autism Spectrum Disorder Discovery (ASDD) strategy usingdata mining techniques based on blood tests[J]. Biomedical Signal Processing and Control,2023, 81: 104419.
[52] DUDAM,MAR,HABERN,etal. Useofmachinelearningforbehavioraldistinctionofautismand ADHD[J]. Translational psychiatry, 2016, 6(2): e732-e732.
[53] WASHINGTON P, PASKOV K M, KALANTARIAN H, et al. Feature selection and dimen-sion reduction of social autism data[C]//Pacific Symposium on Biocomputing 2020. WorldScientific, 2019: 707-718.
[54] WIRATSIN I O, NARUPIYAKUL L. Feature selection technique for autism spectrum disor-der[C]//Proceedings of the 5th International Conference on Control Engineering and ArtificialIntelligence. 2021: 53-56.
[55] THABTAHF,ABDELHAMIDN,PEEBLESD. Amachinelearningautismclassificationbasedon logistic regression analysis[J]. Health information science and systems, 2019, 7: 1-11.
[56] JUNGM,TUY,PARKJ,etal. Surface-basedsharedanddistinctrestingfunctionalconnectivityin attention-deficit hyperactivity disorder and autism spectrum disorder[J]. The British Journalof Psychiatry, 2019, 214(6): 339-344.
[57] HUANG H, LIU X, JIN Y, et al. Enhancing the representation of functional connectivity net-works by fusing multi-view information for autism spectrum disorder diagnosis[J]. Humanbrain mapping, 2019, 40(3): 833-854.
[58] KOHAVI R, JOHN G H. Wrappers for feature subset selection[J]. Artificial intelligence, 1997,97(1-2): 273-324.
[59] WANG C, XIAO Z, WANG B, et al. Identification of autism based on SVM-RFE and stackedsparse auto-encoder[J]. Ieee Access, 2019, 7: 118030-118036.
[60] WANG Y, LIU J, XIANG Y, et al. MAGE: automatic diagnosis of autism spectrum disordersusing multi-atlas graph convolutional networks and ensemble learning[J]. Neurocomputing,2022, 469: 346-353.
[61] RETICO A, GIULIANO A, TANCREDI R, et al. The effect of gender on the neuroanatomyof children with autism spectrum disorders: a support vector machine case-control study[J].Molecular autism, 2016, 7: 1-20.
[62] GUYON I, WESTON J, BARNHILL S, et al. Gene selection for cancer classification usingsupport vector machines[J]. Machine learning, 2002, 46: 389-422.
[63] HAWEEL R, DEKHIL O, SHALABY A, et al. A novel framework for grading autism severityusing task-based fmri[C]//2020 IEEE 17th International Symposium on Biomedical Imaging(ISBI). IEEE, 2020: 1404-1407.
[64] FREDOA,JAHEDIA,REITERM,etal. Diagnosticclassificationofautismusingresting-statefMRI data and conditional random forest[J]. Age, 2018, 12(2.76): 6-41.
[65] ALI M T, ELNAKIEB Y A, SHALABY A, et al. Autism classification using smri: A recursivefeatures selection based on sampling from multi-level high dimensional spaces[C]//2021 IEEE18th International Symposium on Biomedical Imaging (ISBI). IEEE, 2021: 267-270.
[66] YIN W, LI L, WU F X. A semi-supervised autoencoder for autism disease diagnosis[J]. Neu-rocomputing, 2022, 483: 140-147.
[67] ALZUBI R, RAMZAN N, ALZOUBI H. Hybrid feature selection method for autism spectrumdisorderSNPs[C]//2017IEEEConferenceonComputationalIntelligenceinBioinformaticsandComputational Biology (CIBCB). IEEE, 2017: 1-7.
[68] FLEURET F. Fast binary feature selection with conditional mutual information.[J]. Journal ofMachine learning research, 2004, 5(9).
[69] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. nature, 2015, 521(7553): 436-444.
