Share：
Share this content in WeChat
X
[Chinese] [PDF] 457 2
• Clinical Article •
Research on autism brain function network and gradient feature classification based on ensemble learning
ZHOU Yuan  ZHU Yuanqiang  DU Xiangyi  LI Leilei  WANG Chen  ZHENG Jianmin 

Cite this article as: ZHOU Y, ZHU Y Q, DU X Y, et al. Research on autism brain function network and gradient feature classification based on ensemble learning[J]. Chin J Magn Reson Imaging, 2025, 16(7): 6-14. DOI:10.12015/issn.1674-8034.2025.07.002.


[Abstract] Objective To explore the classification performance of resting-state brain functional networks and gradient features in patients with autism spectrum disorders (ASD) based on multimodal machine learning and ensemble learning classification models.Materials and Methods Based on 246 ASD patients and 251 healthy controls (HC), this study used two independent samples t-test to analyse the differences between the results of independent component analysis, gradient analysis, and static functional network connectivity (sFNC) and dynamic functional gradient (dFNG) features to construct a multimodal machine learning classification model. sFNC and dFNG features were used to construct a multimodal machine learning classification model.Results (1) The static connectivity strength of the default network (TN-DM), visual temporal lobe (VI-OT), and visual occipital lobe (VI-OC) networks in ASD patients was significantly weakened, while the connectivity strength between higher cognition frontal lobe (HC-FR) and significant network (TN-SA) was significantly enhanced (P < 0.05, false discovery rate, FDR correction); (2) Dynamic gradient clustering analysis showed that ASD patients remained in the frontal lobe sensorimotor dominant state for a long time in low dimensional space (P < 0.05); (3) The multimodal machine learning model results of sFNC and dFNG show that dFNG and sFNC have significant collaborative classification contributions, significantly improving classification accuracy (accuracy = 99.3%).Conclusions Patients with ASD have systemic abnormalities at both the sFNC and dFNG levels. A multimodal integrated learning model based on the features of sFNC and dFNG can efficiently classify ASD.
[Keywords] autism；magnetic resonance imaging；functional magnetic resonance imaging；gradient feature；ensemble learning

ZHOU Yuan   ZHU Yuanqiang   DU Xiangyi   LI Leilei   WANG Chen   ZHENG Jianmin*  

Department of Diagnostic Radiology, the First Affiliated Hospital of Air Force Military Medical University, Xi'an 710032, China

Corresponding author: ZHENG J M, E-mail: jmzheng1986@126.com

Conflicts of interest   None.

Received  2025-04-19
Accepted  2025-07-06
DOI: 10.12015/issn.1674-8034.2025.07.002
Cite this article as: ZHOU Y, ZHU Y Q, DU X Y, et al. Research on autism brain function network and gradient feature classification based on ensemble learning[J]. Chin J Magn Reson Imaging, 2025, 16(7): 6-14. DOI:10.12015/issn.1674-8034.2025.07.002.

