分享:
分享到微信朋友圈
X
综述
rs-fMRI结合机器学习在肠-脑轴中研究进展
巨妍 王嵩

Cite this article as: JU Y, WANG S. Advances in rs-fMRI combined with machine learning toward the gut-brain axis[J]. Chin J Magn Reson Imaging, 2023, 14(5): 171-174, 180.本文引用格式:巨妍, 王嵩. rs-fMRI结合机器学习在肠-脑轴中研究进展[J]. 磁共振成像, 2023, 14(5): 171-174, 180. DOI:10.12015/issn.1674-8034.2023.05.030.


[摘要] 肠道微生物与大脑之间的双向沟通机制称为肠-脑轴,肠-脑轴的紊乱与多种常见疾病相关,但目前临床确诊方法不完善。静息态功能磁共振成像(resting state functional MRI, rs-fMRI)技术作为重要的影像学工具,帮助提供脑部的功能变化情况;机器学习通过不同的特征提取方法、分类算法等建立预测模型,二者结合常应用于疾病的诊断、分类和预后等方面。本文综述了rs-fMRI结合机器学习应用于肠-脑轴相关的胃肠及主要神经类疾病的研究,旨在为相关模型建立、辅助临床诊断、实现精准医疗提供技术参考。
[Abstract] The two-way communication between gut microbes and the brain is called the gut-brain axis. Disorders of the gut-brain axis are associated with many diseases. However, the current clinical diagnosis method is not perfect. Resting state functional magnetic resonance imaging (rs-fMRI) is an important imaging tool that helps provide information about changes in brain function; machine learning builds prediction models by selecting different feature extraction methods and classification algorithms. The combination of the two is often used in the diagnosis, classification and prognosis of diseases. This article reviews the application of rs-fMRI combined with machine learning to gastrointestinal and major neurological diseases related to the gut-brain axis, aims to provide technical reference for the establishment of relevant models, assist clinical diagnosis, and realize precision medicine.
[关键词] 肠-脑轴;磁共振成像;静息态功能磁共振成像;机器学习;深度学习
[Keywords] gut-brain axis;magnetic resonance imaging;resting state functional magnetic resonance imaging;machine learning;deep learning

巨妍    王嵩 *  

上海中医药大学附属龙华医院放射科,上海 200032

通信作者:王嵩,E-mail:songwangws@163.com

作者贡献声明:王嵩设计本研究的方案,并对稿件的重要内容进行了修改,获得了上海市自然科学基金的资助;巨妍参与构思了这项研究,进行文献的检索并撰写了稿件;全体作者同意发表最后的修改稿,同意对本研究的所有方面负责,确保本研究的准确性和诚信。


基金项目: 上海市自然科学基金 19ZR1457800
收稿日期:2022-09-07
接受日期:2023-05-06
中图分类号:R445.2  R574.4  R749 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2023.05.030
本文引用格式:巨妍, 王嵩. rs-fMRI结合机器学习在肠-脑轴中研究进展[J]. 磁共振成像, 2023, 14(5): 171-174, 180. DOI:10.12015/issn.1674-8034.2023.05.030.

0 前言

       胃肠道和大脑之间的双向沟通通路称为肠-脑轴(gut-brain-axis, GBA)。其通信网络包括中枢神经系统、肠神经系统、自主神经系统和下丘脑-垂体-肾上腺轴[1]。肠道内的微生物区系不仅能维持胃肠道的正常活动,而且影响中枢神经系统活动;神经系统亦通过下丘脑-垂体-肾上腺轴的调节、代谢物反馈等影响肠道微生物的多样性。鉴于GBA的沟通机制和环境因素的复杂性,相同的病因可导致不同的表型,相同的症状可来自不同原因,疾病诊断尚存在延后性。因此目前研究注重寻求数据驱动进行疾病机制的定义,推进生物标志物的发现,寻找最佳诊疗方法。

       静息态功能磁共振成像(resting state functional MRI, rs-fMRI)技术能够精确地显示激活脑区的部位和程度,目前rs-fMRI的多种分析方法已广泛应用于GBA相关疾病的研究中,包括基于种子点的功能连接(functional connectivity, FC)、局部一致性(regional homogeneity, ReHo)、低频振幅(amplitude low-frequency fluctuations, ALFF)等[2]。目前开展的研究已发现相关疾病具有功能异常的差异脑区,但部分脑区在不同疾病中的结果具有普适性。而机器学习(machine learning, ML)与rs-fMRI的结合能够进一步进行数据特征提取,评估鉴别诊断敏感性,以望进行疾病的特异性检查。本文旨在阐述rs-fMRI结合ML在GBA相关胃肠及神经类疾病中应用,揭示文献中关键一致或新颖的结论,为该领域未来发展提供参考。

