这是用户在 2024-5-6 24:57 为 https://app.immersivetranslate.com/pdf-pro/194d0f58-9c84-49a2-b7a5-41c3e82f7972 保存的双语快照页面,由 沉浸式翻译 提供双语支持。了解如何保存?
2024_05_06_6313a8c85e795cdec6c9g

Connectome reorganization associated with temporal lobe pathology and its surgical resection
与颞叶病理学及其手术切除相关的连接组重组

Sara Larivière, Bo-yong Park,, Jessica Royer, Jordan DeKraker, Alexander Ngo, Ella
Sara Larivière, Bo-yong Park,, Jessica Royer, Jordan DeKraker, Alexander Ngo, Ella

Sahlas, Judy Chen, Raúl Rodríguez-Cruces, Yifei Weng, Birgit Frauscher, Ruoting Liu, Zhengge Wang, Golia Shafiei, Bratislav Mišić, Andrea Bernasconi, Neda Bernasconi, , Michael D. Fox, Zhiqiang Zhang and Boris C. Bernhardt
Sahlas, Judy Chen, Raúl Rodríguez-Cruces, Weng Yifei, Birgit Frauscher, Ruoting Liu, Zhengge Wang, Golia Shafiei, Bratislav Mišić, Andrea Bernasconi, Neda Bernasconi, , Michael D.Fox, Zhiqiang Zhang and Boris C. Bernhardt

These authors contributed equally to this work.
这些作者对本研究做出了同等贡献。

Abstract 摘要

Network neuroscience offers a unique framework to understand the organizational principles of the human brain. Despite recent progress, our understanding of how the brain is modulated by focal lesions remains incomplete. Resection of the temporal lobe is the most effective treatment to control seizures in pharmaco-resistant temporal lobe epilepsy (TLE), making this syndrome a powerful model to study lesional effects on network organization in young and middle-aged adults. Here, we assessed the downstream consequences of a focal lesion and its surgical resection on the brain's structural connectome, and explored how this reorganization relates to clinical variables at the individual patient level.
网络神经科学为了解人脑的组织原理提供了一个独特的框架。尽管最近取得了一些进展,但我们对大脑如何受局灶病变调节的理解仍不全面。颞叶切除术是控制药物难治性颞叶癫痫(TLE)发作的最有效治疗方法,因此这种综合征成为研究病变对中青年网络组织影响的有力模型。在这里,我们评估了局灶性病变及其手术切除对大脑结构连接组的下游影响,并探讨了这种重组与患者个体水平的临床变量之间的关系。
We included adults with pharmaco-resistant TLE who underwent anterior temporal lobectomy between two imaging time points, as well as age- and sex-matched healthy controls who underwent comparable imaging ( ). Core to our analysis was the projection of high-dimensional structural connectome data-derived from diffusion MRI tractography from each subject-into lower-dimensional gradients. We then compared connectome gradients in patients relative to controls before surgery, tracked surgicallyinduced connectome reconfiguration from pre- to postoperative time points, and examined associations to patient-specific clinical and imaging phenotypes.
我们纳入了在两个成像时间点之间接受了前颞叶切除术的药物耐受性TLE成人 ,以及接受了类似成像的年龄和性别匹配的健康对照组( )。我们分析的核心是将来自每个受试者扩散磁共振成像束成像的高维结构连接组数据投影到低维梯度中。然后,我们比较了手术前患者与对照组的连接组梯度,追踪了手术诱导的连接组从术前到术后时间点的重构,并研究了与患者特定临床和成像表型的关联。
Before surgery, individuals with TLE presented with marked connectome changes in bilateral temporo-parietal regions, reflecting an increased segregation of the ipsilateral anterior temporal lobe from the rest of the brain. Surgery-induced connectome reorganization was localized to this temporo-parietal subnetwork, but primarily involved postoperative (C) The Author(s) 2024. Published by Oxford University Press on behalf of the Guarantors of Brain. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site-for further information please contact journals.permissions@oup.com.
手术前,TLE患者双侧颞顶区的连接组发生明显变化,反映出同侧颞叶前部与大脑其他部分的分离程度增加。手术诱导的连接组重组被定位在这个颞顶叶子网络中,但主要涉及术后(C) 作者 2024。由牛津大学出版社代表《大脑》担保人出版。这是一篇根据知识共享署名-非商业性许可 ( https://creativecommons.org/licenses/by-nc/4.0/) 条款发布的开放获取文章,该许可允许在任何媒体上进行非商业性的再使用、分发和复制,前提是适当引用了原作。如需商业再利用,请联系 reprints@oup.com 以获取重印和转载的翻译权。所有其他许可均可通过我们网站文章页面上的 "许可 "链接,通过我们的 RightsLink 服务获得,如需更多信息,请联系 journals.permissions@oup.com。

integration of contralateral regions with the rest of the brain. Using a partial least-squares analysis, we uncovered a latent clinical-imaging signature underlying this pre- to postoperative connectome reorganization, showing that patients who displayed postoperative integration in bilateral fronto-occipital cortices also had greater preoperative ipsilateral hippocampal atrophy, lower seizure frequency, and secondarily generalized seizures.
对侧区域与大脑其他区域的整合。通过偏最小二乘法分析,我们发现了从术前到术后连接组重组背后潜在的临床成像特征,显示术后双侧前枕叶皮质出现整合的患者术前同侧海马萎缩程度更大,癫痫发作频率更低,并出现二次泛化癫痫发作。
Our results bridge the effects of focal brain lesions and their surgical resections with largescale network reorganization and inter-individual clinical variability, thus offering new avenues to examine the fundamental malleability of the human brain.
我们的研究结果将局灶性脑损伤及其手术切除的影响与大规模网络重组和个体间临床变异性联系起来,从而为研究人脑的基本可塑性提供了新的途径。

Author affiliations: 作者单位:

