Improved Methods for Single Subject FMRI Analysis for Clinical Application

描述（由申请人提供）：精神疾病对受影响的个人来说是一个巨大的负担，对社会来说是一个巨大的经济负担。据估计，精神障碍的年度费用为1500亿美元，而且每年都在增加，这一总数还不包括因精神障碍而领取残疾津贴的300多万人。当务之急是，我们必须优先考虑研究工作，重点了解大脑功能，以改进诊断策略，发现更有效的治疗方法。功能磁共振成像（fMRI）是可视化和测量典型和非典型认知加工的有力工具。然而，许多重要的认知处理系统，如与记忆、语言、情感和执行控制相关的系统，只产生小的BOLD信号，因此测量结果是嘈杂的，统计置信度很低。因此，功能磁共振成像还没有很好地应用于个别患者的临床诊断。我建议大力改进fMRI数据噪声源的抑制方法，使fMRI从研究人群的工具转变为研究个体认知功能的一致、准确的诊断工具。采用每一种噪声抑制算法必须表现良好才能可靠地检测到单次试验fMRI BOLD信号的策略，我开发了可视化方法来“看到”深入到fMRI数据中，以评估fMRI数据处理的每一步的数据质量。初步的研究表明，有明显的机会，以提高fMRI图像分析技术。拟议的研究将首先开发和测试方法，以改善运动和生理波动的错误抑制。然后，它将通过将这些技术与模式识别相结合来转化这项研究，以表征典型和非典型人群的个体认知激活模式。我的定量科学专长是图像处理、算法设计和模式识别。这项研究直接支持了我跨学科的职业发展，在实验计划、功能磁共振成像扫描仪操作、神经科学课程以及应用于严重脑障碍人群的新软件方法方面拥有实践经验。特别是，本研究的研究对象将包括具有脆性X综合征、特纳综合征、自闭症、威廉姆斯综合征、抑郁症和双相情感障碍等疾病的重要临床精神病学人群，以便所有新开发的方法都可以立即付诸实践。

DESCRIPTION (provided by applicant): Mental illness is a great burden for the affected individual and economically costly for society. The annual cost of mental disorders has been estimated to be $150 billion, increasing every year, and this total does not include more than three million people receiving disability benefits due to mental disorders. It is imperative that we prioritize research efforts focused on understanding brain function in order to improve diagnostic strategies and discover more effective therapies. Functional Magnetic Resonance Imaging (fMRI) is a powerful tool to visualize and measure typical and atypical cognitive processing. However, many important cognitive processing systems, such as those associated with memory, language, emotion and executive control, only produce small BOLD signals and thus measurements are noisy and have low statistical confidence. Hence, fMRI has not been readily adopted for clinical diagnosis of individual patients. I propose to develop greatly improved methods to suppress the noise sources in fMRI data in order to transform fMRI from a research tool about populations to a consistent and accurate diagnostic tool to study individual cognitive functions. Using the strategy that every noise suppression algorithm must perform well to reliably detect single trial fMRI BOLD signals, I developed visualization methods to "see" deeply into fMRI data to evaluate the quality of the data at every step of fMRI data processing. The preliminary studies indicate that there are clear opportunities to improve fMRI image analysis techniques. The proposed research will first develop and test methods to improve suppression of errors from motion and physiological fluctuations. Then it will translate this research by combining these techniques with pattern recognition to characterize individual cognitive activation patterns in typical and atypical populations. My quantitative science expertise is in image processing, algorithm design, and pattern recognition. The research directly supports my interdisciplinary career development with hands-on experience in experiment planning, fMRI scanner operation, neuroscience coursework, and new software methods for application to severely brain disordered populations. In particular, the subjects for this research will include important clinical psychiatric populations with disorders such as fragile X syndrome, Turner syndrome, autism, Williams syndrome, depression, and bipolar disorder, so that all newly developed methods can be immediately put into practice.