[70] DOLZ J, DESROSIERS C, WANG L, et al. Deep CNN ensembles and suggestive annotationsfor infant brain MRI segmentation[J]. Computerized Medical Imaging and Graphics, 2020, 79:101660.
[71] LI X, DVORNEK N C, PAPADEMETRIS X, et al. 2-channel convolutional 3D deep neuralnetwork (2CC3D) for fMRI analysis: ASD classification and feature learning[C]//2018 IEEE15th International Symposium on Biomedical Imaging (ISBI 2018). IEEE, 2018: 1252-1255.
[72] DELANNOY Q, PHAM C H, CAZORLA C, et al. SegSRGAN: Super-resolution and segmen-tationusinggenerativeadversarialnetworks—ApplicationtoneonatalbrainMRI[J]. Computersin Biology and Medicine, 2020, 120: 103755.
[73] RUMELHART D E, HINTONG E, WILLIAMS R J. Learning internal representations by errorpropagation[R]. California Univ San Diego La Jolla Inst for Cognitive Science, 1985.
[74] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neuralnetworks[J]. science, 2006, 313(5786): 504-507.
[75] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J].Neural computation, 2006, 18(7): 1527-1554.
[76] ERHAN D, COURVILLE A, BENGIO Y, et al. Why does unsupervised pre-training help deeplearning?[C]//Proceedings of the thirteenth international conference on artificial intelligenceand statistics. JMLR Workshop and Conference Proceedings, 2010: 201-208.
[77] NG A, et al. Sparse autoencoder[J]. CS294A Lecture notes, 2011, 72(2011): 1-19.
[78] VINCENT P, LAROCHELLE H, BENGIO Y, et al. Extracting and composing robust featureswith denoising autoencoders[C]//Proceedings of the 25th international conference on Machinelearning. 2008: 1096-1103.
[79] KINGMA D P, WELLING M. Auto-encoding variational bayes[A]. 2013.
[80] DOERSCH C. Tutorial on variational autoencoders[A]. 2016.
[81] WANG Y, WANG J, WU F X, et al. AIMAFE: Autism spectrum disorder identification withmulti-atlasdeepfeature representationandensemble learning[J]. Journal ofneurosciencemeth-ods, 2020, 343: 108840.
[82] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recog-nition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324.
[83] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet classification with deep convo-lutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90.
[84] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale imagerecognition[A]. 2014.
[85] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedingsof the IEEE conference on computer vision and pattern recognition. 2016: 770-778.
[86] BENGS M, GESSERT N, SCHLAEFER A. 4d spatio-temporal deep learning with 4d fmri datafor autism spectrum disorder classification[A]. 2020.
[87] EL-GAZZAR A, QUAAK M, CERLIANI L, et al. A hybrid 3DCNN and 3DC-LSTM basedmodel for 4D spatio-temporal fMRI data: an ABIDE autism classification study[C]//OR 2.0Context-Aware Operating Theaters and Machine Learning in Clinical Neuroimaging: SecondInternational Workshop, OR 2.0 2019, and Second International Workshop, MLCN 2019, Heldin Conjunction with MICCAI 2019, Shenzhen, China, October 13 and 17, 2019, Proceedings2. Springer, 2019: 95-102.
[88] SHERKATGHANAD Z, AKHONDZADEH M, SALARI S, et al. Automated detection ofautism spectrum disorder using a convolutional neural network[J]. Frontiers in neuroscience,2020, 13: 1325.
[89] SHRIVASTAVA S, MISHRA U, SINGH N, et al. Control or autism-classification using convo-lutional neural networks on functional MRI[C]//2020 11th International Conference on Com-puting, Communication and Networking Technologies (ICCCNT). IEEE, 2020: 1-6.
[90] DVORNEK N C, VENTOLA P, PELPHREY K A, et al. Identifying autism from resting-statefMRI using long short-term memory networks[C]//Machine Learning in Medical Imaging: 8thInternational Workshop, MLMI 2017, Held in Conjunction with MICCAI 2017, Quebec City,QC, Canada, September 10, 2017, Proceedings 8. Springer, 2017: 362-370.