1 rs-fMRI与ML的结合应用

1.1 胃肠道疾病

       肠易激综合征(irritable bowel syndrome, IBS)表现为复发性腹痛以及粪便形式或频率的改变,严重影响患者生活质量。研究表明IBS患者肠道菌群失衡导致肠黏膜肥大细胞功能受损,分泌介质作用于邻近的内分泌细胞和神经纤维释放神经递质(如5-羟色胺),影响肠蠕动,并将信息传递到神经中枢,导致神经系统高敏感性,从而导致IBS症状[3]。但仍未发现可检测到的器质性原因。MAO等[4]将rs-fMRI数据与支持向量机(support vector machine, SVM)分类器结合进行IBS患者与健康对照(healthy controls, HC)组的区分,以缰核为种子点进行无向FC和有效连接分析,结果发现IBS患者额叶皮层FC显著增强,背外侧丘脑的FC降低,从右丘脑到左丘脑的有效连接降低。并且将全连接模型作为输入特征较单一指标实现了模型准确率的提升,达71.5%,提示特征的全面性更易于寻找疾病自身独特的标志物,为探索特异性诊断指标提供思路。

       功能性消化不良(functional dyspepsia, FD)是一种常见的胃肠功能性疾病,FD患者体内不仅发现十二指肠嗜酸性粒细胞和肥大细胞浸润增加[5],还存在中枢神经变化和下行迷走神经信号传导受损[6]。临床上通常根据上腹痛的症状诊断,缺乏实质的检查指征。目前,rs-fMRI研究同样在FD人群开展,包括与HC人群的区分[7]——差异脑区主要位于与内脏感觉处理和情感反应相关的丘脑、内外囊和前扣带皮层;亚型的分类[8]发现上腹痛综合征亚型患者额叶ALFF值特异性增加,餐后窘迫症亚型患者扣带回皮层ALFF值特异性增加。YIN等[9]进一步通过独立成分分析(independent component analysis, ICA)方法进行脑网络特征的提取,与SVM分类器结合后,不仅进行FD的诊断,且关注FD患者对针灸治疗效果的反应[10],最终获得84%和76%的准确率,表明ICA无监督特征提取方法的可行性,并为体现治疗疗效提供了一种有希望的方法。

       肥胖为肠道微生物群另一个公认的参与领域,通过高脂饮食可改变肠道微生物的组成、大脑和外周组织中的内源性大麻素水平,引发内毒素血症诱导血-脑屏障、白细胞介素1β等介导神经炎症反应[11],从而与食物成瘾相关联。DONG等[12]在女性肥胖群体的研究中发现脑干和壳核FC值增加,并将代谢、rs-fMRI数据及测序等差异特征输入随机森林分类器,5折交叉验证后曲线下面积(area under the curve, AUC)为0.81,其中rs-fMRI变量对机器准确性贡献重大;之后进一步通过rs-fMRI图论方法发现伏隔核是大脑奖励中心的中间区域[13],证实GBA轴参与宿主的代谢、能量消耗以及摄入行为,表明rs-fMRI数据中包含重要的生物学信息,具有可信性,且ML模型是量化指标的有力工具。

