Efficient Inferences in Task-Evoked Functional Magnetic Resonance Imaging Studies

Functional magnetic resonance imaging (fMRI) is a popular technology that measures brain activity by detecting changes associated with blood flow. A fundamental challenge of fMRI has been the limited sample sizes. The main reason for such a small sample size is the high scanning cost. If the sample size could be shrunk without compromising the information, a great amount of money could be saved and many additional cohorts could be explored. On the other hand, the rapid technological advances in brain imaging made it possible and routine to obtain high-resolution imaging data. While traditional methods of analysis may have produced acceptable results when the imaging data was in low resolution, the high dimensional images demand better statistical methods for more precise and efficient estimations for task fMRI studies. The research stands to make a significant contribution by better identifying the key components that play an important role in the learning, memorizing process and shed light on the cognitive growth. The methods to be developed have the potential to be modified for analyzing other types of high-dimensional data with limited sample sizes in medical research, such as genomics data. The program will also significantly impact the training of undergraduate and graduate students.

The goal of this project is to develop a series of comprehensive statistical methodologies for conducting efficient multivariate inferences for task fMRI studies. The investigators provide efficient methods to estimate the effect of exposure and personal characteristics on multivariate responses at a single time point as well as on the time-series data. The importance of the research lies in both the statistical methodology development and the interdisciplinary application. The research will provide theoretical justification for simultaneous dimension reduction and efficient estimation in the high dimensional data setting. Specifically, this research overcomes the technical difficulty that the dimension of brain imaging data is usually hundreds or thousands of times the sample size.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.