基于内源信号的脑功能光学成像血管伪迹检测及去除方法研究
发布时间：2018-10-08 21:26
【摘要】：20世纪70年代以来，一系列新的成像技术的出现使得神经科学家和医生们可以有效地观察到脑的功能性活动，为解释脑的奥秘提供了有效方法。其中基于内源信号的脑功能光学成像是一种前沿的脑成像技术，它因空间分辨率较高，可长时间在体记录，结构简单等特点，可以有效地研究大脑功能特性，在脑功能研究中发挥了重要作用。虽然基于内源信号的脑功能光学成像的原理和系统构成比较简单，但在活体成像实验过程中存在诸如呼吸、心跳以及血管周期性波动等生物噪声，其引起的皮层反射光强变化往往远高于刺激相关的皮层活动信号，这些生物噪声降低了成像信噪比，需要经过一系列图像信号处理之后才能从噪声中提取到所需要的真正反应大脑功能的信息，因而研究适用于光学成像的图像处理方法成为了研究热点。早期光学成像研究中使用的叠加平均及差分等的方法可以有效地从噪声中提取信号，提高信噪比，但在某些情况下，噪声仍然较强，存在的血管伪迹也降低了图像质量，因而工程技术人员一直在致力于研究和建立适于内源信号脑光学成像的图像处理新方法。 在本课题组进行的内源信号脑光学成像实验中，得到的图像信号较强但存在血管伪迹较强的问题，降低了信噪比。因此本文旨在探索多种血管伪迹提取及去除方法，然后从中寻找到最适合于本课题组实际实验结果的血管伪迹提取及去除方法。本文首先介绍了课题中主要应用的血管伪迹提取方法：方差法和局部相似度最小化方法，两者都可以取得对血管伪迹较准确的预测，但是方差法存在着对于局部细节预测不够准确并且有预测错误的情况；而局部相似度最小化方法则不能对血管伪迹的本身形态、长宽粗细等给予很好的表达，因此本文在这里创新性地把两种方法结合使用，得到了比单独使用两种方法更好的血管伪迹预测效果。其次本文系统研究了血管伪迹去除的算法，提出用血管伪迹周围灰度值中位值去除的方法，可以有效的对血管伪迹进行去除，进一步比较了不同参数对该算法的影响，并进行了性能分析。本文在血管伪迹提取去除算法的基础上，创新性地把该方法与主成分分析、独立成分分析相结合，证明了把主成分分析和独立成分分析作为数据预处理手段，再使用血管伪迹提取及去除方法，可以很大程度上改善内源信号脑光学成像图像质量，比单独使用主成分分析和独立成分分析，能够取得更好的效果。 本文的主要结论是：把方差法和局部相似度最小化法相结合可以预测出较好的血管伪迹，在此基础上用中位值去除血管伪迹能得到较好的伪迹处理效果。PCA和ICA作为图像预处理手段，结合上述伪迹检测和去除手段，能进一步增强图像信噪比，提高图像质量。
[Abstract]:Since the 1970s, a series of new imaging techniques have enabled neuroscientists and doctors to effectively observe the functional activities of the brain and provide an effective way to explain the mysteries of the brain. Among them, brain functional optical imaging based on endogenous signal is a kind of advanced brain imaging technology. Because of its high spatial resolution, long time in vivo recording, simple structure and so on, it can effectively study the functional characteristics of the brain. It plays an important role in the study of brain function. Although the principle and system of brain functional optical imaging based on endogenous signals are relatively simple, biological noises such as breathing, heartbeat and periodic fluctuations of blood vessels exist in the experimental process of in-vivo imaging. The cortical reflectance intensity changes are often much higher than the cortical activity signals associated with stimulation, and these biological noises reduce the imaging signal-to-noise ratio (SNR). It takes a series of image signal processing to extract the real information of brain function from noise, so the research of image processing method suitable for optical imaging has become a hot topic. The superposition averaging and differential methods used in the early optical imaging research can effectively extract signals from noise and improve SNR. However, in some cases, the noise is still strong, and the existing vascular artifacts also reduce the image quality. Therefore, engineers have been working to study and establish a new image processing method suitable for endogenous signal brain optical imaging. In the experiment of brain optical imaging with endogenous signal, the result shows that the image signal is strong but the vascular artifact is strong, which reduces the signal-to-noise ratio (SNR). Therefore, the purpose of this paper is to explore a variety of methods for vascular artifact extraction and removal, and then to find out the most suitable methods for vascular artifact extraction and removal. This paper first introduces the main methods of vascular artifact extraction: variance method and local similarity minimization method, both of which can obtain accurate prediction of vascular artifact. However, the variance method can not predict the local details accurately and has the wrong prediction, while the local similarity minimization method can not give a good expression of the shape of the vascular artifacts, length, width, thickness, etc. Therefore, this paper innovatively combines the two methods to obtain a better prediction effect of vascular artifacts than using the two methods alone. Secondly, the algorithm of vascular artifact removal is systematically studied in this paper, and the method of removing the median value of the gray value around the vascular artifact is put forward, which can effectively remove the vascular artifact, and the influence of different parameters on the algorithm is further compared. The performance analysis is also carried out. On the basis of vascular artifact extraction and removal algorithm, this paper innovatively combines this method with principal component analysis and independent component analysis, and proves that principal component analysis and independent component analysis are used as data preprocessing methods. The method of vascular artifact extraction and removal can improve the image quality of endogenous signal brain optical imaging to a great extent, which is better than that of principal component analysis and independent component analysis alone. The main conclusion of this paper is that the combination of variance method and local similarity minimization method can predict better vascular artifacts. On this basis, using median value to remove vascular artifact can get better artifact processing effect. PCA and ICA can be used as image preprocessing means. Combining with the above artifact detection and removal methods, image SNR can be further enhanced and image quality can be improved.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0