1.2 神经系统疾病

       阿尔茨海默病(Alzheimer's disease, AD)是一种起病隐匿且进行性发展的神经系统退行性病变,是生活中痴呆的常见原因,病因仍未明确,其生物标记物被认为是脑脊液淀粉样蛋白以及磷酸化tau等[14]。研究发现衰老可导致肠道微生物成分改变,促炎菌丰度高,诱发炎症导致肠道通透性增强,损害血脑屏障,从而促进神经元的损伤和死亡[15]。目前,rs-fMRI技术结合ML建立模型在AD中全面开展,研究已不局限于单rs-fMRI数据特征,大多采用两种及以上特征指标进行AD的诊断及与前期认知障碍亚型的分类,两项研究均将ReHo值和ALFF值与SVM分类器进行结合[16, 17],亚型分类准确率均大于90%,再次强调了特征融合能够包含更多分类信息,并提出典型相关分析(canonical correlation analysis, CCA)是融合rs-fMRI数据特征的有效方法。不仅如此,XU等[18]通过多核SVM分类器驱动发现主观认知能力下降(subject cognitive decline, SCD)和HC共同枢纽位于前额叶皮质,前扣带和颞极仅存在于SCD,较单独使用rs-fMRI技术结果更接近真实,并指出较SVM模型,使用多核SVM分类器能够提高分类性能。此外,研究不只关注静态脑功能信息,更包括高阶脑网络信号:有效连接及动态FC是目前研究的热点[19, 20, 21],LI等[22]发现有向网络模型比无向表现出更好的分类性能;ABROL等[23]提出动态指标比静态指标更具有预测性。在此类高阶数据特征研究中,研究者更偏向于选择结合3D卷积神经网络(convolutional neural networks, CNN)模型,发挥深度学习(deep learning, DL)的高性能优势,最终几项研究AUC均大于0.8。除此以外,研究首次将脑白质的功能信息与SVM结合[24, 25],同样发现白质中静、动态FC信息,在AD预测中准确率同样大于80%,提出一种了解AD功能进展障碍的新途径。不只限于AD的诊断,DUC等[26]使用ICA方法提取脑网络结合3D-CNN架构,在认知状态评价量表评分的预测方面同样具有可行性,并实践证明最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)具有提高预测准确性、减少过拟合的优点;MILLAR等[27]将FC结合高斯回归模型进行AD患者年龄的预测,结果表明FC预测的脑年龄存在作为AD敏感标志物的可能,目前此方面未得到充分探索,今后可进一步研究。

       帕金森病(Parkinson's disease, PD)是第二常见的神经退行性病变,病因机制不清晰,病情呈持续性进展[28]。BRAAK等[29]首次提出“肠脑”假说:引起PD的病原体通过胃肠道黏膜屏障,在肠神经系统引起α-突触核蛋白聚集,再通过迷走神经逆行到大脑。近期研究同样发现PD患者的胃肠病理变化可能先于中枢神经系统的改变[30]。目前,ICA结合SVM同样是PD疾病中常用的分类模型[31, 32]。不仅如此,RUBBERT等[33]训练logistic回归模型进行PD与HC的二分类,同样具有高准确性及敏感性。总结发现,不只限于进行PD与HC的预测、亚型分类,以及BAGGIO等[31]首次进行的PD与多系统萎缩患者的高精度区分,ML模型验证后均强调了rs-fMRI数据中小脑区域的重要性,提示小脑区域可能是PD疾病重要的高精度特异性区域。SHI等的三项研究[34, 35, 36]将采集的多中心数据的ALFF值与SVM结合,发现PD可能与感觉运动网络和顶叶外侧皮层的异常大脑活动相关,在外部验证集中具有良好的泛化;并且发现脑网络组246图谱分类性能优于常用的自动解剖标记90图谱。此外不仅局限于大脑图谱的特征,还提取了ALFF直方图特征,包括最小值、标准差等,全面反映ALFF值信息,将AUC提升到0.97,为模型准确性的提高提供了可行方法。除疾病的诊断应用之外,BOUTET等[37]将应用范围拓展到深部脑刺激护理疗效预测,关注丘脑、辅助运动等区域的特征,准确率达89%,研究者认为rs-fMRI对PD患者深部脑刺激的反映可以代表临床反应的客观生物标志物,为rs-fMRI结合ML应用拓展了新的领域。

       rs-fMRI技术结合ML模型在神经类疾病中开展较为成熟,本文只选取两类代表性神经退行性疾病进行总结,在其他神经类疾病如多发性硬化[38, 39]、自闭症[40, 41]、焦虑[38]、抑郁[39, 40, 41]等精神类疾病中同样发挥优势。

2 影响模型准确性因素

2.1 数据属性

       数据属性的主要方面之一为样本量的大小,总结发现涉及研究大多来自于自行收集并非开放数据库,病例的实际收取具有困难性,所以样本量大多集中于200以内。研究表明小样本可能会导致结果不可靠[42],所以小样本研究均将扩大样本量作为今后的改进方向。且目前对训练集大小的确定方法研究稀缺[43],未来应强调样本量确定方法的标准化。另一方面发现涉及研究中收集的样本纳入标准之间存在差异,包括年龄范围差异、单中心或者多中心收取等,如此,将人群甚至扫描机器的差异引进数据中可导致数据异质性过大而降低模型性能。因此,今后的病例收集中应在保证降低异质性的前提下扩大样本量实现高精度分类,降低数据属性对模型准确性的影响。

