【作者】 曾翎；

【导师】 尧德中；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2008， 博士

【摘要】 脑功能磁共振成像(Functional magnetic resonance imaging fMRI)主要依据血氧水平依赖性(blood oxygenation level dependent，BOLD)对比增强原理进行成像，是目前人们掌握的有效的无侵入、可精确定位脑功能活动的研究手段，具有很高的空间分辨率和存在进一步提高时间分辨率的潜力，非常适合神经活动的时空分析和脑的高级功能研究，已受到神经、认知和临床等领域的极大关注。本文围绕功能磁共振成像(fMRI)在脑科学研究中的应用，对磁共振图像的分割和配准、脑功能的定量分析、采用独立成分分析(ICA)和支持向量机(SVM)对脑功能活动图像的研究，成果有比较系统性的创新；同时结合认知科学研究中的热点问题进行了脑功能的应用研究。详细内容如下：1)提出了基于互信息脑功能磁共振图像配准新方法。互信息作为衡量两幅图像配准的相似性测度函数，假设当两幅图像配准时，互信息达到最大值。采用了无需计算梯度的Powell直接搜索算法。磁共振成像(MRI)的配准实验证明，互信息法能准确地实现多模态医学图像的配准，并且能达到亚象素的精度。2)提出了利用空间独立成分分析来处理静息态fMRI数据方法，首次将静息态脑功能的低频振荡理论应用于独立成分分析静态数据的成分选择，通过Z分数选择了静息态下的活动点和去除独立噪点，然后通过频谱分析选择主要能量集中在0．01Hz到0．1Hz的独立成分，进而采用聚类分析得出脑功能连接网络。最后对结果进行了分析与评价验证了我们的方法的可靠性，并得出本文所采用的方法和对时间序列的频谱求相关的结论是类似。3)运用SPM(Statistical Parameter Mapping)对功能磁共振数据分析处理，并用主成分分析进行时间压缩等综合数据处理方法，去训练功能磁共振成像数据，选取最佳的支持向量机核函数，从训练结果中提取重要的体现大脑活动差异的权重向量，由此探测左右手动状态下大脑活动区域。4)用功能磁共振成像血氧水平依赖性变化率的范数来定量分析左右脑功能区之间不同的血氧水平依赖性动态反应活动，揭示大脑功能的不对称性。6个右利手功能磁共振实验表明：不论是单手还是双手运动实验左脑区的血氧水平信号变化的强度比右脑区的弱，该结果提供了脑功能不对称性的功能磁共振证据：右利手被试左脑的功能区形成了自适用系统，仅需要少量的神经元就可以完成运动功能。右手经常活动会使左脑的功能区域活动强度变弱。右脑的功能区域的活动强度比左脑的功能区域的活动强度大。更多还原

【Abstract】 Functional magnetic resonance imaging (fMRI) is mainly based on blood oxygenation level dependent (BOLD). It is the most efficient instrument that can be used to precisely locate brain function activities without invasion. With very high spatial resolution and potential high temporal resolution, fMRI is well fit for the spatial and temporal analysis of neural action and the research of advanced brain function. So it is being concerned by many science branches such as neuroscience, cognition and clinic et al.Focused on the application of fMRI in brain science, systematic improvements are conducted from MRI image segmentation and registration, brain functional quantitative analysis, brain function activation analysis based on independent component analysis and support vector machine. Meanwhile, combined with hot question of cognitive science, brain functional activation application study is implemented. The details are shown as follow:1)A new image registration algorithm based on the mutual information was proposed. As the measured function of scaling the similarity of two images, the mutual information reach to maximization when the images registration of two pictures is well done. Powell direct searching method is adopted to accelerate the image registration and avoid calculating gradient. An experiment result demonstrates that our method can achieve the registration of multimodal medical image precisely, and could reach sub-pixel precision.2) Technique of data processing on functional magnetic resonance imaging (fMRI) by using spatial independent component analysis (sICA) method in the resting state was proposed. Firstly, the low-frequency oscillations theory is applied to the choice of components of interest (COI) by sICA method. The activation voxels and noise voxels are specified by Z value of separated sICA component. Then COI whose energy concentrates between 0.01Hz and 0.1Hz were chosen through spectrum analysis. And then, the functional connectivity networks were obtained using hierarchical clustering. Finally, by analyzing and evaluating the results, It has proved that our method can be feasible and our results can be related to the frequency method of the time courses.3) SPM(Statistical parameter mapping) is adopted to process fMRI data .Then principal components analysis is proposed to make time series data compression. These comprehensive methods are combined to train fMRI data. Finally, the weight vector that manifests the cerebrum activity difference was acquired by selecting most superior nuclear function of support vector mapping. The region of cerebrum activity was detected in the state of right and left hands movement activity.4) The norm of fMRI blood oxygenation level dependent (BOLD) signal percent change was introduced to quantitatively measure BOLD signal intensity change difference between left and right motor areas. The results of the data collected from six subjects show that the norm of BOLD signal percent change in right motor area is higher than that in left motor area both for two hand movement and single hand movement with right handedness. These results from fMRI show the asymmetry of motor areas and may suggest that the left brain motor area comes into being as an adaptation system, which only need few neuron cell to finish movement task for right handedness. The left motor area activation intensity is reduced with normal right finger movement. The right motor activation intensity is higher than the left motor activation intensity.更多还原