本文编号：2258320
[Abstract]:Since the 1970s, a series of new imaging techniques have enabled neuroscientists and doctors to effectively observe the functional activities of the brain and provide an effective way to explain the mysteries of the brain. Among them, brain functional optical imaging based on endogenous signal is a kind of advanced brain imaging technology. Because of its high spatial resolution, long time in vivo recording, simple structure and so on, it can effectively study the functional characteristics of the brain. It plays an important role in the study of brain function. Although the principle and system of brain functional optical imaging based on endogenous signals are relatively simple, biological noises such as breathing, heartbeat and periodic fluctuations of blood vessels exist in the experimental process of in-vivo imaging. The cortical reflectance intensity changes are often much higher than the cortical activity signals associated with stimulation, and these biological noises reduce the imaging signal-to-noise ratio (SNR). It takes a series of image signal processing to extract the real information of brain function from noise, so the research of image processing method suitable for optical imaging has become a hot topic. The superposition averaging and differential methods used in the early optical imaging research can effectively extract signals from noise and improve SNR. However, in some cases, the noise is still strong, and the existing vascular artifacts also reduce the image quality. Therefore, engineers have been working to study and establish a new image processing method suitable for endogenous signal brain optical imaging. In the experiment of brain optical imaging with endogenous signal, the result shows that the image signal is strong but the vascular artifact is strong, which reduces the signal-to-noise ratio (SNR). Therefore, the purpose of this paper is to explore a variety of methods for vascular artifact extraction and removal, and then to find out the most suitable methods for vascular artifact extraction and removal. This paper first introduces the main methods of vascular artifact extraction: variance method and local similarity minimization method, both of which can obtain accurate prediction of vascular artifact. However, the variance method can not predict the local details accurately and has the wrong prediction, while the local similarity minimization method can not give a good expression of the shape of the vascular artifacts, length, width, thickness, etc. Therefore, this paper innovatively combines the two methods to obtain a better prediction effect of vascular artifacts than using the two methods alone. Secondly, the algorithm of vascular artifact removal is systematically studied in this paper, and the method of removing the median value of the gray value around the vascular artifact is put forward, which can effectively remove the vascular artifact, and the influence of different parameters on the algorithm is further compared. The performance analysis is also carried out. On the basis of vascular artifact extraction and removal algorithm, this paper innovatively combines this method with principal component analysis and independent component analysis, and proves that principal component analysis and independent component analysis are used as data preprocessing methods. The method of vascular artifact extraction and removal can improve the image quality of endogenous signal brain optical imaging to a great extent, which is better than that of principal component analysis and independent component analysis alone. The main conclusion of this paper is that the combination of variance method and local similarity minimization method can predict better vascular artifacts. On this basis, using median value to remove vascular artifact can get better artifact processing effect. PCA and ICA can be used as image preprocessing means. Combining with the above artifact detection and removal methods, image SNR can be further enhanced and image quality can be improved.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0
【参考文献】
相关期刊论文 前3条
1 张濵,俞洪波,寿天德;应用基于内源信号的光学成像技术的视觉脑研究现状[J];动物学研究;2000年01期
2 陈昕,寿天德;用脑光学成像精确测定猫初级视皮层视野拓扑投射关系[J];生理学报;2003年05期
3 张昆鸟,俞洪波,寿天德;基于内源信号的脑光学成像系统的研制[J];生物物理学报;1999年03期
，本文编号：2258320
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2258320.html