2.2 分类模型

       研究者大多选择SVM分类器,因其在小样本研究中能够发挥减少过拟合并且提供良好的泛化性能[44]的优势。且基于不同的内核进行线性和非线性的转化时,内核的选择可能会极大地影响模型性能,例如XU等[18]的研究中通过将SVM转化为多核SVM性能得到了提升。此外在复杂特征以及大样本研究中CNN等复杂网络具有更好的表现,能够发挥DL网络深、容量足的优势。在THOMAS等[45]的总结中发现,SVM与CNN分类性能表现相当,但同时有研究称在相同条件下DL比传统ML表现更好[46]。因CNN同时具有需要调参数、数据要求高、黑盒等劣势,在实际选择中需根据自身数据属性进行优势分类器的选择。

2.3 特征选择方法

       多项研究[4,16, 17]证明适当包含更多的生物学信息以及增加输入特征可以作为提高准确性的方法。但由于rs-fMRI数据自身具有高维特性,数据的直接串联可能会导致特征冗余,因此研究中使用的CCA、LASSO算法、递归特征消除支持向量机等能够帮助识别最有效特征,消除无关变量以达到降低特征维数、进行特征选择的目的,增强模型判别力。文献[22,25, 26]中通过增加是否进行特征选择的对照研究,证实适当进行特征选择能够提高分类性能。

2.4 rs-fMRI指标

       先前试验中研究者大多选择静态指标,包括静态FC、ALFF等,由于大脑具有动态特性,静态指标对脑功能信息反映不全面。近期研究者侧重于关注动态指标以及具有方向性的有效连接,作为高阶输入特征或者静态指标的补充,检测除静态指标以外的组间差异。实践证明,静态转化为动态[23]、无向转化为有向[22]、动态作为补充的全连接模型[4]均实现了性能的提升。但同时有研究具有矛盾的结果[47],今后可进一步扩展全连接指标的应用,充分了解潜在机制后恰当使用。

2.5 图谱

       图谱的选择会影响研究区域的定位,自动解剖标记图谱在rs-fMRI研究中使用最广泛。目前研究者开始使用不同图谱测试准确性[9],从而作为评价模型性能稳健性的一种方法,并且在对比试验中SHI等[35]证实脑网络组246图谱分类性能优于常用的自动解剖标记90图谱。因此,图谱不仅影响图像分割的精度和计算效率,并且是对判别性结果解释的关键,对模型精度起重要作用。

3 小结

       IBS、FD、肥胖以及AD、PD等与GBA的紊乱相关,GBA的通路均参与发病机制,并且研究已证实以上疾病能够相互影响[48, 49, 50],具有关联性。目前rs-fMRI技术在上述疾病中已广泛开展,数据中包含丰富的生物学信息,与ML的结合广泛用于分类模型的构建,不仅验证了rs-fMRI数据特征的可靠性,同时更加扩展了应用范围,帮助寻找疾病潜在的神经影像学的生物标志物,并帮助疾病诊断、亚型分类以及疗效预测。但仍存在以下不足:病例的收集存在差异、ML模型框架不公开、缺乏实施细节的问题,试验难以复制,甚至在研究之间无法比较机器性能,为模型的优化造成困难;并且,模型的建立大多集中于疾病的诊断和亚型的分类,疾病的预后情况如预测并发症、药物治疗反应等可在今后进一步研究。近年来关于GBA的研究稳步增加,相信基于现有的改进技术,结合不同方法的优越性,能够获得针对rs-fMRI与ML结合的先进模型,未来GBA疾病的合并治疗成为可能。