1 Multimodal Imaging and Connectome Analysis Laboratory, MeConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, QC H3A 2B4, Canada
1 加拿大蒙特利尔 H3A 2B4 麦克吉尔大学蒙特利尔神经研究所和医院 MeConnell 脑成像中心多模态成像和连接组分析实验室
2 Center for Brain Circuit Therapeutics, Brigham and Women's Hospital, Harvard University, Boston, MA 02115, USA
2 美国马萨诸塞州波士顿市哈佛大学布里格姆妇女医院脑回路治疗中心 邮编:02115
3 Department of Data Science, Inha University, Incheon 22212, Republic of Korea
3 仁荷大学数据科学系,大韩民国仁川 22212
4 Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon 34126, Republic of Korea
4 基础科学研究所神经科学成像研究中心,大韩民国水原 34126
5 Department of Medical Imaging, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China
5 南京大学医学院附属金陵医院医学影像科,南京 210002
6 Analytical Neurophysiology Laboratory, Montreal Neurological Institute, McGill University, Montreal, QC H3A 2B4, Canada
6 加拿大蒙特利尔 H3A 2B4 麦克吉尔大学蒙特利尔神经研究所神经生理学分析实验室
7 Department of Radiology, Nanjing Drum Tower Hospital, Affiliated Hospital of Nanjing University Medical School, Nanjing 210008, China
7 南京大学医学院附属鼓楼医院放射科,南京 210008
8 Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA
8 美国宾夕法尼亚州费城 19104 宾夕法尼亚大学佩雷尔曼医学院精神病学系
9 McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, QC H3A 2B4, Canada
9 麦克吉尔大学蒙特利尔神经研究所和医院麦康奈尔脑成像中心,加拿大蒙特利尔,QC H3A 2B4
10 Neuroimaging of Epilepsy Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, QC H3A 2B4, Canada
10 加拿大蒙特利尔 H3A 2B4 麦克吉尔大学蒙特利尔神经研究所麦康奈尔脑成像中心癫痫神经成像实验室
Correspondence to: Boris C. Bernhardt, PhD
通讯作者:Boris C. Bernhardt 博士Boris C. Bernhardt 博士
Montreal Neurological Institute (NW-256)
蒙特利尔神经研究所(NW-256)
3801 University Street 大学街 3801 号
Montreal, Quebec, Canada H3A 2B4
加拿大魁北克省蒙特利尔 H3A 2B4
E-mail: boris.bernhardt@mcgill.ca
电子邮件: boris.bernhardt@mcgill.ca
Correspondence may also be addressed to: Sara Larivière,
来函也可寄往Sara Larivière、
Brigham and Women's Hospital
布里格姆妇女医院
60 Fenwood Rd
Boston, MA 02115, USA
美国马萨诸塞州波士顿 02115
E-mail: slariviere@bwh.harvard.edu
电子邮件: slariviere@bwh.harvard.edu
Running title: Surgery-induced connectome reorganization in TLE
运行标题:手术诱导的 TLE 连接组重组
Keywords: connectome; brain networks; focal lesion; surgery; longitudinal; gradients
关键词:连接组;大脑网络;病灶;手术;纵向;梯度

Introduction 导言

Human brain organization is increasingly conceptualized and analyzed from a network perspective, allowing great strides in understanding both health and disease. In recent years, these approaches have been guided by advances in neuroimaging acquisition and complex data analytics. In particular, the advent of diffusion MRI tractography has enabled the approximation of structural connections in vivo, and the systematic characterization of brain connectivity-so called connectomes. In parallel, a series of multivariate analytics has been developed to capture organizational principles of large-scale brain connectivity. These range from graph theoretical assessments of network topology to modular decompositions that identify interacting systems. More recently, dimensionality reduction methods that derive compact and continuous embeddings from high dimensional connectomes have begun to reveal salient spatial axes of connectivity. In the current work, we adopted the latter
人们越来越多地从网络角度对人脑组织进行概念化和分析,从而在了解健康和疾病方面取得了长足进步。近年来,神经成像采集和复杂数据分析技术的进步为这些方法提供了指导。尤其是弥散核磁共振成像束成像技术的出现,使得体内结构连接的近似性和大脑连接性(即所谓的connectomes)的系统表征成为可能。 与此同时,还开发了一系列多元分析方法来捕捉大规模大脑连接的组织原理。这些方法包括对网络拓扑的图论评估 ,以及识别相互作用系统的模块分解。 最近,从高维连接组中推导出紧凑连续嵌入的降维方法开始揭示连接性的显著空间轴。 在目前的工作中,我们采用了后者