[91] BYEON K, KWON J, HONG J, et al. Artificial neural network inspired by neuroimagingconnectivity: application in autism spectrum disorder[C]//2020 IEEE International Conferenceon Big Data and Smart Computing (BigComp). IEEE, 2020: 575-578.
[92] ZHOU J, CUI G, HU S, et al. Graph neural networks: A review of methods and applications[J]. AI open, 2020, 1: 57-81.
[93] ASIF N A, SARKER Y, CHAKRABORTTY R K, et al. Graph neural network: A comprehen-sive review on non-euclidean space[J]. IEEE Access, 2021, 9: 60588-60606.
[94] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[A]. 2016.
[95] VELIČKOVIĆ P, CUCURULL G, CASANOVA A, et al. Graph attention networks[A]. 2017.
[96] HAMILTON W, YING Z, LESKOVEC J. Inductive representation learning on large graphs[J].Advances in neural information processing systems, 2017, 30.
[97] YANG C, WANG P, TAN J, et al. Autism spectrum disorder diagnosis using graph attentionnetworkbasedonspatial-constrainedsparsefunctionalbrainnetworks[J]. ComputersinBiologyand Medicine, 2021, 139: 104963.
[98] NOMAN F, YAP S Y, PHAN R C W, et al. Graph Autoencoder-Based Embedded Learning inDynamic Brain Networks for Autism Spectrum Disorder Identification[C]//2022 IEEE Interna-tional Conference on Image Processing (ICIP). IEEE, 2022: 2891-2895.
[99] YAO D, YANG E, SUN L, et al. Integrating Multimodal MRIs for Adult ADHD Identifica-tion with Heterogeneous Graph Attention Convolutional Network[C]//Predictive Intelligencein Medicine: 4th International Workshop, PRIME 2021, Held in Conjunction with MICCAI2021, Strasbourg, France, October 1, 2021, Proceedings 4. Springer, 2021: 157-167.
[100] SONG X, FRANGI A, XIAO X, et al. Integrating similarity awareness and adaptive calibrationin graph convolution network to predict disease[C]//Medical Image Computing and ComputerAssistedIntervention–MICCAI2020: 23rdInternationalConference, Lima, Peru, October4–8,2020, Proceedings, Part VII 23. Springer, 2020: 124-133.
[101] SABOURS,FROSSTN,HINTONGE. DynamicRoutingBetweenCapsules[C/OL]//GUYONI, LUXBURG U V, BENGIO S, et al. Advances in Neural Information Processing Systems:volume 30. Curran Associates, Inc., 2017. https://proceedings.neurips.cc/paper/2017/file/2cad8fa47bbef282badbb8de5374b894-Paper.pdf.
[102] JIAO Z, LI H, FAN Y. Improving diagnosis of autism spectrum disorder and disentanglingits heterogeneous functional connectivity patterns using capsule networks[C]//2020 IEEE 17thInternational Symposium on Biomedical Imaging (ISBI). IEEE, 2020: 1331-1334.
[103] ELSKEN T, METZEN J H, HUTTER F. Neural architecture search: A survey[J]. The Journalof Machine Learning Research, 2019, 20(1): 1997-2017.
[104] ZOPH B, VASUDEVAN V, SHLENS J, et al. Learning transferable architectures for scal-able image recognition[C]//Proceedings of the IEEE conference on computer vision and patternrecognition. 2018: 8697-8710.
[105] REAL E, AGGARWAL A, HUANG Y, et al. Aging evolution for image classifier architecturesearch[C]//AAAI conference on artificial intelligence: volume 2. 2019: 2.
[106] CHEN L C, COLLINS M, ZHU Y, et al. Searching for efficient multi-scale architectures fordense image prediction[J]. Advances in neural information processing systems, 2018, 31.
[107] REAL E, MOORE S, SELLE A, et al. Large-scale evolution of image classifiers[C]//International Conference on Machine Learning. PMLR, 2017: 2902-2911.
