The research themes involve the continuing effort to develop more robust, informative, and quantitative methods for mapping human brain function. We made significant progress this year - both in publications and unpublished findings that will be written up quickly. To summarize, the following were among the highlights, organized by research project corresponding to the listed post doc or section/colaborating staff scientist: Rasmus Birn: Significant MRI signal changes can be induced by variations in the depth and rate of breathing. These occur throughout gray matter, but are largest in many of the same brain regions involved in ?default mode? brain function during rest, which can lead to errors in functional connectivity analysis. The relative latency of these signal changes is correlated with the latency of breath-hold induced signal changes, suggesting that the signal changes are induced by similar mechanisms, most likely slow CO2-mediated blood flow changes. Removing these respiration-related fluctuations, in addition to more conventional physiological corrections, can improve the temporal signal-to-noise ratio (and therefore the detection of neuronal activity) by up to 50%. Furthermore, the average response to a single deep breath can be used to model (and predict) fMRI signal changes in response to other respiratory manipulations, such as breath-holding, rate changes, and depth changes. In an additional study, we showed that a blocked design with 10s task and 10s rest conditions can map functional activation during an overt speech task with no significant task-related motion artifact. Using this design in an fMRI study of verbal fluency, we found greater activation in the left inferior frontal gyrus during letter fluency, and greater activation in the left fusiform gyrus, and anterior and superior left frontal gyrus during category fluency, suggesting that letter and category fluency performance is dependent on partially distinct neural circuitry. David Knight: Fear memories are composed of distinct declarative (aware) and nondeclarative (unaware) features. To identify brain regions thatsupport these separate memory processes, we measured conditional fear, contingency awareness, and functional magnetic resonance imaging signal during a conditioning procedure in which tones that predicted an aversive event were presented at supra and sub-threshold volumes. Contingency awareness developed in conjunction with learning-related parahippocampal and fusiform activity on perceived conditioning trials only. In contrast, conditional fear and differential amygdala activity developed on both perceived and unperceived trials. These findings demonstrate the distinct roles separate neural circuits play in declarative and nondeclarative fear memory. Jerzy Bodurka: Due to recent NIH advances in MRI technology, resulting in 3-fold sensitivity improvements, it was possible to explore fMRI spatial resolution limits more carefully. A recently published NeuroImage paper (3) addresses how increased resolution reduces signal dropout throughout the brain and shows improved BOLD detection in MTL using parallel imaging at high spatial resolution. Another important question related to fMRI spatial limit is, what would be the MRI voxel volume at which BOLD detection is still feasible? Simulations and experiments were conducted to address this practical issue(5). This work introduced ?suggested? voxel volume for fMRI at which thermal or system related noise equals physiological noise. The obtained cubical suggested voxel volume for brain gray matter of 1.8 mm^3 shows high spatial resolution fMRI with adequate BOLD sensitivity is possible. Sean Marrett: We have continued to investigate the neural mechanisms involved in perceptual decision making using both fMRI and MEG. In a paper published in PNAS (5) in 2006 we reported that involvement of the left posterior DLPFC during decision making transcends both task and response specificity, which suggests it provides a key flexibility in linking sensory evidence, decision making and behavior. Wenming Luh: Pulsed Arterial Spin Labeling (PASL) techniques spatially label the blood proximal to the imaging volume and allow the tagged blood to arrive at the imaging brain tissues after a certain transit time that varies among imaging voxels and with functional tasks. For quantitative perfusion, saturation pulses can be applied at time TI1 to control the temporal width of the tag such as in QUIPSS II. However, when tagging below the brain for measuring perfusion of lower brain tissues or attempting whole brain coverage, the majority of the tag resides in large arteries with fast flowing blood of varying velocity respect to cardiac phases. We have demonstrated the ineffectiveness of TI1 saturation when tagging below the brain for quantitative PASL experiments. The variation of tag temporal width with cardiac phases will increase ASL signal fluctuation and reduce quantitative accuracy. The short temporal tag width will limit achievable SNR and volume coverage. Niko Kriegeskorte: The analysis of functional magnetic resonance imaging (fMRI) data benefits from multivariate analysis of contiguous sets of voxels targeting the information they jointly encode. We found that such information-based analyses are sensitive to neuronal information undetected by conventional massive univariate analyses. In particular, these analyses revealed information about the identity of visually presented individual faces in anterior inferotemporal cortex, but not ? as current theory would suggest ? in the fusiform face area. We also found that the temporal structure of fMRI responses may contain information about the spatial structure of neuronal activity patterns and vice versa and that information-based analysis should be designed for sensitivity to nonseparable spatiotemporal hemodynamic patterns. Regarding the noise, we found that close-by fMRI voxels tend to be positively correlated as an artefact of the imaging technique. Anthony Boemio: To understand the transformation that occurs at the interface between the low-level auditory representation of speech and stored linguistic codes, we created a novel acoustic stimulus in which the formant structure from fully connected American English sentences is extracted and parametrically shifted in time to produce varying temporal coherence. Functional magnetic resonance imaging (fMRI) shows that the auditory-linguistic transformation occurs both in the anterior and posterior superior temporal gyri -- in agreement with previous speech studies -- as well as in the inferior frontal as well as in the inferior frontal gyrus and basal ganglia which may play a role in parsing the speech stream. Kevin Murphey: Reliable detection of activation in single subjects at high resolution is becoming a more common desire among fMRI researchers who are interested in comparing individuals rather than populations. Detection of activation at higher resolutions requires longer scan durations. The relationship between TSNR and the necessary fMRI scan duration required to obtain significant results at varying P values was determined both experimentally and theoretically(11). We showed that with voxel volumes of ~10mm^3 at 3T, and a corresponding TSNR of ~50, the required number of time points that guarantees detection of signal changes of 1% is about 860, but if TSNR increases by only 20%, the time for detection decreases by more than 30%. This implies that imaging of columnar resolution (effect size = 1% and assuming a TR of 1sec) at 3T will require either 10 minutes for a TSNR of 60 or 40 minutes for a TSNR of 30. This implies that sucess or failure of imaging at high resolution critically depends on Temporal Signal to Noise (TSNR).
Showing the most recent 10 out of 36 publications