approach to establish how structural connectivity embeddings can be modulated by focal lesions.
的方法来确定病灶病变如何调节结构连接嵌入。
We specifically study patients with pharmaco-resistant temporal lobe epilepsy (TLE), who suffer from seizures originating from the temporal lobe despite anti-seizure medical treatment. Compared to patients with adequate seizure control, pharmaco-resistant patients present with elevated risk for comorbidity and mortality, impaired quality of life and wellbeing, cognitive dysfunction and psychiatric difficulties, as well as progressive brain damage. In these patients, randomized controlled trials have shown that resection of the affected temporal lobe is currently the most effective treatment to control seizures, to restore psychosocial functioning, and to improve quality of life. By applying dimensional decomposition approaches to structural connectivity data acquired in healthy individuals as well as patients before and after surgery, we can investigate the impact of focal lesions on connectome-level organization.
我们专门研究了药物耐药性颞叶癫痫(TLE)患者,这些患者尽管接受了抗癫痫药物治疗,但仍会出现源于颞叶的癫痫发作。 与癫痫发作得到充分控制的患者相比,药物耐药性患者的合并症和死亡风险较高, ,生活质量和幸福感受损, ,认知功能障碍和精神障碍, ,以及进行性脑损伤。 随机对照试验表明,对于这些患者,切除受影响的颞叶是目前控制癫痫发作、恢复社会心理功能和改善生活质量的最有效治疗方法。 通过对健康人以及患者手术前后获得的结构连通性数据进行维度分解,我们可以研究病灶病变对连通组水平组织的影响。
Our novel framework, thereby, complements previous investigations of structural connectivity alterations in TLE patients relative to controls before surgery, which have reported microstructural and architectural damage seldomly limited to the temporal lobe or to the hemisphere ipsilateral to the seizure focus. Similarly, prior work has assessed topological changes based on graph theoretical parameterization, and has accumulated evidence for a modular rearrangement in networks that extends beyond the temporal lobe. Such findings are compatible with the notion that "focal" pathologies generally affect large-scale brain connectivity, and that large-scale compromise contributes to patient presentation and clinically relevant outcomes. In line with these findings, studies assessing brain network changes following epilepsy surgery suggest that a focal resection does not only jimpact the primary lesioned site, but rather has effects that cascade along direct and indirect connections into mutually interconnected networks. As such, TLE represents a powerful human lesion model to assess how focal pathology and its resection affect wholebrain organization.
因此,我们的新框架补充了以往对TLE患者与对照组手术前结构连接性改变的研究,这些研究报告的微观结构和架构损伤很少局限于颞叶或癫痫病灶同侧半球。 同样,先前的工作也根据图论参数化评估了拓扑变化,并积累了网络模块化重新排列的证据,其范围超出了颞叶。 这些发现与 "局灶性 "病变通常会影响大脑大范围连接性,而大范围连接性受损会导致患者表现和临床相关结果的观点相吻合。 与这些发现相一致的是,评估癫痫手术后大脑网络变化的研究表明,病灶切除术不仅会影响原发病变部位,而且会沿着直接和间接连接串联成相互关联的网络。 因此,TLE 是一个强大的人类病变模型,可用于评估病灶病理及其切除如何影响整个脑组织。
Here, we leveraged pre- as well as postoperative diffusion MRI data acquired in a cohort of pharmaco-resistant TLE patients. Our analytical paradigm derived whole brain connectomes from diffusion tractographic modelling, followed by non-linear dimensionality reduction to derive a continuous coordinate system in which confluent spatial trends of structural connectivity variation can be investigated. We compared our patients to a group of healthy controls, to investigate preoperative alterations and we tracked connectome reorganization
在这里,我们利用了一组药物耐受性TLE患者术前和术后的弥散核磁共振成像数据。我们的分析范式从弥散牵引建模中得出全脑连接组,然后进行非线性降维,得出一个连续坐标系,在该坐标系中可以研究结构连接变化的汇合空间趋势。 我们将患者与一组健康对照组进行了比较,以调查术前改变,并跟踪了连接组的重组情况。

from pre-to postoperative time points in patients. Exploiting inter-patient heterogeneity, we furthermore examined associations of pre- to postoperative connectivity trends with clinical and other imaging phenotypes at the individual patient level.
从患者术前到术后的时间点。利用患者间的异质性,我们进一步研究了术前到术后的连通性趋势与单个患者的临床和其他影像表型之间的关联。

Materials and methods 材料和方法

Participants 与会者

We studied 37 consecutive people suffering from pharmaco-resistant TLE, who (i) underwent anterior temporal lobectomy as a treatment of their seizures at Jinling Hospitar between 2009 and 2018, (ii) had postoperative histological confirmation of hippocampal sclerosis, (iii) underwent a research-dedicated, high-resolution 3T MRI before and after surgery that included T1-weighted (T1w) and diffusion MRI scans, and (iv) had at least one year of postoperative follow-up. Patients were diagnosed according to the classification of the International League Against Epilepsy based on a comprehensive examination that includes clinical history, seizure semiology, continuous video-electroencephalographic (EEG) telemetry recordings, neuroimaging, and neuropsychology. No patient had encephalitis, malformations of cortical development (e.g., tumors, vascular malformations), or a history of traumatic brain injury. We also studied 31 age- and sex-matched healthy individuals who underwent identical 3T MRI at one time point. Demographic and clinical information on both cohorts is provided in Supplementary Table 1. Our study was approved by the research ethics board of Jinling Hospital, Nanjing University School of Medicine, and written informed consent was obtained from all participants.
我们连续研究了37名药物耐药性TLE患者,他们(i)在2009年至2018年期间在金陵医院接受了前颞叶切除术作为癫痫发作的治疗方法,(ii)术后组织学证实海马硬化、(iii) 手术前后均接受了研究专用的高分辨率 3T 磁共振成像,包括 T1 加权(T1w)和弥散磁共振成像扫描,以及 (iv) 术后随访至少一年。根据国际抗癫痫联盟(International League Against Epilepsy)的分类,对患者进行了全面检查,包括临床病史、癫痫发作半定量分析、连续视频脑电图(EEG)遥测记录、神经影像学检查和神经心理学检查。没有患者患有脑炎、大脑皮层发育畸形(如肿瘤、血管畸形)或脑外伤史。我们还研究了 31 名年龄和性别匹配的健康人,他们在一个时间点接受了相同的 3T MRI 检查。两组患者的人口统计学和临床信息见补充表 1。我们的研究获得了南京大学医学院附属金陵医院研究伦理委员会的批准,并获得了所有参与者的书面知情同意。