[1]
ZHOU H, XU X, YAN W L, et al. Prevalence of autism spectrum disorder in China: a nationwide multi-center population-based study among children aged 6 to 12 years[J]. Neurosci Bull, 2020, 36(9): 961-971. DOI: 10.1007/s12264-020-00530-6.
[2]
LIU Y, ZHOU F, QIN J X, et al. The intervention dilemma and high burden of children with autism in Guizhou province, Southwest China[J/Ol]. Front Psychiatry, 2022, 13: 929833 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/36405893/. DOI: 10.3389/fpsyt.2022.929833.
[3]
SANTANA C P, DE CARVALHO E A, RODRIGUES I D, et al. Rs-fMRI and machine learning for ASD diagnosis: a systematic review and meta-analysis[J/OL]. Sci Rep, 2022, 12(1): 6030 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/35411059/. DOI: 10.1038/s41598-022-09821-6.
[4]
THABTAH F, PEEBLES D. A new machine learning model based on induction of rules for autism detection[J]. Health Informatics J, 2020, 26(1): 264-286. DOI: 10.1177/1460458218824711.
[5]
YANG J C, YIN X L, HUANG L, et al. Brain functional connectome gradients analysis study of chronic back pain[J]. Rehabil Med, 2024, 34(5): 457-464. DOI: 10.3724/SP.J.1329.2024.05006.
[6]
SONG C, JIANG Z Q, HU L F, et al. A machine learning-based diagnostic model for children with autism spectrum disorders complicated with intellectual disability[J/OL]. Front Psychiatry, 2022, 13: 993077 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/36213933/. DOI: 10.3389/fpsyt.2022.993077.
[7]
AMEMIYA S, TAKAO H, ABE O. Resting-state fMRI: emerging concepts for future clinical application[J]. J Magn Reson Imaging, 2024, 59(4): 1135-1148. DOI: 10.1002/jmri.28894.
[8]
JIANG X, SHOU X J, ZHAO Z B, et al. A brain structural connectivity biomarker for autism spectrum disorder diagnosis in early childhood[J/OL]. Psychoradiology, 2023, 3: kkad005 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/38666122/. DOI: 10.1093/psyrad/kkad005.
[9]
RABANY L, BROCKE S, CALHOUN V D, et al. Dynamic functional connectivity in schizophrenia and autism spectrum disorder: Convergence, divergence and classification[J/OL]. Neuroimage Clin, 2019, 24: 101966 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/31401405/. DOI: 10.1016/j.nicl.2019.101966.
[10]
CHEN B S, LINKE A, OLSON L, et al. Greater functional connectivity between sensory networks is related to symptom severity in toddlers with autism spectrum disorder[J]. J Child Psychol Psychiatry, 2021, 62(2): 160-170. DOI: 10.1111/jcpp.13268.
[11]
ASSAF M, JAGANNATHAN K, CALHOUN V D, et al. Abnormal functional connectivity of default mode sub-networks in autism spectrum disorder patients[J]. Neuroimage, 2010, 53(1): 247-256. DOI: 10.1016/j.neuroimage.2010.05.067.
[12]
QIN J L, LI H X, WU Y, et al. Functional connectivity within the default mode network can predict the sleep disturbance scores of the patients with depression[J]. Chin J Magn Reson Imag, 2024, 15(7): 51-57. DOI: 10.12015/issn.1674-8034.2024.07.009.
[13]
SUN J W, FAN R, WANG Q, et al. Identify abnormal functional connectivity of resting state networks in Autism spectrum disorder and apply to machine learning-based classification[J/OL]. Brain Res, 2021, 1757: 147299 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/33516816/. DOI: 10.1016/j.brainres.2021.147299.
[14]
BERNAS A, BARENDSE E M, ALDENKAMP A P, et al. Brain resting-state networks in adolescents with high-functioning autism: Analysis of spatial connectivity and temporal neurodynamics[J/OL]. Brain Behav, 2018, 8(2): e00878 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/29484255/. DOI: 10.1002/brb3.878.
[15]
HE M R, MO X, YOU W F, et al. Changes of cerebral functional hierarchy in AD patients based on cerebral-cerebellar connectivity gradients[J]. Chin J Magn Reson Imag, 2024, 15(5): 1-7. DOI: 10.12015/issn.1674-8034.2024.05.001.
[16]
URCHS S G W, TAM A, ORBAN P, et al. Functional connectivity subtypes associate robustly with ASD diagnosis[J/OL]. Elife, 2022, 11: e56257 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/36444973/. DOI: 10.7554/eLife.56257.
[17]
RUAN L L, CHEN G X, YAO M L, et al. Brain functional gradient and structure features in adolescent and adult autism spectrum disorders[J/OL]. Hum Brain Mapp, 2024, 45(11): e26792 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39037170/. DOI: 10.1002/hbm.26792.
[18]
BILGEN I, GUVERCIN G, REKIK I. Machine learning methods for brain network classification: Application to autism diagnosis using cortical morphological networks[J/OL]. J Neurosci Methods, 2020, 343: 108799 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/32574639/. DOI: 10.1016/j.jneumeth.2020.108799.
