原文链接:https://authors.elsevier.com/sd/article/S0960-9822(22)00734-5
图1. 基于电刺激结合内源信号光学成像得到的皮层内功能拓扑连接。依次刺激皮层上相邻的位点(A),对应的连接位点也会随着刺激位点的移动而发生位移(B-D)。
图2. 常见的柱状结构微网络。刺激(黄色闪电)猕猴V2的单个功能柱(黑点)激活的微网络。颜色功能柱相关网络:红色箭头。朝向功能柱相关网络:蓝色箭头。插图:猕猴大脑。
Self-portrait painted by Famous Dutch painter Rembrandt (left). The details on his face reflects his personal tragedy in his later years. Smoothing of fine features in his face leads to dramatic loss of the evidence of these tragic events (right). (https://en.wikipedia.org/wiki/Self-Portrait_with_Beret_and_Turned-Up_Collar#cite_note-A1-1). This underscores the importance of fine detail.
One of the amazing things developed in the last century is the ability to non-invasively peer into the human brain using a technology called Magnetic Resonance Imaging (MRI). This imaging method has provided us a breathtaking universe of images depicting the brain’s activity during sensory, motor, emotional, and cognitive behavior. Research in nonhuman primates, an excellent model of human brain function, has revealed that the brain is functionally organized at fine (submillimeter) scale. The use of ultra-high field (≥ 7 T) magnetic resonance (UHF-MRI) technology, combined with advanced RF coils and pulse sequences, has improved the resolution of MRI images by 3-50 folds, making it feasible to image these submillimeter structures. However, mapping these fine functional structures has remained quite challenging.
One of the reasons it is challenging is that the hard-won improvements in resolution are often lost due to inadvertent data blurring caused by data processing procedures. Normally, fMRI data undergo multiple standard steps of data processing, termed a ‘pipeline’. These processing steps have been used for many functional MRI studies and have proven robust. However, for extracting the finest details of the images, this study shows they actually introduce blurring that is counterproductive. The result is inability to resolve functionally distinct structures in the brain (e.g., inability to distinguish color vs disparity processing locations). In addition, as highlighted by a recent report in Nature (Botvinik-Nezer et al., 2020), such implicit smoothing is a primary reason for inconsistent results in fMRI studies.
In this study, the authors, for the first time, systematically and quantitatively examined how fMRI data processing strategies influence the detection of fine-scale features acquired in ultrahigh field 7T MRI. To answer this question, the research team proposed both quantitative and qualitative methods and evaluated the effect of different data processing strategies on spatial blur. Using synthesized data (white noise, voxel index etc.) and data from human vision experiments (spatial interleaved color and disparity stripes), they found that voxel upsampling, surface mesh refinement and intracortical smoothing can effectively retain resolution. For example, surface mesh refinement is key to rescuing voxels that would have been lost with lower surface resolution (Fig. 1A, 1B); the greater the refinement, the better the retention of valuable voxels (FIG. 1C). Furthermore, the researchers used a well-known organization in the brain as a benchmark of spatial blur. In the second visual cortical area V2 of the brain, the millimeter-sized ‘stripes’ represent visual feature information related to color and depth. These methods permit visualization of voxels not seen in with default processing (Fig. 2B, 2C, arrows) and results in more distinct color and depth stripes (Fig. 2D). In fact, the simple voxel upsampling method works as well as the combined image transformation method, and is also beneficial for improving the accuracy of head motion estimation. The results also point out the inhomogeneity of effective resolution across the brain, something fundamental for the interpretation of experimental results.
Figure 1. Evaluation of voxel loss during cortical representation using synthetic voxel index data. (A) Schematic diagram of the projection of voxels to surface grids. When the spacing between adjacent voxels of high-resolution data is less than that between adjacent vertices of surface mesh, some voxels are lost (gray squares). Voxel loss can be reduced by adding surface mesh vertices (purple circles). (B) The lost voxels of the original cortex can be retained as the surface mesh refined. A zoom in view shows the surface (black outline) inside the calcarine sulcus. The color represents the voxel index value. As the cortical vertices increase, more voxels are projected into the surface. (C) Quantitative evaluation of the surface refinement effect. The number of fMRI voxels increased with the increase of cortical mesh vertices. Each refinement iteration will steadily contain more voxels.