MRI acquisition and processing
磁共振成像采集和处理

All MRI data were obtained on a Siemens Trio 3T scanner (Siemens, Erlangen, Germany) and included: (i) high-resolution 3D T1w MRI using a magnetization-prepared rapid gradient-echo sequence (repetition time , echo time , flip angle , voxel size , field of view slices) and (ii) diffusion MRI using a spin echo-based echo planar imaging sequence with four b0 images (dMRI, TR , flip angle , voxel size , b-value , diffusion directions .
所有磁共振成像数据均在西门子 Trio 3T 扫描仪(西门子,德国埃尔兰根)上获得,包括(i) 使用磁化预处理快速梯度回波序列进行高分辨率三维 T1w MRI(重复时间 ,回波时间 ,翻转角 ,体素大小 ,视场 切片);(ii) 使用基于自旋回波的回波平面成像序列进行弥散 MRI,包含四幅 b0 图像(dMRI,TR ,翻转角 ,体素大小 ,b 值 ,弥散方向
Multimodal MRI preprocessing was performed using micapipe, an open-access image processing and data fusion pipeline ( ; https://micapipe.readthedocs.io).. In brief, native T1w structural images were deobliqued, reoriented to standard neuroscience orientation, corrected for intensity nonuniformity, intensity normalized, skull-stripped, and submitted to FreeSurfer (v6.0) to extract models of the cortical surfaces. Subcortical structures were segmented using FSL FIRST. Native DWI images were denoised, underwent b0 intensity normalization, and were corrected for susceptibility distortion, head motion, and edfy currents.
多模态磁共振成像预处理是使用开放式图像处理和数据融合管道 micapipe ( ; https://micapipe.readthedocs.io)进行的。 简而言之,原始 T1w 结构图像经过去漂白、重新定向到标准神经科学方向、校正强度不均匀性、强度归一化、颅骨切片,并提交到 FreeSurfer (v6.0) 提取皮质表面模型。使用 FSL FIRST 对皮层下结构进行分割。 原始 DWI 图像经过去噪、b0 强度归一化处理,并对易感失真、头部运动和刃电流进行了校正。

Surgical cavity mapping 手术洞穴绘图

Patient-specific surgical cavities were automatically segmented by registering both pre- and postoperative T1w images to the MNI152 standard template through linear transformations and subtracting the postoperative scan from the preoperative scan. Segmented cavities were visually inspected and manually edited to ensure that the extent of the resections was correctly identified. Cavities were then mapped to the surface template, and a consensus label was generated as defined by the union of all segmentations.
通过线性变换将术前和术后的 T1w 图像注册到 MNI152 标准模板上,并从术前扫描中减去术后扫描,从而自动分割患者特定的手术腔。对分割后的腔隙进行目视检查和手动编辑,以确保正确识别切除范围。然后将腔隙映射到表面模板上,并根据所有分割结果的组合生成共识标签。

Structural connectome generation
结构连接组的生成

Structural connectomes were generated with MRtrix3 from pre-processed DWI data. Examples of the images used in our image processing pipeline are illustrated in Fig. 1A. We performed anatomically-constrained tractography using tissue types (cortical and subcortical grey matter, white matter, cerebrospinal fluid) segmented from each participant's preprocessed T1w images registered to native DWI space. We estimated multi-tissue response functions, performed constrained spherical-deconvolution and intensity normalization. Seeding from all white matter voxels, we generated a tractogram with streamlines (maximum tract length ; fractional anisotropy cut-off ) and applied spherical deconvolution informed filtering of tractograms (SIFT2) to reconstruct whole-brain streamlines weighted by cross-sectional multipliers. Reconstructed streamlines were mapped onto 400 similar-sized cortical parcels constrained within the boundaries of the DesikanKilliany atlas and 14 regions in the subcortex and the hippocampus to produce subjectspecific structural connectivity matrices. Group-average normative structural connectomes were defined using a distance-dependent thresholding procedure, which preserved the edge length distribution in individual patients, and were transformed to reduce connectivity strength variance. As such, structural connectivity was defined by the number of streamlines
使用 MRtrix3 根据预处理后的 DWI 数据生成结构连接组。 图 1A 展示了我们的图像处理流程中使用的图像示例。我们使用从每位受试者的预处理 T1w 图像中分割出的组织类型(皮质和皮质下灰质、白质、脑脊液),并将其注册到原始 DWI 空间,进行了解剖学约束的束成像。我们估算了多组织响应函数,进行了约束球形解卷积和强度归一化。从所有白质体素开始,我们生成了带有 流线的束图(最大束长 ;分数各向异性截断 ),并应用球形解卷积对束图进行知情滤波(SIFT2),以重建按横截面乘数加权的全脑流线。将重建的流线映射到 400 个类似大小的皮层区块上,这些区块限制在 DesikanKilliany 地图集 以及皮层下和海马的 14 个区域的边界内,以生成特定受试者的结构连接矩阵。使用距离相关阈值化程序定义了组平均标准结构连接矩阵,该程序保留了单个患者的边缘长度分布, ,并对 ,以减少连接强度方差。因此,结构连通性由流线的数量定义

between two regions (i.e., fiber density). Structural connectomes were thresholded to retain only the top of connections.
即纤维密度)。对结构连接组进行了阈值化处理,只保留顶部的连接

