CASIA OpenIR  > 毕业生  > 硕士学位论文
基于神经影像的精准经颅磁刺激方法研究
程鑫乐
2024-05-14
Pages79
Subtype硕士
Abstract

经颅磁刺激(Transcranial Magnetic ResonanceTMS)是一种大脑皮质神经的无创性刺激技术,其原理在于通过颅内感应电刺激实现神经调控。尽管传统的TMS方法在临床实践中应用广泛,但其靶点定位的精度和稳定性一直是制约其治疗效果的关键因素。医生手持线圈的治疗方法难以保证治疗过程中刺激位置的稳定,而5cm法则既不能根据个体头部尺寸引入个体化信息,也缺少理论支撑。精准TMS的提出旨在解决这些问题。目前大部分精准TMS已经通过机器人手持线圈解决治疗过程中保证刺激位置稳定的问题,但是刺激靶点的个体化定位方法及其理论研究还比较少。神经影像为个体大脑功能和结构的研究提供了丰富的数据支持,并已在诸多领域取得显著成果。本研究主要借助神经影像数据,提取相关的大脑功能与结构信息,对精准TMS应用中手部热点和抑郁症治疗靶点的精准定位进行研究。研究分为两个部分展开。

第一部分,提出了一种基于图谱映射的手部热点快速定位方法。手部热点是大脑运动皮层对刺激响应最为明显的区域,在每个被试的TMS治疗方案中扮演着至关重要的角色。传统手部热点的定位方案往往比较耗时,需要在病人运动区周围施加多次的单脉冲刺激,并记录相应的运动诱发电位(Motor Evoked PotentialMEP),随后通过数据拟合确定手部热点的位置。本研究以34名抑郁症患者为对象,利用个体手部热点位置绘制了群组手部热点概率图谱和频率图谱,并应用于被试的手部热点定位。同时,在研究过程中,发现存在一些被试的手部热点位置过于分散,无法直接应用手部热点图谱进行定位。因此,从磁共振影像(Magnetic Resonance ImagingMRI)中提取的大脑在运动皮层的结构信息,主要包括皮层沟回指数、皮层曲率、皮层厚度等数据,被用于训练一个机器学习分类模型,帮助从人群中筛选适合应用图谱定位手部热点位置的被试。研究结果表明,手部热点概率图谱能够准确反应群组手部热点的分布规律,且个体手部热点位置的空间分布与抑郁症患者的状况无关,验证了图谱在抑郁症患者中的适用性。手部热点频率图谱表明定位的准确率能够达到58%。而基于大脑结构信息数据训练的机器学习分类模型,可以有效筛选能够应用图谱的被试,提高图谱定位的准确率,最终使该快速定位方法准确率提升至88%,相较于针对每个被试采集MEP的手部热点定位方法,平均节约了50%的时间,提高了临床手部热点定位的效率。

第二部分,提出了一个TMS抑郁症治疗靶点定位方法。抑郁症是一种常见且难以治疗的精神疾病,其核心症状包括情绪低落、动力减退、失去兴趣和乐趣,其治疗一直是临床研究的重点。TMS作为一种有效的治疗方法,其靶点定位的准确性直接关系到治疗效果。然而,目前关于TMS对大脑的影响机制以及如何利用该机制进行靶点定位的研究尚显不足。本研究基于35名抑郁症患者在治疗期间三个时间点的MRI和汉密尔顿抑郁量表(Hamilton Depression ScaleHAMD)评分,利用偏最小二乘回归(Partial Least-Squares RegressionPLSR)模型分析了实际刺激靶点功能连接与治疗效果之间的相关性,挖掘一系列与抑郁症疗效高度相关的脑网络组图谱(Brainnetome AtlasBNA)定义的脑区,并基于这些BNA脑区与疗效之间的相关性,提出了一个抑郁症治疗靶点定位方法。其中相关性较高的11个(前5%BNA脑区被作为感兴趣脑区,用于进一步分析功能连接和功能活动的变化情况。研究结果表明,基于实际刺激靶点与BNA脑区的功能连接可以成功预测抑郁症的疗效以及最优治疗靶点的位置。进一步的分析发现,治疗响应组和非响应组的脑区功能连接和功能活动有着不一样的变化。治疗响应组功能连接保持稳定,非响应组功能连接先减少后增加。响应组治疗靶点处的功能活动保持稳定,非响应组功能活动在治疗第二周后发生显著变化。这些结果说明虽然TMS在治疗期间会对大脑神经活动进行调控,但是TMS治疗后一周以及两周,刺激位置的功能活动变化不会发生明显变化,刺激位置的功能活动变化更可能是受到内在脑网络的控制。这些结果还为抑郁症治疗靶点的选择提供了重要的理论依据和实践指导,即靶点定位应尽量避开那些与感兴趣脑区存在高功能连接的刺激位置。