[108] LIU Y, SUN Y, XUE B, et al. A survey on evolutionary neural architecture search[J]. IEEEtransactions on neural networks and learning systems, 2021.
[109] XIE L, YUILLE A. Genetic cnn[C]//Proceedings of the IEEE international conference on com-puter vision. 2017: 1379-1388.
[110] HOLLAND J H. Genetic algorithms[J]. Scientific american, 1992, 267(1): 66-73.
[111] SUN Y, XUE B, ZHANG M, et al. Evolving deep convolutional neural networks for imageclassification[J]. IEEE Transactions on Evolutionary Computation, 2019, 24(2): 394-407.
[112] SUN Y, XUE B, ZHANG M, et al. Completely automated CNN architecture design based onblocks[J]. IEEEtransactionsonneuralnetworksandlearningsystems, 2019, 31(4): 1242-1254.
[113] CAMERO A, TOUTOUH J, STOLFI D H, et al. Evolutionary deep learning for car parkoccupancy prediction in smart cities[C]//Learning and Intelligent Optimization: 12th Interna-tional Conference, LION 12, Kalamata, Greece, June 10–15, 2018, Revised Selected Papers 12.Springer, 2019: 386-401.
[114] GAOZ,LIY,YANGY,etal. AGPSO-optimizedconvolutionalneuralnetworksforEEG-basedemotion recognition[J]. Neurocomputing, 2020, 380: 225-235.
[115] ELSAID A, JAMIY F E, HIGGINS J, et al. Using ant colony optimization to optimize longshort-term memory recurrent neural networks[C]//Proceedings of the Genetic and EvolutionaryComputation Conference. 2018: 13-20.
[116] DE FALCO I, DE PIETRO G, DELLA CIOPPA A, et al. Evolution-based configuration opti-mization of a deep neural network for the classification of obstructive sleep apnea episodes[J].Future Generation Computer Systems, 2019, 98: 377-391.
[117] BYLA E, PANG W. Deepswarm: Optimising convolutional neural networks using swarm in-telligence[C]//AdvancesinComputationalIntelligenceSystems: ContributionsPresentedatthe19th UK Workshop on Computational Intelligence, September 4-6, 2019, Portsmouth, UK 19.Springer, 2020: 119-130.
[118] WANG B, SUN Y, XUE B, et al. A hybrid ga-pso method for evolving architecture and shortconnections of deep convolutional neural networks[C]//PRICAI 2019: Trends in Artificial In-telligence: 16th Pacific Rim International Conference on Artificial Intelligence, Cuvu, YanucaIsland, Fiji, August 26-30, 2019, Proceedings, Part III 16. Springer, 2019: 650-663.
[119] KWASIGROCH A, GROCHOWSKI M, MIKOLAJCZYK M. Deep neural network architec-ture search using network morphism[C]//2019 24th International Conference on Methods andModels in Automation and Robotics (MMAR). IEEE, 2019: 30-35.
[120] SUN Y, XUE B, ZHANG M, et al. A particle swarm optimization-based flexible convolutionalautoencoder for image classification[J]. IEEE transactions on neural networks and learningsystems, 2018, 30(8): 2295-2309.
[121] LIU P, EL BASHA M D, LI Y, et al. Deep evolutionary networks with expedited geneticalgorithms for medical image denoising[J]. Medical image analysis, 2019, 54: 306-315.
[122] ASSUNÇÃO F, LOURENÇO N, MACHADO P, et al. DENSER: deep evolutionary networkstructured representation[J]. Genetic Programming and Evolvable Machines, 2019, 20: 5-35.
[123] YANG Z, WANG Y, CHEN X, et al. Cars: Continuous evolution for efficient neural archi-tecture search[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and PatternRecognition. 2020: 1829-1838.
[124] DI MARTINO A, YAN C G, LI Q, et al. The autism brain imaging data exchange: towardsa large-scale evaluation of the intrinsic brain architecture in autism[J]. Molecular psychiatry,2014, 19(6): 659-667.