Connectome gradient estimation
连接组梯度估计

Cortex- and subcortex-wide structural connectome gradients were generated using BrainSpace (v0.1.10; https://brainspace.readthedocs.io). First, a group-level structural connectome (combining cortical and subcortical data) was obtained using an iterative, leaveone-out procedure (Fig. 1B). For both the group-level and the left-out subject-specific connectome, an affinity matrix was constructed with a normalized angle kernel, and eigenvectors were estimated via diffusion map embedding, an unsuperyised nonlinear dimensionality reduction technique that projects connectome features into low-dimensional gradients. This technique is only controlled by only féw parameters, computationally efficient, relatively robust to noise compared to other nonlinear techniques, and has been extensively used in the previous literature. Algorithm parameters were identical to those from prior applications, specifically and diffusion time to retain the global relations between data points in the embedded space and to emphasize direct connections. Interhemispheric connections were not included in the gradient computation; left and right gradients were generated separately and aligned with Procrustes. The eigenvectors estimated from this technique provide a connectivity coordinate system - a diffusion mapwhere Euclidean distances in the manifold correspond to diffusion times between the nodes of the network. In this gradient space, regions with similar connectivity profiles are closely located, whereas regions with different connectivity profiles are located farther apart. To analyze gradient changes within and across participants, each left-out subject-specific gradient was aligned to the group template gradient (estimated using data from all other participants) yia Procrustes alignment and scaled between 0 and 1 . To improve sensitivity, patients were sorted into ipsilateral/contralateral to the seizure focus. The scree plot describing the eigenvalue decay function revealed that four principal eigenvectors (i.e., gradients) corresponded to the cut-off point, after which eigenvalues leveled off at small values and only explained marginal information.
使用 BrainSpace(v0.1.10;https://brainspace.readthedocs.io)生成皮层和皮层下结构连接组梯度。 首先,使用迭代、剔除程序(图 1B)获得组级结构连接组(结合皮层和皮层下数据)。对于组级和剔除的特定受试者连接组,使用归一化角度核构建亲和矩阵,并通过扩散图嵌入估算特征向量,扩散图嵌入是一种非上乘的非线性降维技术,可将连接组特征投射到低维梯度中。 该技术仅受féw参数控制,计算效率高,与其他非线性技术相比对噪声具有相对的鲁棒性, ,并已在之前的文献中得到广泛应用。 算法参数与之前的应用相同,特别是 和扩散时间 ,以保留嵌入空间中数据点之间的全局关系,并强调直接连接。 大脑半球间的连接不包括在梯度计算中;左右梯度分别生成,并用 Procrustes 对齐。 这项技术估算出的特征向量提供了一个连接坐标系--一个扩散图,流形中的欧氏距离对应于网络节点之间的扩散时间。 在这个梯度空间中,具有相似连通性特征的区域距离较近,而具有不同连通性特征的区域距离较远。为了分析参与者内部和参与者之间的梯度变化,每个被排除在外的特定受试者的梯度都与组模板梯度(使用所有其他参与者的数据估算)进行了对齐,yia Procrustes alignment ,并在 0 和 1 之间进行缩放。为了提高灵敏度,将患者分为癫痫病灶的同侧/外侧。描述特征值衰减函数的scree图显示,四个主要特征向量(即梯度)与临界点相对应,之后特征值在小值时趋于平稳,只能解释边缘信息。

Quantifying pre- and postoperative changes in structural
量化手术前后的结构变化

gradients
We assessed cross-sectional connectome changes between individuals with TLE (preoperatively) and controls using univariate fixed effects models performed on each eigenvector independently. Eigenvector-specific score histograms were then generated across the entire cortex, and split into ipsilateral/contralateral as well as patients/controls to examine changes in the distribution of significant scores and to assess inter-hemispheric asymmetry. To signify an overall load of alterations, we also compared the aggregate of the first four eigenvectors in patients relative to controls using a multivariate fixéd effect linear model. By statistically combining multidimensional gradients, the latter approach leverages their covariance to obtain a substantial gain in sensitivity, and thus unveil subthreshold properties not readily identified in a single eigenvector. Both uni- and multivariate linear models tested for differences between groups while controlling for effects of age and sex using BrainStat ( https://brainstat.readthedocs.io). 49
我们使用对每个特征向量独立执行的单变量固定效应模型,评估了TLE患者(术前)和对照组之间的横断面连接组变化。 然后在整个皮层生成特征向量特定得分直方图,并将其分为同侧/对侧以及患者/对照组,以检查重要得分分布的变化,并评估半球间的不对称性。为了显示整体负荷的变化,我们还使用多变量固定效应线性模型比较了患者与对照组前四个特征向量的总和。通过对多维梯度进行统计组合,后一种方法利用了它们的协方差,大大提高了灵敏度,从而揭示了单一特征向量不易识别的亚阈值特性。单变量和多变量线性模型都测试了组间差异,同时使用 BrainStat ( https://brainstat.readthedocs.io) 控制年龄和性别的影响。49
To test for changes over time (i.e., pre- vs. postsurgery), we fitted univariate and multivariate linear mixed effects models, a flexible framework for the analysis of repeated measures. Our models controlled for effects of age and sex, and included a subject-specific random intercept to improve model fit in longitudinal designs. With this approach, we were able to test for pre- to postoperative within-subject individual and multivariate eigenvector changes.
为了检验随时间的变化(即手术前与手术后),我们建立了单变量和多变量线性混合效应模型,这是一种灵活的重复测量分析框架。 我们的模型控制了年龄和性别的影响,并加入了特定受试者的随机截距,以改善纵向设计中的模型拟合度。通过这种方法,我们能够检验从术前到术后受试者内部个体和多变量特征向量的变化。
Findings from all fixed and mixed effects models were corrected for multiple comparisons using the false discovery rate (FDR) procedure.
所有固定效应和混合效应模型的研究结果都使用误发现率(FDR)程序进行了多重比较校正。

4D gradient deformations
4D 梯度变形

To simplify multidimensional changes into a single scalar feature, we computed the Euclidean distance (also referred to as 'eccentricity') between the group template center and eaçh subject-specific data point in the 4D gradient space (Fig. 2B). Previous work from our group and others has shown an inverse relationship between a region's eccentricity and its level of integration across different networks. For instance, highly eccentric regions (i.e., those distant from the 4D group template center) are interpreted to be segregated from other networks in the brain, but tightly interconnected with regions of the same network. On the other hand, regions with lower eccentricity (i.e., those with close proximity to the template center) are thought to reflect greater covariance with the rest of the brain, and consequently,
为了将多维变化简化为单个标量特征,我们计算了组模板中心与 4D 梯度空间中特定受试者数据点之间的欧氏距离(也称为 "偏心率")(图 2B)。我们的研究小组 和其他研究小组 以前的研究表明,一个区域的偏心率与其在不同网络中的整合程度成反比关系。例如,高偏心率区域(即远离 4D 组模板中心的区域)被解释为与大脑中的其他网络隔离,但与同一网络的区域紧密相连。 另一方面,偏心率较低的区域(即与模板中心距离较近的区域)被认为反映了与大脑其他区域的更大共变性,从而反映了与大脑其他区域的更大共变性、