[1]
JARRET A, JACKSON R, DUIZER C, et al. Enteric Nervous System-Derived IL-18 Orchestrates Mucosal Barrier Immunity[J]. Cell, 2020, 180(1): 50-63. DOI: 10.1016/j.cell.2019.12.016">10.1016/j.cell.2019.12.016">10.1016/j.cell.2019.12.016.
[2]
LI B Z, CAO Y, ZHANG Y, et al. Relation of Decreased Functional Connectivity Between Left Thalamus and Left Inferior Frontal Gyrus to Emotion Changes Following Acute Sleep Deprivation[J]. Front Neurol, 2021, 12: 642411. DOI: 10.3389/fneur.2021.642411">10.3389/fneur.2021.642411">10.3389/fneur.2021.642411.
[3]
MAYER E A, SAVIDGE T, SHULMAN R J. Brain-gut microbiome interactions and functional bowel disorders[J]. Gastroenterology, 2014, 146(6): 1500-1512. DOI: 10.1053/j.gastro.2014.02.037">10.1053/j.gastro.2014.02.037">10.1053/j.gastro.2014.02.037.
[4]
MAO C P, CHEN F R, HUO J H, et al. Altered resting-state functional connectivity and effective connectivity of the habenula in irritable bowel syndrome: A cross-sectional and machine learning study[J]. Hum Brain Mapp, 2020, 41(13): 3655-3666. DOI: 10.1002/hbm.25038">10.1002/hbm.25038">10.1002/hbm.25038.
[5]
CEULEMANS M, JACOBS I, WAUTERS L, et al. Immune Activation in Functional Dyspepsia: Bystander Becoming the Suspect[J]. Front Neurosci, 2022, 16: 831761. DOI: 10.3389/fnins.2022.831761">10.3389/fnins.2022.831761">10.3389/fnins.2022.831761.
[6]
LI H, PAGE A J. Altered Vagal Signaling and Its Pathophysiological Roles in Functional Dyspepsia[J]. Front Neurosci, 2022, 16: 858612. DOI: 10.3389/fnins.2022.858612">10.3389/fnins.2022.858612">10.3389/fnins.2022.858612.
[7]
LIU P, FAN Y, WEI Y, et al. Altered structural and functional connectivity of the insula in functional dyspepsia[J/OL]. Neurogastroenterol Motil, 2018, 30(9): e13345 [2022-11-01]. https://doi.org/10.1111/nmo.13345. DOI: 10.1111/nmo.13345">10.1111/nmo.13345">10.1111/nmo.13345.
[8]
QI R, SHI Z, WENG Y, et al. Similarity and diversity of spontaneous brain activity in functional dyspepsia subtypes[J]. Acta Radiol, 2020, 61(7): 927-935. DOI: 10.1177/0284185119883391">10.1177/0284185119883391">10.1177/0284185119883391.
[9]
YIN T, SUN R, HE Z, et al. Subcortical-cortical functional connectivity as a potential biomarker for identifying patients with functional dyspepsia[J]. Cereb Cortex, 2022, 32(15): 3347-3358. DOI: 10.1093/cercor/bhab419">10.1093/cercor/bhab419">10.1093/cercor/bhab419.
[10]
YIN T, HE Z, CHEN Y, et al. Predicting acupuncture efficacy for functional dyspepsia based on functional brain network features: a machine learning study[J]. Cereb Cortex, 2023, 21, 33(7): 3511-3522. DOI: 10.1093/cercor/bhac288">10.1093/cercor/bhac288">10.1093/cercor/bhac288.
[11]
FORTE N, FERNÁNDEZ-RILO A C, Palomba L, et al. Obesity Affects the Microbiota-Gut-Brain Axis and the Regulation Thereof by Endocannabinoids and Related Mediators[J]. Int J Mol Sci, 2020, 21(5): 1554. DOI: 10.3390/ijms21051554">10.3390/ijms21051554">10.3390/ijms21051554.
[12]
DONG T S, MAYER E A, OSADCHIY V, et al. A Distinct Brain-Gut-Microbiome Profile Exists for Females with Obesity and Food Addiction[J]. Obesity (Silver Spring), 2020, 28(8): 1477-1486. DOI: 10.1002/oby.22870">10.1002/oby.22870">10.1002/oby.22870.
[13]
DONG T S, GUAN M, MAYER E A, et al. Obesity is associated with a distinct brain-gut microbiome signature that connects Prevotella and Bacteroides to the brain's reward center[J]. Gut Microbes, 2022, 14(1): 2051999. DOI: 10.1080/19490976.2022.2051999">10.1080/19490976.2022.2051999">10.1080/19490976.2022.2051999.
