论文部分内容阅读
大脑是掌控人体所有活动的枢纽,哪怕是一个极小的动作,大脑也会产生大量的信息去完成。而f MRI就是致力于找出人体活动或者身体异常现象与大脑皮层各个功能区域对应关系的技术。本文的研究的对象是在已知刺激下采集到的f MRI数据。首先用MDL算法估计去噪后的数据的信源个数。然后利用混合凸集分析算法(CAM)对f MRI数据进行特征提取。CAM能够有效克服一般盲信号分离算法假设信源独立的局限性。同时,针对f MRI数据量大、维数高的特点,采用PCA对数据进行降维,然后用K-mean方法聚类,实现对CAM算法的优化,提高了CAM的运算效率。最后设计实验,比较CAM算法和ICA算法特征提取的结果。实验证明CAM算法得到较高的相关度,且能够更好的解释大脑激活区与刺激信号的对应关系。
The brain is the hub of all the activities of the human body, even if it is a very small move, the brain will produce a lot of information to complete. And f MRI is to find out the relationship between human activities or body abnormalities and the various functional areas of the cerebral cortex. The study in this paper is based on f MRI data acquired with known stimuli. First, we use MDL algorithm to estimate the number of sources of denoised data. Then, the feature extraction of f MRI data is performed by using the CAM algorithm. CAM can effectively overcome the limitations of the blind source separation algorithm. At the same time, according to the characteristics of large amount of f MRI data and high dimensionality, PCA is used to reduce the dimensionality of the data and then the K-mean method is used to optimize the CAM algorithm and to improve the computational efficiency of CAM. Finally, the experiment is designed to compare the results of CAM algorithm and ICA algorithm feature extraction. The experiment proves that the CAM algorithm obtains a higher correlation and can better explain the correspondence between the brain activation area and the stimulation signal.