be more structurally and functionally integrated. Collectively, these findings support the notion that changes in a region's segregation or integration can be assessed through variations in manifold eccentricity. In turn, we interpreted eccentricity reductions as connectome integration and eccentricity increases as connectome segregation. The template center was defined as the centroid of the first four eigenvectors. Shifts in structural connectivity patterns lead to a displacement of eigenvector scores in the 4D gradient space, which in turn affects their proximity to the center. Gradient contraction/expansion thus quantifies global brain reorganization in connectivity space. Gradient deformation was quantified only in regions of interest, as determined by areas that showed significant cross-sectional and longitudinal changes, respectively. The rationale was to determine whether preoperative and surgeryinduced 4D gradient deformations were observed in areas that showed significant eigenvector changes.
在结构和功能上更加一体化。总之,这些发现支持了这样一种观点,即一个区域的分离或整合的变化可以通过歧面偏心率的变化来评估。反过来,我们将偏心率降低解释为连接组整合,将偏心率增加解释为连接组分离。模板中心被定义为前四个特征向量的中心点。结构连接模式的变化会导致特征向量得分在四维梯度空间中发生位移,进而影响其与中心的距离。因此,梯度收缩/扩张可以量化连接空间中的全球大脑重组。梯度变形仅在感兴趣的区域进行量化,这些区域由分别显示出显著横截面和纵向变化的区域决定。这样做的目的是为了确定在出现显著特征向量变化的区域是否观察到术前和手术诱导的四维梯度变形。

Inclusion and exclusion of surgical cavities
纳入和排除手术空腔

To ensure a rigorous and balanced comparison between preoperative TLE patients and controls, we included all brain regions in the construction of subject-specific structural connectome gradients. This approach was adopted to preserve the integrity of the whole-brain analysis and facilitate accurate gradient construction. However, to maintain consistency in the number of brain regions, and facilitate a balanced statistical comparison between preoperative and postoperative time points, we subsequently excluded surgical cavities from the analysis. In preoperative patients, the surgical cavities were excluded after constructing the wholebrain gradients to avoid biasing the gradient construction process. For postoperative patients, the surgical cavities were excluded at an earlier step, specifically during the construction of the connectivity matrix by omitting the corresponding rows and columns.
为确保对术前TLE患者和对照组进行严格、平衡的比较,我们在构建特定受试者结构连接组梯度时纳入了所有脑区。采用这种方法是为了保持全脑分析的完整性,便于准确构建梯度。不过,为了保持脑区数量的一致性,并便于对术前和术后时间点进行均衡的统计比较,我们随后将手术腔排除在分析之外。对于术前患者,我们在构建全脑梯度后排除了手术腔,以避免梯度构建过程出现偏差。对于术后患者,我们在更早的步骤中排除了手术腔,特别是在构建连接矩阵时省略了相应的行和列。
To understand the effects of cavity removal on gradient alterations and ensure that reported findings reflect bona fide alterations in connectome reorganization post-surgery, we performed a twofold evaluation process. First, we excluded the cavity at two different stages in the generation of whole-brain gradients in controls and preoperative patients: (i) during the construction of the connectivity matrix and (ii) after constructing the whole-brain gradients. Using spatial correlation analyses, we then evaluated the consistency of gradient patterns obtained from both approaches at an individual level. Second, to assess robustness of our preto postoperative findings, we repeated our multivariate analysis with surgical cavities being
为了了解龋洞去除对梯度改变的影响,并确保报告的结果反映了手术后连接组重组的真正改变,我们进行了双重评估。首先,我们在对照组和术前患者生成全脑梯度的两个不同阶段排除了空腔:(i)在构建连接矩阵时;(ii)在构建全脑梯度后。然后,我们利用空间相关性分析评估了两种方法在个体水平上获得的梯度模式的一致性。其次,为了评估术前和术后研究结果的稳健性,我们重复了我们的多变量分析,将手术腔室作为术前和术后研究对象。

excluded during the connectivity matrix construction at both preoperative and postoperative time points.
术前和术后时间点的连接矩阵构建过程中都排除了这些数据。

Spatial associations between preoperative and surgery-induced reorganization
术前重组与手术诱导重组之间的空间关联

To assess whether preoperative alterations constrained surgery-induced changes, we performed spatial associations between cross-sectional (relative to controls) and longitudinal (pre- to postoperative) connectome findings. Statistical significance of spatial correlations was assessed using variogram-matching models that take into account the spatial dependencies in the data. This method generates surrogate brain maps with matched spatial autocorrelation to that of a target brain, and has previously been applied to cortical and subcortical structures. As recommended, surrogate maps were generated using surfacebased geodesic distance between cortical regions and three-dimensional Euclidean distance between subcortical and cortical/subcortical regions. Variogram-matching null distributions were then generated from randomly shuffling surrogate maps while preserving the distancedependent correlation between elements of the brain map. The empirical correlation coefficients were compared against the null distribution determined by the ensemble of spatially permuted correlation coefficients tó estimate a -value (termed ).
为了评估术前改变是否制约了手术引起的改变,我们在横断面(相对于对照组)和纵向(术前到术后)连接组结果之间建立了空间关联。空间相关性的统计意义使用变异图匹配模型进行评估,该模型考虑了数据的空间依赖性。 这种方法生成的代用脑图与目标脑图具有匹配的空间自相关性,以前曾应用于皮层和皮层下结构。根据建议, ,利用皮层区域之间基于表面的大地测量距离以及皮层下和皮层/皮层下区域之间的三维欧氏距离生成代脑图。然后,在保留脑图元素之间与距离有关的相关性的前提下,通过随机洗牌代用图生成变异图匹配空分布。将经验相关系数与空间包络相关系数集合确定的空分布进行比较,以估算 -值(称为 )。