Figure 2. Mesoscopic scale functional imaging achieved in this study and evaluation of the effects of different data processing strategies. (A) The activation map was displayed in the reconstructed cerebral cortex space. The area indicated by the black dotted line is the second visual area, which is shown in (B), (C) and (D) after zooming in. (B) Functional domains selective to color information. (C) Functional domains selective to disparity information. Data processing strategies, such as voxel up-sampling, spatial transformation composing, cortical mesh refinement and intra-cortical smoothing, can present finer functional domains (indicated by arrows). (D) The spatial relationship between the two distinct functional domains. Red represents "color columns" and green represents "disparity columns". The blue is the overlap. (see main text for quantitative analysis).
This study provides systematic analysis of standard methods used in fMRI analysis and provides instructive steps on how to effectively retain resolution for the purpose of fine-scale functional mapping. This investigation provides a methodological reference critical for mesoscopic scale functional imaging. Reviewers expressed high praise for this study, noting its significant contribution to understanding mesoscale units of brain function, which include functional cortical columns, laminae, and organizations within subcortical nuclei.
The corresponding authors of this study are Professor Anna Wang Roe (王菁), Interdisciplinary Institute of Neuroscience and Technology, Zhejiang University, and Jonathan R. Polimeni, Massachusetts General Hospital, Harvard University. The first author is Jianbao Wang (王剑葆), a doctoral candidate in Professor Anna Wang Roe's research group. The project was co-supervised by Professor Nasr Shahin of Massachusetts General hospital, Harvard University.
Tips from the authors:
This study can be simply understood as avoiding the "beautify function" of scientific imaging. Many image software in daily life, such as skin smoothing, whitening and other functions can help us beautify the picture. But for scientific research, every detail has meaning, so studies need to keep the image authentic or at least understand what is being lost with beautification. Thus, this study helps neuroimagers retain important information needed for visualizing important millimeter-scale functional structures in the brain.
Some members from Anna Wang Roe(the 3rd from the left) Group and Jonathan Polimeni(the 1st from the left)
Full text link:
https://onlinelibrary.wiley.com/doi/10.1002/hbm.25867
Lab website:
http://www.ziint.zju.edu.cn/Anna/index.php/Index/index1.html
原文链接:https://www.cell.com/cell-reports/fulltext/S2211-1247(22)00428-4
Hisashi Tanigawa 副教授
博士于2001年在日本大阪大学医学院获得博士学位。 随后,他分别于日本理化学研究所脑科学综合研究中心(2002年到2006年) 、 美国范德比尔特大学心理所(2007年到2011年)从事博士后工作。 随后,他担任了日本新潟大学医学院的助理教授(2011年至2013年) 、 以及新潟大学跨学科学术中心的副教授(2013年到2017年) 。 在2017年5月,他正式加入浙江大学。
研究方向:
课题组的研究目标是在灵长类大脑皮层中探索高级认知功能(如物体识别,注意,工作记忆,长时程记忆)背后的神经机制。我们的研究着手于猕猴腹侧视觉通路,以及前额叶皮层的功能和解剖结构。实验室在猕猴上采用多种实验技术(包括内源信号光学成像,皮层脑电图,多电极阵列记录, 近红外神经刺激,神经示踪)开展这些研究工作。
实验室主页:
http://www.ziint.zju.edu.cn/index.php/Index/Czindex.html?userid=34
课题组诚聘博士后一名,欢迎点击下方招聘链接!