[14]
HENEKA M T, CARSON M J, EL K J, et al. Neuroinflammation in Alzheimer's disease[J]. Lancet Neurol, 2015, 14(4): 388-405. DOI: 10.1016/S1474-4422(15)70016-5">10.1016/S1474-4422(15)70016-5">10.1016/S1474-4422(15)70016-5.
[15]
KESIKA P, SUGANTHY N, SIVAMARUTHI B S, et al. Role of gut-brain axis, gut microbial composition, and probiotic intervention in Alzheimer's disease[J]. Life Sci, 2021, 264: 118627. DOI: 10.1016/j.lfs.2020.118627">10.1016/j.lfs.2020.118627">10.1016/j.lfs.2020.118627.
[16]
ZHANG Q, WANG Q, HE C, et al. Altered Regional Cerebral Blood Flow and Brain Function Across the Alzheimer's Disease Spectrum: A Potential Biomarker[J]. Front Aging Neurosci, 2021, 13: 630382. DOI: 10.3389/fnagi.2021.630382">10.3389/fnagi.2021.630382">10.3389/fnagi.2021.630382.
[17]
JIA H, WANG Y, DUAN Y, et al. Alzheimer's Disease Classification Based on Image Transformation and Features Fusion[J]. Comput Math Methods Med, 2021, 2021: 9624269. DOI: 10.1155/2021/9624269">10.1155/2021/9624269">10.1155/2021/9624269.
[18]
XU X, LI W, TAO M, et al. Effective and Accurate Diagnosis of Subjective Cognitive Decline Based on Functional Connection and Graph Theory View[J]. Front Neurosci, 2020, 14: 577887. DOI: 10.3389/fnins.2020.577887">10.3389/fnins.2020.577887">10.3389/fnins.2020.577887.
[19]
LEI B, YU S, ZHAO X, et al. Diagnosis of early Alzheimer's disease based on dynamic high order networks[J]. Brain Imaging Behav, 2021, 15(1): 276-287. DOI: 10.1007/s11682-019-00255-9">10.1007/s11682-019-00255-9">10.1007/s11682-019-00255-9.
[20]
WU Y, ZHOU Y, SONG M. Classification of patients with AD from healthy controls using entropy-based measures of causality brain networks[J]. J Neurosci Methods, 2021, 361: 109265. DOI: 10.1016/j.jneumeth.2021.109265">10.1016/j.jneumeth.2021.109265">10.1016/j.jneumeth.2021.109265.
[21]
LIN K, JIE B, DONG P, et al. Convolutional Recurrent Neural Network for Dynamic Functional MRI Analysis and Brain Disease Identification[J]. Front Neurosci, 2022, 16: 933660. DOI: 10.3389/fnins.2022.933660">10.3389/fnins.2022.933660">10.3389/fnins.2022.933660.
[22]
LI X, YANG C, XIE P, et al. The diagnosis of amnestic mild cognitive impairment by combining the characteristics of brain functional network and support vector machine classifier[J]. J Neurosci Methods, 2021, 363: 109334. DOI: 10.1016/j.jneumeth.2021.109334">10.1016/j.jneumeth.2021.109334">10.1016/j.jneumeth.2021.109334.
[23]
ABROL A, FU Z, DU Y, et al. Multimodal Data Fusion of Deep Learning and Dynamic Functional Connectivity Features to Predict Alzheimer's Disease Progression[J]. Annu Int Conf IEEE Eng Med Biol Soc, 2019, 2019: 4409-4413. DOI: 10.1109/EMBC.2019.8856500">10.1109/EMBC.2019.8856500">10.1109/EMBC.2019.8856500.
[24]
ZHAO J, DING X, DU Y, et al. Functional connectivity between white matter and gray matter based on fMRI for Alzheimer's disease classification[J/OL]. Brain Behav, 2019, 9(10): e1407 [2022-11-01]. https://doi.org/10.1002/brb3.1407. DOI: 10.1002/brb3.1407">10.1002/brb3.1407">10.1002/brb3.1407.
[25]
GAO Y, SENGUPTA A, LI M, et al. Functional connectivity of white matter as a biomarker of cognitive decline in Alzheimer's disease[J/OL]. PLoS One, 2020, 15(10): e240513 [2022-11-01]. https://doi.org/10.1371/journal.pone.0240513. DOI: 10.1371/journal.pone.0240513">10.1371/journal.pone.0240513">10.1371/journal.pone.0240513.
[26]