Other Abstract

Transcranial Magnetic Resonance (TMS) is a non-invasive stimulation of the cerebral cortical nerves that is essentially an intracranial inductive electrical stimulation. However, the current method of TMS treatment is flawed both in terms of targeting precision and targeting method. It is difficult to ensure the stability of the stimulation position during treatment by relying on the physician's hand-held coil treatment method, and the 5 cm rules cannot be used to incorporate individualised information according to the individual's head size and lacks theoretical justification. Therefore, precision TMS has been proposed to solve these problems. Most precision TMS ensure the stability of the stimulation position during treatment using robotic hand-held coils, but there is still less research on the individualised positioning method of the stimulation target and its theoretical study. Currently, neuroimaging is widely used to study the function and structure of the individual brain, and many results have been obtained. The present study focuses on extracting relevant information about brain function and structure using neuroimaging data to investigate the precise localisation of hand motor hotspots (hMHS) and therapeutic targets for depression in the application of precision TMS. The study consists of two parts.

In the first part, the hMHS is the most obvious region in the motor cortex that responds to stimulation and is mainly used to determine the intensity of individualised TMS, and the therapeutic targets for some diseases are also determined based on the location of the hMHS. Thus, the identification of the hMHS plays an important role in the TMS treatment of each subject. Current hMHS identification protocols are often time-consuming, requiring the acquisition of multiple single-pulse stimuli around the patient's motor cortex and the recording of motor evoked potentials (MEPs), which are then fitted to the data to obtain the hMHS. In this study, 34 depressed patients were used to draw group hMHS probability maps and frequency maps using individual hMHS locations, which were applied to locate hMHS in the subjects. Meanwhile, there were some subjects whose hMHS could not be directly applied to hMHS mapping for localisation because they were too scattered. Therefore, structural information of the brain in the motor cortex extracted from magnetic resonance imaging (MRI), which mainly includes data such as cortical sulcus, curvature, and thickness, was used to train a machine learning model that helped to select those subjects who were able to identify the hMHS by applying the atlas. The results of the study showed that the hMHS probability mapping successfully identified the hMHS and verified that the spatial distribution of individual hMHS was not related to the condition of the depressed patients, indicating the feasibility of applying the mapping to the depressed patients; and the hMHS frequency mapping showed that the identification accuracy could reach 58%. The machine learning model, which was trained based on the brain structure information data, can effectively select the subjects who can apply the atlas and improve the accuracy rate of the atlas. Finally, the fast identification method achieved an accuracy of 88%, saving an average of 50% time compared to the identification method that collects MEP for each subject.

In the second part, a TMS depression therapeutic target localization method is proposed. Depression is a common and difficult-to-treat mental illness, whose core symptoms include depressed mood, decreased motivation, and loss of interest and pleasure, and its treatment has always been the focus of clinical research.TMS, as an effective therapeutic method, has a direct relationship with the accuracy of its target localization to the therapeutic effect. However, there is a lack of research on the mechanism by which TMS affects the brain and how it can be utilized for target localization. In this study, based on the MRI and Hamilton Depression Scale (HAMD) scores of 35 depressed patients at three time points during the treatment period, we analyzed the correlation between the functional connectivity of the real stimulation targets and the treatment effect using Partial Least-Squares Regression (PLSR) model. The correlation between real stimulation target functional connectivity and therapeutic efficacy was analyzed using PLSR modeling, a series of Brainnetome Atlas (BNA)-defined brain regions that are highly correlated with the therapeutic efficacy of depression were excavated, and a therapeutic target localization method for depression was proposed based on the correlation between these BNA brain regions and the therapeutic efficacy. Eleven (top 5%) of the BNA brain regions with high correlations were used as brain regions of interest for further analysis of changes in functional connectivity and functional activity. The results of the study showed that functional connectivity based on real stimulation targets with BNA brain regions can successfully predict the efficacy of depression and the location of optimal therapeutic targets. Further analysis revealed that the functional connectivity and functional activity of brain regions in the treatment-responsive and non-responsive groups had different changes. Functional connectivity remained stable in the treatment-responsive group and decreased and then increased in the non-responsive group. Functional activity at the treatment target remained stable in the response group and changed significantly after the second week of treatment in the non-response group. These results suggest that TMS does not affect functional activity changes at the stimulation location, and that functional activity changes are more likely to be controlled by intrinsic brain networks. These results also provide an important theoretical basis and practical guidance for the selection of therapeutic targets for depression, i.e., target localization should try to avoid stimulation locations that have high functional connectivity with brain regions of interest.

Keyword经颅磁刺激 神经影像 手部热点 抑郁症 机器学习
Language中文
IS Representative Paper
Sub direction classification脑网络分析
planning direction of the national heavy laboratory脑启发多模态智能模型与算法
Paper associated data
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/56627
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
程鑫乐. 基于神经影像的精准经颅磁刺激方法研究[D],2024.
Files in This Item:
File Name/Size DocType Version Access License
毕业论文.pdf(4207KB)学位论文 限制开放CC BY-NC-SA
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[程鑫乐]'s Articles
Baidu academic
Similar articles in Baidu academic
[程鑫乐]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[程鑫乐]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.