Clinical associations 临床关联

We used partial least squares (PLS) analysis to unravel within-patient relationships between surgery-induced 4D gradient deformations and standard neuroimaging (i.e., hippocampal atrophy ) and clinicál parameters (Fig. 4A). PLS analysis is a multivariate associative technique that maximizes the covariance between two sets of variables. Briefly, these two variable sets are correlated with each other across patients. The resulting correlation matrix is submitted to singular value decomposition, which identifies linear combinations of the original variables to generate new latent variables that have maximum covariance. We evaluated the significance of our model using nonparametric methods: (i) permutation tests were used to assess statistical significance of each latent variable ( 10,000 permutations; termed ), (ii) bootstrap ratios were used to assess the reliability of singular vector weights (analogous to -scores, such that confidence interval corresponds to a bootstrap ratio of and , respectively), and (iii) cross-validation was used to assess out-ofsample correlations between projected scores using a four-fold approach (100 randomized
我们使用偏最小二乘法(PLS)分析来揭示手术诱导的 4D 梯度变形与标准神经影像学(即海马萎缩 )和临床参数之间的患者内部关系(图 4A)。PLS 分析是一种多元关联技术,可最大化两组变量之间的协方差。 简而言之,这两组变量在不同患者之间相互关联。由此产生的相关矩阵经过奇异值分解,找出原始变量的线性组合,生成具有最大协方差的新潜变量。 我们使用非参数方法评估了模型的显著性:(i) 使用置换检验来评估每个潜变量的统计意义(10,000 次置换;称为 ),(ii) 使用自举比率来评估奇异向量权重的可靠性(类似于 分数,因此 置信区间分别对应于 的自举比率),(iii) 使用交叉验证来评估预测分数之间的样本外相关性。

train-test splits of the original data, where of the data was treated as a training set and of the data was treated as an out-of-sample test set) and a leave-one-out approach. Mathematical details of the analysis and inferential methods are described elsewhere.
其中 数据被视为训练集,而 数据被视为样本外测试集)和 "留一弃一 "方法。分析和推论方法的数学细节在其他地方有详细描述。

Cortical morphology and microstructure
皮质形态和微观结构

It has been shown that grey and white matter is significantly altered following resective surgery in TLE. To improve our understanding of surgery-induced gradient changes at the morphological and microstructural levels, we aligned our longitudinal multivariate gradient analyses with corresponding alterations in grey and white matter measures. For grey matter assessments, subject-specific cortical thickness was estimated using FreeSurfer and was measured as the Euclidean distance between corresponding pial and white matter vertices. For white matter assessments, we studied the superficial white matter located beneath the cortex. This compartment harbors termination zones of long-range tracts and short-range U-fibers and is, therefore, pivotal in maintaining cortico-cortical connectivity. We used surface-based linear models to compare grey (cortical thickness) and white (multivariate measures of fractional anisotropy and mean diffusivity) matter changes before and after surgery, restricting our analysis to regions showing significant longitudinal gradient alterations.
研究表明,TLE患者在接受切除手术后,灰质和白质会发生显著改变。 为了更好地了解手术在形态学和微观结构层面引起的梯度变化,我们将纵向多变量梯度分析与灰质和白质测量的相应变化进行了比对。在灰质评估中,使用 FreeSurfer 估算受试者特定的皮质厚度,并以相应的髓质和白质顶点之间的欧氏距离进行测量。对于白质评估,我们研究了位于皮层下方 的表层白质。 该区域是长程束和短程U型纤维的终止区,因此在维持皮层与皮层之间的连接方面起着关键作用。 我们使用基于表面的线性模型来比较手术前后灰质(皮质厚度)和白质(分数各向异性和平均扩散率的多变量测量)的变化,并将我们的分析局限于显示出显著纵向梯度变化的区域。

Results 成果

Connectome gradient analysis at baseline
基线连接组梯度分析

For every participant, we generated cortex- and subcortex-wide structural connectome gradients using non-linear dimensionality reduction. The first four gradients explained approximately (baseline/preoperatively) and (postoperatively) of variance across participants, with each gradient representing a different axis of connectivity variation (Fig. 1). Across both time points and in both groups, gradients generally depicted smooth differentiation in structural connectivity profiles of frontal vs temporo-occipital regions (Gradient 1), orbitofrontal vs sensorimotor regions (Gradient 2), medial occipital vs anterior temporal lobe regions (Gradient 3), and sensory vs. transmodal regions (Gradient 4).
我们使用非线性降维方法为每位参与者生成了皮层和皮层下范围的结构连接组梯度。 前四个梯度分别解释了大约 (基线/术前)和 (术后)参与者的方差,每个梯度代表不同的连接变异轴(图 1)。在两个时间点和两个组别中,梯度总体上描述了额叶与颞枕叶区域(梯度 1)、眶额叶与感觉运动区域(梯度 2)、内侧枕叶与前颞叶区域(梯度 3)以及感觉与跨模态区域(梯度 4)的结构连接特征的平滑分化。
Compared to healthy controls, presurgical TLE patients showed significant gradient changes in bilateral temporo-parietal ( ) and orbitofrontal ( ) cortices, together with the ipsilateral hippocampus ( ) and the contralateral amygdala ; Fig. 2A). Considering each gradient separately, global histogram analysis revealed significant asymmetry between ipsilateral and contralateral gradient alterations in TLE. Most notably, ipsilateral mesiotemporal lobe regions were shifted toward the most extreme, segregated positions along Gradients 2, 3, and 4, whereas their contralateral homologues were shifted to a more central, integrated position (Supplementary Fig. 1). This ipsilateral mesiotemporal lobe segregation was further confirmed by quantifying their positional change in the 4D gradient space relative to controls. Constraining this analysis to areas of significant preoperative gradient changes, individuals with TLE showed highest deformation values (i.e., depicting increased expansion/segregation) in ipsilateral anterior temporal lobe ( 0.005 ) and parahippocampal gyrus ( ; Fig. 2B).
与健康对照组相比,手术前的 TLE 患者双侧颞顶叶( )和眶额叶( )皮层,以及同侧海马( )和对侧杏仁核( ;图 2A)均显示出明显的梯度变化。分别考虑每个梯度,全局直方图分析表明,在 TLE 中,同侧和对侧梯度的改变明显不对称。最值得注意的是,同侧中颞叶区域沿梯度 2、3 和 4 向最极端、最分离的位置移动,而其对侧同源区域则向更中心、更整合的位置移动(补充图 1)。通过量化它们在四维梯度空间中相对于对照组的位置变化,进一步证实了这种同侧颞中叶分离现象。将这一分析限制在术前梯度变化显著的区域,TLE 患者同侧颞叶前部 ( 0.005 ) 和海马旁回 ( ; 图 2B) 的变形值最高(即显示扩展/分离增加)。