DUC N T, RYU S, QURESHI M, et al. 3D-Deep Learning Based Automatic Diagnosis of Alzheimer's Disease with Joint MMSE Prediction Using Resting-State fMRI[J]. Neuroinformatics, 2020, 18(1): 71-86. DOI: 10.1007/s12021-019-09419-w">10.1007/s12021-019-09419-w">10.1007/s12021-019-09419-w.
[27]
MILLAR P R, LUCKETT P H, GORDON B A, et al. Predicting brain age from functional connectivity in symptomatic and preclinical Alzheimer disease[J]. Neuroimage, 2022, 256: 119228. DOI: 10.1016/j.neuroimage.2022.119228">10.1016/j.neuroimage.2022.119228">10.1016/j.neuroimage.2022.119228.
[28]
TYSNES O B, STORSTEIN A. Epidemiology of Parkinson's disease[J]. J Neural Transm (Vienna), 2017, 124(8): 901-905. DOI: 10.1007/s00702-017-1686-y">10.1007/s00702-017-1686-y">10.1007/s00702-017-1686-y.
[29]
BRAAK H, RÜB U, GAI W P, et al. Idiopathic Parkinson's disease: possible routes by which vulnerable neuronal types may be subject to neuroinvasion by an unknown pathogen[J]. J Neural Transm (Vienna), 2003, 110(5): 517-536. DOI: 10.1007/s00702-002-0808-2">10.1007/s00702-002-0808-2">10.1007/s00702-002-0808-2.
[30]
CRYAN J F, O'RIORDAN K J, SANDHU K, et al. The gut microbiome in neurological disorders[J]. Lancet Neurol, 2020, 19(2): 179-194. DOI: 10.1016/S1474-4422(19)30356-4">10.1016/S1474-4422(19)30356-4">10.1016/S1474-4422(19)30356-4.
[31]
BAGGIO H C, ABOS A, SEGURA B, et al. Cerebellar resting-state functional connectivity in Parkinson's disease and multiple system atrophy: Characterization of abnormalities and potential for differential diagnosis at the single-patient level[J]. Neuroimage Clin, 2019, 22: 101720. DOI: 10.1016/j.nicl.2019.101720">10.1016/j.nicl.2019.101720">10.1016/j.nicl.2019.101720.
[32]
PANG H, YU Z, YU H, et al. Use of machine learning method on automatic classification of motor subtype of Parkinson's disease based on multilevel indices of rs-fMRI[J]. Parkinsonism Relat Disord, 2021, 90: 65-72. DOI: 10.1016/j.parkreldis.2021.08.003">10.1016/j.parkreldis.2021.08.003">10.1016/j.parkreldis.2021.08.003.
[33]
RUBBERT C, MATHYS C, JOCKWITZ C, et al. Machine-learning identifies Parkinson's disease patients based on resting-state between-network functional connectivity[J]. Br J Radiol, 2019, 92(1101): 20180886. DOI: 10.1259/bjr.20180886">10.1259/bjr.20180886">10.1259/bjr.20180886.
[34]
SHI D, ZHANG H, WANG S, et al. Application of Functional Magnetic Resonance Imaging in the Diagnosis of Parkinson's Disease: A Histogram Analysis[J]. Front Aging Neurosci, 2021, 13: 624731. DOI: 10.3389/fnagi.2021.624731">10.3389/fnagi.2021.624731">10.3389/fnagi.2021.624731.
[35]
SHI D, YAO X, LI Y, et al. Classification of Parkinson's disease using a region-of-interest- and resting-state functional magnetic resonance imaging-based radiomics approach[J]. Brain Imaging Behav, 2022, 16(5): 2150-2163. DOI: 10.1007/s11682-022-00685-y">10.1007/s11682-022-00685-y">10.1007/s11682-022-00685-y.
[36]
SHI D, ZHANG H, WANG G, et al. Machine Learning for Detecting Parkinson's Disease by Resting-State Functional Magnetic Resonance Imaging: A Multicenter Radiomics Analysis[J]. Front Aging Neurosci, 2022, 14: 806828. DOI: 10.3389/fnagi.2022.806828">10.3389/fnagi.2022.806828">10.3389/fnagi.2022.806828.
[37]
BOUTET A, MADHAVAN R, ELIAS G, et al. Predicting optimal deep brain stimulation parameters for Parkinson's disease using functional MRI and machine learning[J]. Nat Commun, 2021, 12(1): 3043. DOI: 10.1038/s41467-021-23311-9">10.1038/s41467-021-23311-9">10.1038/s41467-021-23311-9.
[38]