[19]
BARIK K, WATANABE K, BHATTACHARYA J, et al. Functional connectivity based machine learning approach for autism detection in young children using MEG signals[J/OL]. J Neural Eng, 2023, 20(2): 026012 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/36812588/. DOI: 10.1088/1741-2552/acbe1f.
[20]
XU M, CALHOUN V, JIANG R T, et al. Brain imaging-based machine learning in autism spectrum disorder: methods and applications[J/OL]. J Neurosci Methods, 2021, 361: 109271 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/34174282/. DOI: 10.1016/j.jneumeth.2021.109271.
[21]
GRAÑA M, SILVA M. Impact of machine learning pipeline choices in autism prediction from functional connectivity data[J/OL]. Int J Neural Syst, 2021, 31(4): 2150009 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/33472548/. DOI: 10.1142/S012906572150009X.
[22]
ZHAO X H, GE M L, CHEN S H, et al. Features fusion of brain networks and its application to autism recognition by machine learning based on resting-state functional magnetic resonance imaging[J]. Chin J Magn Reson Imag, 2021, 12(12): 55-61. DOI: 10.12015/issn.1674-8034.2021.12.011.
[23]
HONG S J, DE WAEL R VOS, BETHLEHEM R A I, et al. Atypical functional connectome hierarchy in autism[J/OL]. Nat Commun, 2019, 10(1): 1022 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/30833582/. DOI: 10.1038/s41467-019-08944-1.
[24]
KANG L J, WANG W, ZHANG N, et al. Anhedonia and dysregulation of an angular gyrus-centred and dynamic functional network in adolescent-onset depression[J]. J Affect Disord, 2023, 324: 82-91. DOI: 10.1016/j.jad.2022.12.057.
[25]
LI G S, LIU Y J, ZHENG Y T, et al. Large-scale dynamic causal modeling of major depressive disorder based on resting-state functional magnetic resonance imaging[J]. Hum Brain Mapp, 2020, 41(4): 865-881. DOI: 10.1002/hbm.24845.
[26]
SAKOĞLU U, PEARLSON G D, KIEHL K A, et al. A method for evaluating dynamic functional network connectivity and task-modulation: application to schizophrenia[J]. MAGMA, 2010, 23(5/6): 351-366. DOI: 10.1007/s10334-010-0197-8.
[27]
JENSEN K M, TURNER J A, UDDIN L Q, et al. Addressing inconsistency in functional neuroimaging: a replicable data-driven multi-scale functional atlas for canonical brain networks[J/OL]. bioRxiv, 2024: 2024.09.09.612129 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39314443/. DOI: 10.1101/2024.09.09.612129.
[28]
WANG X G, CHEN H, WEN R, et al. Exploring functional and structural connectivity disruptions in spinocerebellar Ataxia type 3: Insights from gradient analysis[J/OL]. CNS Neurosci Ther, 2024, 30(7): e14842 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39014518/. DOI: 10.1111/cns.14842.
[29]
DE WAEL R VOS, BENKARIM O, PAQUOLA C, et al. BrainSpace: a toolbox for the analysis of macroscale gradients in neuroimaging and connectomics datasets[J/OL]. Commun Biol, 2020, 3(1): 103 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/32139786/. DOI: 10.1038/s42003-020-0794-7.
[30]
YAO C Y, SHAN Y, CUI B X, et al. Hyperconnectivity and connectome gradient dysfunction of cerebello-thalamo-cortical circuitry in Alzheimer's disease spectrum disorders[J/OL]. Cerebellum, 2025, 24(2): 43 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39913059/. DOI: 10.1007/s12311-025-01792-4.
[31]
MARGULIES D S, GHOSH S S, GOULAS A, et al. Situating the default-mode network along a principal gradient of macroscale cortical organization[J]. Proc Natl Acad Sci USA, 2016, 113(44): 12574-12579. DOI: 10.1073/pnas.1608282113.
[32]
ROYER J, PAQUOLA C, VALK S L, et al. Gradients of brain organization: smooth sailing from methods development to user community[J]. Neuroinformatics, 2024, 22(4): 623-634. DOI: 10.1007/s12021-024-09660-y.
[33]
JANG H, DAI R, MASHOUR G A, et al. Classifying unconscious, psychedelic, and neuropsychiatric brain states with functional connectivity, graph theory, and cortical gradient analysis[J/OL]. Brain Sci, 2024, 14(9): 880 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39335376/. DOI: 10.3390/brainsci14090880.
[34]
WANG Y W, CHEN X, YAN C G. Comprehensive evaluation of harmonization on functional brain imaging for multisite data-fusion[J/OL]. Neuroimage, 2023, 274: 120089 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/37086875/. DOI: 10.1016/j.neuroimage.2023.120089.
[35]
INGALHALIKAR M, SHINDE S, KARMARKAR A, et al. Functional connectivity-based prediction of autism on site harmonized ABIDE dataset[J]. IEEE Trans Biomed Eng, 2021, 68(12): 3628-3637. DOI: 10.1109/TBME.2021.3080259.
