Protocol number 93-M-0170, NCT00001360 The Section on Functional Imaging Methods (SFIM) has advanced functional MRI (fMRI) methodology through development of novel paradigms, acquisition methods, and processing methods Our work over the past year can be grouped into the following categories: Layer-Specific fMRI, Dynamic Functional Connectivity and Activity, Naturalistic tasks, Individual Differences, Simultaneous EEG-fMRI, Underpinnings of resting state fMRI signal changes, and Pulse Sequence Development and Optimization. Layer-specific fMRI methods: We continued to develop acquisition fMRI methods that can measure brain activity with ultra-high resolution, aiming to distinguish activity across cortical layers and columns. The novel parameter space explored since October 2017 includes the following: A study comparing the capabilities of layer-fMRI at ultra-high field strengths of 9.4T compared to 7T has been completed. We could investigate layer-dependent signal features across different cortical columns in M1 in a participant-specific surface-based signal evaluation scheme to identify individual finger representations in primary motor cortex. We could investigate the layer-dependent activity in the primary sensory system distinguishing feed-forward input during finger touching vs. expectation feed-back input during a mental prediction task. We developed a new software suite in C++, LAYNII for more streamlined analysis of layer fMRI. We could translate the layer-fMRI technology to be applicable in cognitive brain areas including the dorso-lateral-prefrontal cortex and applied it with a working memory task. Dynamic Functional Connectivity and Activity: Understanding how FC states populate such high dimensional space is important, as it will help us address questions such as whether FC dynamics should be modeled as a continuous or discrete space, what mathematical tools are best suited to group FC states and what aspects of FC spatial distributions are affected by mental conditions. In this project we are exploring the use of different linear (e.g., PCA) and non-linear (e.g., ISOMAP, Spectral Embedding, t-SNE) dimensionality reduction techniques to compute low (e.g., 2 or 3) dimensional embeddings of FC states during both task and rest. We also explore how these embeddings can help us make better informed decisions regarding how to compare and group FC states. Naturalistic Tasks: In the last year, we have continued to make advances in collecting and analyzing fMRI data while subjects are engaged in naturalistic paradigms (i.e., watching a movie or listening to story). We are exploring the possibility of using such tasks as brain stress tests to draw out individual variation as it relates to intrinsic personality and cognitive traits. Specific accomplishments in the past year include the following: a) We have demonstrated that healthy individuals who score higher on a measure of trait-level paranoia have distinct brain and behavioral responses to an ambiguous social narrative. This serves as proof-of-concept that an identical stimulus can evoke different neural and behavioral reactions across individuals, and that these differences stem from an intrinsic trait that is relevant to psychiatric illness. b) Using a large scale, publicly available data set, we have explored a large space of potential approaches to analyzing naturalistic fMRI data. We found that inter-subject correlation, which takes advantage of the time-locked nature of the stimulus across subjects but does not make any assumptions about the structure of the task, is more sensitive to phenotypic differences across subjects than either traditional stimulus-derived regressors or functional connectivity. We are currently confirming this result using other data sets and exploring additional analyses using regions of interest delineated by inter-subject correlation. c) We have begun developing and piloting new and innovative experimental paradigms that combine naturalistic tasks during fMRI with detailed behavioral and phenotypic assessment. In the next year we expect to launch a full-scale data collection effort on both healthy controls and psychiatric patients. Individual Differences: In the last year, we have continued to make progress using functional brain connectivity to characterize individuals and predict behavior at the single-subject level. Specifically, we have made the following advances: a) We have shown that functional connectivity profiles extracted from fMRI data acquired during a task, as opposed to during the resting state, yields more accurate predictions of a sophisticated cognitive trait (fluid intelligence). We are currently characterizing the differences in predictions derived from different tasks, and extending these results to additional behaviors. b) We have shown that relative differences in functional connectivity across subjects are more informative than absolute differences. We have demonstrated this via a comparison of various data processing methods that emphasize either relative or absolute changes, with the aim of optimizing a pipeline for behavioral prediction from functional connectivity data. A manuscript on these results is in preparation. Simultaneous EEG-fMRI: While fMRI has set the standard for spatial resolution in millimeters (and sub-millimeter in 7-Tesla), it sacrifices precision in temporal resolution most often averaging over the span of 12 seconds. EEG provides excellent temporal resolution in milliseconds, and also represents actual neuronal firings/activity. With technological advances in the past ten years, multiple manufacturers have developed simultaneous EEG-fMRI systems, capable of recording both modalities (EEG, fMRI) at the same time. This technological advance, also paired with the increase in number of EEG sensors, provides the possibility of more sophisticated methods of combining data with high temporal and high spatial information. We have begun to identify how simultaneous EEG-fMRI changes the structure of signal and noise in the fMRI data. To do this, we have combined simultaneous data with our methods developed for multi-echo fMRI denoising (particularly those with independent component analysis; ICA) to identify if the sensors themselves change the detection of signal and noise components, and whether the EEG can similarly be used as an additional variable in the ICA to aid in denoising of the multi-echo fMRI data. Underpinnings of resting state fMRI signal changes: We have started analyzing data as co-investigators on a BRAIn grant with Elizabeth Hillman of Columbia University. Dr. Hillman has been using wide field optical mapping in awake, behaving mice to acquire GCaMP (a Calcium-based neural activity measure) and hemodynamic signals across much of the rostral cortex. With these data, we can apply and attempt to validate the dynamic functional connectivity metrics that are often used with fMRI. Pulse Sequence Development and Optimization: Our work developing and utilizing ME fMRI continues. In collaboration with Dr. Caballero-Gaudes, we are working on the detection of individual BOLD events with no information about the task timing or location. Preliminary tests show that the algorithm can reliably detect BOLD events associated with each individual trial in a manner similar to traditional GLM methods, yet without any information on paradigm timing. This new method has important applications including: uncovering hemodynamic events driving dynamic resting functional connectivity, and detecting inter-ictal events in epilepsy patients without the need for concurrent EEG recordings.
Showing the most recent 10 out of 84 publications
