Methods Development

Image Acquisition Methods Development

It has been the philosophy of the FNL from its inception that the examination of neuropsychiatric questions using imaging should not be limited by existing techniques. There has always therefore been an emphasis on developing new methodology, as needed, to better address specific neurobiological questions. Multiple PET and fMRI acquisition methods have been developed. Many of these techniques have been incorporated into the ongoing studies of the FNL, such as the reduction of susceptibility artifact at the base of the brain using z-shim echo planar imaging methods,silent, event-related, sparse-sampling fMRI method which avoids the confound of scanner noise (which is especially important for auditory and some psychiatric studies), and high resolution fMRI.

Processing and Analysis Methods for Functional Magnetic Resonance Imaging

In the FNL, fMRI analysis methods are continually being developed and optimized to improve detection of the small fMRI signals associated with the neuropsychological paradigms used in the MRI scanner. This work is being done with and by Dr. Hong Pan, with his trainees. Analytic techniques have been developed to improve detection power and estimation accuracy: (1) Based on variance component analysis, appropriate statistical modeling of residuals in fMRI data has been incorporated, which includes additive confounds (such as estimated head motion, physiological noise, global CBF fluctuation, scanner drift, experimental periods) and temporal correlation. (2) Based on mixed-effects statistical models, the intra- and inter-subject variabilities have been taken into account to allow results to be more generalizable to the populations studied. (3) Multi-level hierarchical mixed-effects models have been developed, with more advanced statistical modeling of the variance-mean relationship, for the multi-level nested data structures (such as trials/ runs/ sessions/ subjects/ subject groups/ study sites) and heteroscedastic variances (i.e. differing variance at different values of a random variable) common in fMRI data sets, to take advantage of and to meet the challenge presented by multi-site studies. (4) Correlation analyses of behavioral, physiological and clinical measures with fMRI data provides additional and important information about underlying pathophysiology. (5) Extraction of temporal trends in fMRI time-course provides information about changes in brain activity over the course of the experiment, which is important in understanding modulation of function. (6) Region-of-interest analyses, looking with greater specificity at the structure and function of particular brain areas (including the amygdala, hippocampus, and ventromedial prefrontal cortex), are being conducted. (7) Significant improvements have been made in existing, external software for performing voxel-based morphometry (VBM), a technique for examining differences in structure across the entire brain. (8) Multivariate statistical techniques have been employed to allow the study of interactions among brain areas, which is crucial, as we now understand that it is how brain regions function together, rather than activity in isolated areas, that is also likely to be relevant in the study of neuropsychiatric disorders. Overall, these novel approaches to characterizing and modeling many aspects of the MRI signal are resulting in improved ability to detect subtle but important changes in key brain regions. In addition to statistical and computational advances, numerous improvements in automation, visualization, and software user interfaces have been made as well.

Neuropsychological Paradigm Development, Psychophysics and Psychophysiological Measures

The FNL has significant expertise in neuropsychological paradigm development. The interpretation of functional differences detected between particular groups of subjects is largely dependent upon what neuropsychological task the subject is performing during the activation study. Task design must be targeted to probe either a brain function of interest, or particular neurocircuitry. Exquisite attention must be paid to creating appropriate control conditions to isolate the cognitive component of interest (e.g. in a subtraction paradigm), or to parametrically vary one component of the task to hone in on the corresponding brain function.

An alternative to using activation paradigms is to examine differences in resting state patterns of BOLD fMRI signal in different subject populations, which can provide complementary information.  In addition, the FNL routinely uses MRI blood flow techniques to obtain resting activity measures.

Multimodal fMRI/EEG

Combining non-invasive methods with high spatial and high temporal resolution of neural activity is essential for resolving the complex neural dynamics and functional connectivity of human perceptual and cognitive networks. We use multimodal fMRI/EEG to probe neural networks with high spatiotemporal resolution. We previously reported a linear correlation in human auditory cortex between hemodynamic fMRI and mass-neural EEG measurements, supporting the idea that common neural sources contribute to these responses, as demonstrated unequivocally in invasive animal studies.