[36]
YU M C, LINN K A, SHINOHARA R T, et al. Childhood trauma history is linked to abnormal brain connectivity in major depression[J]. Proc Natl Acad Sci USA, 2019, 116(17): 8582-8590. DOI: 10.1073/pnas.1900801116.
[37]
LIDSTONE D E, ROCHOWIAK R, MOSTOFSKY S H, et al. A data driven approach reveals that anomalous motor system connectivity is associated with the severity of core autism symptoms[J/OL]. Autism Res, 2021 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/33484109/. DOI: 10.1002/aur.2476.
[38]
WYMBS N F, NEBEL M B, EWEN J B, et al. Altered inferior parietal functional connectivity is correlated with Praxis and social skill performance in children with autism spectrum disorder[J]. Cereb Cortex, 2021, 31(5): 2639-2652. DOI: 10.1093/cercor/bhaa380.
[39]
YU H, QU H, CHEN A G, et al. Alteration of effective connectivity in the default mode network of autism after an intervention[J/OL]. Front Neurosci, 2021, 15: 796437 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/35002608/. DOI: 10.3389/fnins.2021.796437.
[40]
BOS D J, VAN RAALTEN T R, ORANJE B, et al. Developmental differences in higher-order resting-state networks in autism spectrum disorder[J/OL]. Neuroimage Clin, 2014, 4: 820-827 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/24936432/. DOI: 10.1016/j.nicl.2014.05.007.
[41]
ZHANG W K, CAI K L, XIONG X, et al. Alterations of triple network dynamic connectivity and repetitive behaviors after mini-basketball training program in children with autism spectrum disorder[J/OL]. Sci Rep, 2025, 15(1): 2629 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39838077/. DOI: 10.1038/s41598-025-87248-5.
[42]
SHANG J, SHEN E W, YU Y, et al. Relationship between abnormal intrinsic functional connectivity of subcortices and autism symptoms in high-functioning adults with autism spectrum disorder[J/OL]. Psychiatry Res Neuroimaging, 2024, 337: 111762 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/38043369/. DOI: 10.1016/j.pscychresns.2023.111762.
[43]
VÁZQUEZ-RODRÍGUEZ B, SUÁREZ L E, MARKELLO R D, et al. Gradients of structure-function tethering across neocortex[J]. Proc Natl Acad Sci USA, 2019, 116(42): 21219-21227. DOI: 10.1073/pnas.1903403116.
[44]
DE ROSA A P, D'AMBROSIO A, BISECCO A, et al. Functional gradients reveal cortical hierarchy changes in multiple sclerosis[J/OL]. Hum Brain Mapp, 2024, 45(6): e26678 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/38647001/. DOI: 10.1002/hbm.26678.
[45]
UDDIN L Q, SUPEKAR K, MENON V. Reconceptualizing functional brain connectivity in autism from a developmental perspective[J/OL]. Front Hum Neurosci, 2013, 7: 458 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/23966925/. DOI: 10.3389/fnhum.2013.00458.
[46]
REN P C, BI Q S, PANG W B, et al. Stratifying ASD and characterizing the functional connectivity of subtypes in resting-state fMRI[J/OL]. Behav Brain Res, 2023, 449: 114458 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/37121277/. DOI: 10.1016/j.bbr.2023.114458.
[47]
SHAN X L, UDDIN L Q, MA R, et al. Disentangling the individual-shared and individual-specific subspace of altered brain functional connectivity in autism spectrum disorder[J]. Biol Psychiatry, 2024, 95(9): 870-880. DOI: 10.1016/j.biopsych.2023.09.012.
[48]
MAGNÉE M J C M, DE GELDER B, VAN ENGELAND H, et al. Multisensory integration and attention in autism spectrum disorder: evidence from event-related potentials[J/OL]. PLoS One, 2011, 6(8): e24196 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/21887382/. DOI: 10.1371/journal.pone.0024196.
[49]
NOONAN S K, HAIST F, MÜLLER R A. Aberrant functional connectivity in autism: evidence from low-frequency BOLD signal fluctuations[J/OL]. Brain Res, 2009, 1262: 48-63 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/19401185/. DOI: 10.1016/j.brainres.2008.12.076.
[50]
ROYER J, KEBETS V, PIGUET C, et al. Multimodal neural correlates of childhood psychopathology[J/OL]. Elife, 2024, 13: e87992 [2025-04-18]. https://pubmed.ncbi.nlm.nih.gov/39625475/. DOI: 10.7554/eLife.87992.

PREV Regulatory effect of gut-targeted therapy on brain function and emotional changes in patients with diarrhea-predominant irritable bowel syndrome: A resting-state fMRI study
NEXT Predictive efficiency of magnetic resonance vessel wall imaging on vertebrobasilar dolichoectasia with posterior circulation infarction
  


LINK


Tel & Fax: +8610-67113815    E-mail: editor@cjmri.cn