Tracking surgery-induced deformations
跟踪手术引起的变形

Directly comparing structural connectome changes before and after anterior temporal lobectomy (Fig. 3A), patients showed significant multivariate gradient changes in posterior temporal regions near the resection site and contralateral temporo-parietal cortices ( ; Fig. 3B). As evidenced in both univariate and 4D gradient deformation analyses, contralateral temporo-parietal regions were shifted towards more integrated positions in the 4D gradient space after surgery, particularly along the axis of Gradient 1 ( ; Supplementary Fig. 2). Findings remained significant after controlling for time after surgery (spatial correlation with main multivariate findings: ; Supplementary Fig. 3). Similarly, comparing the effects of removing the surgical cavity at different stages in our analysis pipeline revealed minimal impact on baseline/preoperative connectome gradients (all ; Supplementary Fig. 4A) and yielded similar pre- to postoperative gradient findings ; Supplementary Fig. 4B).
直接比较颞叶前部切除术前后的结构连接组变化(图 3A),患者在切除部位附近的颞后部区域 和对侧颞叶皮质 ( ; 图 3B)显示出显著的多变量梯度变化。单变量和四维梯度变形分析表明,术后对侧颞顶叶区域向四维梯度空间中更整合的位置移动,尤其是沿梯度 1 轴 ( ; 补充图 2)。在控制术后时间后,研究结果仍有意义(与主要多变量研究结果的空间相关性: ;附图 3)。同样,在我们的分析管道中比较不同阶段切除手术腔的影响,发现对基线/术前连通组梯度的影响极小(所有 ;补充图 4A),术前与术后梯度结果相似 ;补充图 4B)。
As all regions showing pre- to postoperative gradient changes fell within the boundaries of significant preoperative changes, we next assessed the extent to which preoperative connectome reorganization relates to surgery-induced changes. The degree of preoperative multivariate gradient changes was correlated with pre- to postoperative multivariate gradient changes across all regions ( ), as well as when considering only areas of significant preoperative ( ) and pre- to postoperative
由于所有显示术前到术后梯度变化的区域都在显著术前变化的范围内,我们接下来评估了术前连接组重组与手术引起的变化的相关程度。术前多变量梯度变化的程度与术前至术后所有区域的多变量梯度变化相关( ),当仅考虑术前显著变化的区域( )和术前至术后的梯度变化时也是如此。
0.05) gradient changes (Supplementary Fig. 3). In other words, regions with profound connectome alterations before surgery (relative to controls) corresponded to regions that also underwent the largest pre- to postoperative reorganization. We also found a significant, albeit weaker, negative spatial associations between whole-brain preoperative and surgery-induced 4D gradient deformations ( ). In this case, a negative correlation indicates that regions that were initially segregated from the rest of the brain before surgery subsequently shifted towards a more integrated position in the connectome space after surgery.
0.05)的梯度变化(补充图 3)。换句话说,手术前连接组发生严重改变的区域(相对于对照组)对应的区域也经历了最大的术前至术后重组。我们还发现,全脑术前和手术诱导的四维梯度变形之间存在明显的负空间关联( ),尽管这种关联较弱。在这种情况下,负相关表明,术前最初与大脑其他部分分离的区域在术后向连接组空间中更一体化的位置转移。

Associations with clinical variables
与临床变量的关联

Multivariate PLS analysis identified one statistically significant latent variable (LV-1 accounting for of covariance, ) that maximized the covariance between whole-brain 4D surgery-induced deformation patterns and standard neuroimaging parameters (e.g., hippocampal atrophy) and clinical measures (Fig. 4A). Loadings (i.e., correlations of individual clinical measures with LV-1) revealed ipsilateral hippocampal atrophy in Cornu Ammonis (CA1-3) and dentate gyrus subfields (all ), lower seizure frequency ( , and presence of secondarily generalized seizures as the strongest contributors of LV-1. Patients presenting with these clinical features showed increased postoperative connectome integration in bilateral fronto-occipital cortices, ipsilateral postcentral gyrus, and contralateral mesiotemporal cortex (all bootstrap ratio < confidence interval). Clinical measures and 4D surgery-induced gradient deformations were correlated at the individual patient level ( ), while out-of-sample correlations averaged (four-fold cross-validation) and (leave-one-out cross-validation), both of which exceeded mean permuted out-of-sample correlations (four-fould: , leaveone-out: ; Fig. 4B).
多变量 PLS 分析确定了一个具有统计意义的潜在变量(LV-1,占协方差的