http://www.ziint.zju.edu.cn/index.php/join/zindex.html?tid=6&userid=34
图3. DCBT纳米颗粒标记的小鼠脑血管的活体三光子荧光成像图。(A)通过颅窗的三光子荧光成像示意图;(B-K)不同深度脑血管的成像图(100-1010 μm);(L-O)FWHM曲线。
图4. 不同三光子荧光脑血管成像深度的三维重构图。
图5. DCBT纳米颗粒标记的小鼠脑血管的活体穿颅三光子荧光成像图。(A)穿颅三光子荧光成像示意图;(B-G)颅下不同深度脑血管的成像图(0-500 μm);(H)FWHM曲线和SBR分析;(I,J)三光子荧光颅下脑血管成像的三维重构图。
鉴于DCBT纳米颗粒优异的三光子吸收性能和近红外发光特性,DCBT纳米颗粒可成功应用于颅骨去除(图4-5)或完整小鼠(图6)的三光子荧光脑血管成像,成像结果显示出较高的空间分辨率。得益于DCBT纳米颗粒的近红外二区三光子激发(1550 nm)和近红外一区发射,其通过开颅窗的组织成像深度可达1010 μm,穿颅的组织成像深度为500 μm,显示出优异的组织穿透深度。这项研究将为开发用于生物医学应用的多光子吸收材料提供更多的启发。同时也为三光子大深度活体成像探针提供了新的可能,帮助理解脑内深部血管在正常以及疾病状态下的结构以及动态功能变化。
课题组简介
奚望 副教授
个人主页:
https://person.zju.edu.cn/0008720
唐本忠 院士
个人主页:
https://lhs.cuhk.edu.cn/teacher/264
郑正 研究员
个人主页:
http://hgxy.hfut.edu.cn/2018/1117/c9456a240602/page.htm
浙江大学系统神经与认知科学研究所奚望副研究员以及浙江大学药学院陈忠教授团队合作研究了颞叶癫痫中从急性到长期的病理过程中神经血管耦合功能的动态变化,发现了神经血管耦合功能在癫痫病理过程中长期的耦合功能变化规律。相关成果以标题为“Long-term development of dynamic changes in neurovascular coupling after acute temporal lobe epilepsy”发表在Brain Research,文章链接https://authors.elsevier.com/a/1ej5f1aPVenMy。文章的第一作者为浙江大学系统神经与认知科学研究所研究助理柳荫。系统所多光子成像平台为本文提供了技术支持。浙江大学系统所副研究员奚望以及药学院陈忠教授为本文的共同通讯作者。
颞叶癫痫(TLE)是一种常见的神经病理学,其特征是神经元异常同步激活,导致反复发作。TLE治疗受到临床问题的困扰,例如耐药性和伴随的共病,包括认知和运动功能障碍。此外,TLE的病因复杂,形成过程漫长;这种机制还不完全清楚。已经提出了许多TLE模型来阐明癫痫发生机制,包括电生理离子通道、神经元高兴奋性、神经可塑性和慢性神经炎症的改变。除了神经元机制外,能量代谢和脑细胞基质供应异常等血管机制与癫痫也有复杂的联系。
为了验证神经血管耦合(NVC)在癫痫发作中的功能障碍,我们开发了长期双光子成像来记录从急性到慢性TLE过程中的神经元和血管活动变化。这是第一次对清醒小鼠在致痫过程中神经元和血管系统的变化进行如此长时间的成像。我们发现,在清醒小鼠中,皮层扩散性抑制(CSD)期间神经血管耦合表现出强相关性,而在急性癫痫期间则表现出弱相关性。我们发现,癫痫持续状态后,NVC转向非线性动力学。此外,在自发和足底刺激输入状态下,血管与神经元信号的相关性表现出急性到慢性TLE的病理过程相对应动态变化过程。我们的研究结果可能提供一种新的、可靠的方法来描述癫痫从急性期到慢性期的神经血管耦合功能动态变化。为癫痫疾病的治疗提供新的可能的治疗靶点。
课题组简介
浙江大学系统神经与认知科学研究所副研究员奚望:先后在南京大学和中国科学院神经所获得学士和博士学位。主要研究兴趣为应用多光子成像技术研究从啮齿类到非人灵长类的神经血管耦合相关机制,以及神经血管耦合机制在脑功能如睡眠、注意、意识等神经认知功能中的作用。相关工作以第一作者及通讯作者发表于Biomaterials、J. Nucl. Med.、Theronatics等国际知名杂志。主持与参与多项国家自然科学基金、浙江省自然科学基金等研究项目。