Power Analysis and Optimization for FMRI Experiments

在广泛使用的功能磁共振方法学中 成像(FMRI)，正式控制功率(即1-假阴性率)有 比起假阳性率的控制，受到的关注较少。对…的控制 要量化两者的阴性fMRI结果的可信度，功率是必要的 基本的科学和临床环境。临床应用的一个显著例子 能够通过先验功率通知功能磁共振实验设计的效用 计算是在神经外科手术的术前功能磁共振测试中 职能本地化候选人。显然，在这种情况下， 阴性可能会给患者带来不良后果。在基础科学中， 功能磁共振实验的先验功率计算将提供以下能力 用期望的否定结果驳斥脑认知模型 自信。具体目标1是制定一种算法来确定 对fMRI数据进行参数(即t和F)检验的体素幂。相比之下， 与以前的方法相比，该算法将考虑到时间上的影响 噪声的自相关结构，完整的实验设计， 感兴趣的假设、假设的信噪比和滤波 神经活动到功能磁共振成像转换的特性。具体目标2是 探讨神经活动转化的饱和特征 改变为功能磁共振信号改变。更好地描述饱和度将 改善对fMRI效应大小的预测，从而提高 功率计算。在有意义的生理范围内的饱和度将是 通过测试与fMRI反应叠加的偏差进行检查(在 主要感觉-运动和早期视觉区域)由一组1，2， 或三个紧密间隔的(1秒视觉刺激。具体目标3是 研究与事件相关的功能磁共振设计的能力对 它们的时间结构。提高实验设计能力的能力 通过操纵它们的时间结构(不改变成本)将 使高功率fMRI实验更具可行性。

In the widely used methodology of functional magnetic resonance imaging (fMRI), the formal control of power (i.e., 1-false-negative rate) has received less attention than the control of false-positive rate. The control of power is necessary for quantifying confidence in negative fMRI results in both basic scientific and clinical settings. A striking example of the clinical utility of being able to inform fMRI experimental design by a priori power calculations is in pre-operative fMRI testing of neurological surgery candidates for localization of function. Clearly in such circumstances a false negative could have undesirable consequences for the patient. In basic science, a priori power calculations for fMRI experiments would afford the ability to refute brain-cognition models based on negative results with a desired confidence. Specific Aim 1 is to formulate an algorithm for determining voxel-wise power of parametric (i.e., t and F) tests on fMRI data. In contrast to previous methods, this algorithm will account for the temporal autocorrelation structure of the noise, the complete experimental design, the hypothesis of interest, assumed signal:noise ratios, and the filtering properties of the neural activity-to-fMRI transform. Specific Aim 2 is to investigate saturation characteristics of the transformation of neural activity change to fMRI signal change. Better characterization of saturation will improve predictions of fMRI effect magnitude and hence improve the accuracy of power calculations. Saturation over meaningful physiological ranges will be examined by testing for deviations from superposition of fMRI responses (in primary sensori-motor and early visual regions) elicited by sets of one, two, or three closely spaced (1 sec visual stimuli. Sepcific Aim 3 is to investigate the dependence of the power of event-related fMRI designs on their temporal structure. The ability to improve power of experimental designs by manipulation of their temporal structure (without changing the cost) will make high power fMRI experiments more feasible.