[125] TZOURIO-MAZOYER N, LANDEAU B, PAPATHANASSIOU D, et al. Automated anatomi-cal labeling of activationsin SPMusing a macroscopic anatomical parcellation ofthe MNIMRIsingle-subject brain[J]. Neuroimage, 2002, 15(1): 273-289.
[126] HILLMAN E M. Coupling mechanism and significance of the BOLD signal: a status report[J].Annual review of neuroscience, 2014, 37: 161-181.
[127] DOSENBACH N U, NARDOS B, COHEN A L, et al. Prediction of individual brain maturityusing fMRI[J]. Science, 2010, 329(5997): 1358-1361.
[128] EICKHOFF S B, STEPHAN K E, MOHLBERG H, et al. A new SPM toolbox for combiningprobabilistic cytoarchitectonic maps and functional imaging data[J]. Neuroimage, 2005, 25(4):1325-1335.
[129] KENNEDY D N, LANGE N, MAKRIS N, et al. Gyri of the human neocortex: an MRI-basedanalysis of volume and variance.[J]. Cerebral Cortex (New York, NY: 1991), 1998, 8(4): 372-384.
[130] MAKRIS N, MEYER J W, BATES J F, et al. MRI-based topographic parcellation of humancerebral white matter and nuclei: II. Rationale and applications with systematics of cerebralconnectivity[J]. Neuroimage, 1999, 9(1): 18-45.
[131] LANCASTER J L, WOLDORFF M G, PARSONS L M, et al. Automated Talairach atlas labelsfor functional brain mapping[J]. Human brain mapping, 2000, 10(3): 120-131.
[132] DING C, PENG H. Minimum redundancy feature selection from microarray gene expressiondata[J]. Journal of bioinformatics and computational biology, 2005, 3(02): 185-205.
[133] PENG H, LONG F, DING C. Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy[J]. IEEE Transactions on pattern analysisand machine intelligence, 2005, 27(8): 1226-1238.
[134] CHEN Y W, LIN C J. Combining SVMs with various feature selection strategies[J]. Featureextraction: foundations and applications, 2006: 315-324.
[135] YANG X, SCHRADER P T, ZHANG N. A deep neural network study of the ABIDE repositoryon autism spectrum classification[J]. International Journal of Advanced Computer Science andApplications, 2020, 11(4).
[136] SEWANI H, KASHEF R. An autoencoder-based deep learning classifier for efficient diagnosisof autism[J]. Children, 2020, 7(10): 182.
[137] XIAO X, YAN M, BASODI S, et al. Efficient hyperparameter optimization in deep learningusing a variable length genetic algorithm[A]. 2020.
[138] DA SILVA G L F, VALENTE T L A, SILVA A C, et al. Convolutional neural network-basedPSOforlungnodulefalsepositivereductiononCTimages[J]. Computermethodsandprogramsin biomedicine, 2018, 162: 109-118.
[139] DUFOURQ E, BASSETT B A. Eden: Evolutionary deep networks for efficient machine learn-ing[C]//2017 Pattern Recognition Association of South Africa and Robotics and Mechatronics(PRASA-RobMech). IEEE, 2017: 110-115.
[140] JHA R R, BHARDWAJ A, GARG D, et al. MHATC: Autism Spectrum Disorder identifica-tionutilizingmulti-headattentionencoderalongwithtemporalconsolidationmodules[C]//202244th Annual International Conference of the IEEE Engineering in Medicine & Biology Society(EMBC). IEEE, 2022: 337-341.
[141] ESLAMI T, MIRJALILI V, FONG A, et al. ASD-DiagNet: a hybrid learning approach fordetection of autism spectrum disorder using fMRI data[J]. Frontiers in neuroinformatics, 2019,13: 70.
[142] LU H, LIU S, WEI H, et al. Multi-kernel fuzzy clustering based on auto-encoder for fMRIfunctional network[J]. Expert Systems with Applications, 2020, 159: 113513