KIM B H, KIM M K, JO H J, et al. Predicting social anxiety in young adults with machine learning of resting-state brain functional radiomic features[J]. Sci Rep, 2022, 12(1): 13932. DOI: 10.1038/s41598-022-17769-w">10.1038/s41598-022-17769-w">10.1038/s41598-022-17769-w.
[39]
SHI Y, ZHANG L, WANG Z, et al. Multivariate Machine Learning Analyses in Identification of Major Depressive Disorder Using Resting-State Functional Connectivity: A Multicentral Study[J]. ACS Chem Neurosci, 2021, 12(15): 2878-2886. DOI: 10.1021/acschemneuro.1c00256">10.1021/acschemneuro.1c00256">10.1021/acschemneuro.1c00256.
[40]
DAI P, XIONG T, ZHOU X, et al. The alterations of brain functional connectivity networks in major depressive disorder detected by machine learning through multisite rs-fMRI data[J]. Behav Brain Res, 2022, 435: 114058. DOI: 10.1016/j.bbr.2022.114058">10.1016/j.bbr.2022.114058">10.1016/j.bbr.2022.114058.
[41]
CHUN J Y, SENDI M, SUI J, et al. Visualizing Functional Network Connectivity Difference between Healthy Control and Major Depressive Disorder Using an Explainable Machine-learning Method[J]. Annu Int Conf IEEE Eng Med Biol Soc, 2020, 2020: 1424-1427. DOI: 10.1109/EMBC44109.2020.9175685">10.1109/EMBC44109.2020.9175685">10.1109/EMBC44109.2020.9175685.
[42]
AN C, PARK Y W, AHN S S, et al. Radiomics machine learning study with a small sample size: Single random training-test set split may lead to unreliable results[J/OL]. PLoS One, 2021, 16(8): e256152 [2022-11-01]. https://doi.org/10.1371/journal.pone.0256152. DOI: 10.1371/journal.pone.0256152.
[43]
BALKI I, AMIRABADI A, LEVMAN J, et al. Sample-Size Determination Methodologies for Machine Learning in Medical Imaging Research: A Systematic Review[J]. Can Assoc Radiol J, 2019, 70(4): 344-353. DOI: 10.1016/j.carj.2019.06.002.
[44]
MO J, LIU Z, SUN K, et al. Automated detection of hippocampal sclerosis using clinically empirical and radiomics features[J]. Epilepsia, 2019, 60(12): 2519-2529. DOI: 10.1111/epi.16392.
[45]
THOMAS R M, GALLO S, CERLIANI L, et al. Classifying Autism Spectrum Disorder Using the Temporal Statistics of Resting-State Functional MRI Data With 3D Convolutional Neural Networks[J]. Front Psychiatry, 2020, 11: 440. DOI: 10.3389/fpsyt.2020.00440.
[46]
YANG M, CAO M, CHEN Y, et al. Large-Scale Brain Functional Network Integration for Discrimination of Autism Using a 3-D Deep Learning Model[J]. Front Hum Neurosci, 2021, 15: 687288. DOI: 10.3389/fnhum.2021.687288.
[47]
THOMPSON G J. Neural and metabolic basis of dynamic resting state fMRI[J]. Neuroimage, 2018, 180(Pt B): 448-462. DOI: 10.1016/j.neuroimage.2017.09.010.
[48]
LU Y, CHEN M, HUANG Z, et al. Antidepressants in the Treatment of Functional Dyspepsia: A Systematic Review and Meta-Analysis[J/OL]. PLoS One, 2016, 11(6): e157798 [2022-11-01]. https://doi.org/10.1371/journal.pone.0157798. DOI: 10.1371/journal.pone.0157798.
[49]
FU P, GAO M, YUNG K. Association of Intestinal Disorders with Parkinson's Disease and Alzheimer's Disease: A Systematic Review and Meta-Analysis[J]. ACS Chem Neurosci, 2020, 11(3): 395-405. DOI: 10.1021/acschemneuro.9b00607.
[50]
PRADILLO J M, HERNÁNDEZ-JIMÉNEZ M, FERNÁNDEZ-VALLE M E, et al. Influence of metabolic syndrome on post-stroke outcome, angiogenesis and vascular function in old rats determined by dynamic contrast enhanced MRI[J]. J Cereb Blood Flow Metab, 2021, 41(7): 1692-1706. DOI: 10.1177/0271678X20976412.

上一篇 基于脑磁共振成像探讨冠状动脉钙化与认知障碍的潜在机制研究进展和展望
下一篇 基于多模态MRI影像组学与深度学习的脑胶质瘤诊断及预后预测研究进展
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2