Daniel Rio

Statistical analysis of functional MRI data in the wavelet domain

The use of the wavelet transform is explored for the detection of differences between brain functional magnetic resonance images (fMRIs) acquired under two different experimental conditions. The method benefits from the fact that a smooth and spatially localized signal can be represented by a small set of localized wavelet coefficients, while the power of white noise is uniformly spread throughout the wavelet space. Hence, a statistical procedure is developed that uses the imposed decomposition orthogonality to locate wavelet-space partitions with large signal-to-noise ratio (SNR), and subsequently restricts the testing for significant wavelet coefficients to these partitions. This results in a higher SNR and a smaller number of statistical tests, yielding a lower detection threshold compared to spatial-domain testing and, thus, a higher detection sensitivity without increasing type I errors. The multiresolution approach of the wavelet method is particularly suited to applications where the signal bandwidth and/or the characteristics of an imaging modality cannot be well specified. The proposed method was applied to compare two different fMRI acquisition modalities, Differences of the respective useful signal bandwidths could be clearly demonstrated; the estimated signal, due to the smoothness of the wavelet representation, yielded more compact regions of neuroactivity than standard spatial-domain testing.

The Corticotropin Releasing Hormone-1 (CRH1) Receptor Antagonist Pexacerfont in Alcohol Dependence: A Randomized Controlled Experimental Medicine Study

Extensive preclinical data implicate corticotropin-releasing hormone (CRH), acting through its CRH1 receptor, in stress- and dependence-induced alcohol seeking. We evaluated pexacerfont, an orally available, brain penetrant CRH1 antagonist for its ability to suppress stress-induced alcohol craving and brain responses in treatment seeking alcohol-dependent patients in early abstinence. Fifty-four anxious alcohol-dependent participants were admitted to an inpatient unit at the NIH Clinical Center, completed withdrawal treatment, and were enrolled in a double-blind, randomized, placebo-controlled study with pexacerfont (300 mg/day for 7 days, followed by 100 mg/day for 23 days). After reaching steady state, participants were assessed for alcohol craving in response to stressful or alcohol-related cues, neuroendocrine responses to these stimuli, and functional magnetic resonance imaging (fMRI) responses to alcohol-related stimuli or stimuli with positive or negative emotional valence. A separate group of 10 patients received open-label pexacerfont following the same dosing regimen and had cerebrospinal fluid sampled to estimate central nervous system exposure. Pexacerfont treatment had no effect on alcohol craving, emotional responses, or anxiety. There was no effect of pexacerfont on neural responses to alcohol-related or affective stimuli. These results were obtained despite drug levels in cerebrospinal fluid (CSF) that predict close to 90% central CRH1 receptor occupancy. CRH1 antagonists have been grouped based on their receptor dissociation kinetics, with pexacerfont falling in a category characterized by fast dissociation. Our results may indicate that antagonists with slow offset are required for therapeutic efficacy. Alternatively, the extensive preclinical data on CRH1 antagonism as a mechanism to suppress alcohol seeking may not translate to humans.

Decreased cingulate and precuneate glucose utilization in alcoholic Korsakoff's syndrome

Localized cerebral glucose utilization was determined for nine abstinent alcoholic men with Korsakoffs syndrome and 10 age-matched normal men who underwent positron emission tomography with [18F]2-fluoro-2-deoxyglucose (FDG). Patients with Korsakoffs syndrome showed relatively decreased glucose utilization in cingulate and precuneate areas. These decreases persisted even after correction for group differences in ventricular and sulcal cerebrospinal fluid measured on computed tomography. Electroencephalographic recordings at the time of FDG uptake showed no group differences, a finding that demonstrates that the metabolic differences could not be explained by differences in physiological arousal at the time of scanning. It is concluded that the decreased glucose utilization in the patients reflects a disruption of memory circuitry, the Papez circuit, caused by diencephalic lesions induced by thiamine deficiency.

Intensity-adaptive segmentation of single-echo T1-weighted magnetic resonance images

A procedure for segmentation of intracranial tissues, including cerebrospinal fluid surrounding the brain, cortical and subcortical gray matter, and white matter, in a T1‐weighted magnetic resonance image of the brain, has been developed. The proposed method utilizes information from the histogram of pixel intensities of the intracranial image. Based on this information, an unsupervised K‐means clustering procedure separates various tissue regions. Information about the approximate location of anatomical regions within the intracranial space is used to detect ventricles and the caudate nuclei. First a description and justification for the procedure is presented. Then the performance of the procedure is evaluated by analysis of variance. In conclusion, the results of applying this procedure to 31 healthy subjects are presented and future improvements are discussed. Hum. Brain Mapping 5:194–205, 1997.

Head motion during positron emission tomography : is it significant?

High sensitivity for detecting local brain function differences from subsequent PET images acquired at different cerebral stimulation states requires interscan head motion to be minimized. This motion was measured by an optical lever system during scanning (130 min) of 15 subjects in a dual-dose injection study. Despite motion restriction by a face-mask restraint system, rotations in the sagittal and coronal planes (up to 4.1 degrees and 2.4 degrees, respectively) significantly influenced the measured means and variances of local metabolic differences between states. Hence, adjustments for head movement by retrospective, digital slice realignment or, better, real-time corrections are important.

Monitoring the source of memory in detoxified alcoholics

The ability to monitor the source of remembered information and related reflective cognitive processes was examined in normal volunteers and detoxified alcoholics. Normal volunteers were very accurate judges of whether remembered events were presented as stimuli or were self-generated, even when memory was tested 2 days later. In contrast, a subgroup of otherwise cognitively unimpaired alcoholics demonstrated impairments in the ability to track the source of remembered knowledge and were also less able to inhibit intrusion errors in recalling information from memory. These findings provide preliminary evidence of an impairment in cognitive control functions in certain alcoholics. This conclusion is supported by associated findings indicating that, among alcoholics, performance on explicit memory tasks that required reflective cognitive operations were positively correlated with glucose utilization rates in left prefrontal, temporal, and posterior orbital frontal cortical regions.

Statistical analysis of image differences by wavelet decomposition

The wavelet transform was studied for the analysis of glucose utilization differences between subject groups shown in PET images. To strengthen statistical inference, it was of particular interest investigating the trade-off between signal localization and image decomposition into uncorrelated components. This trade-off is governed by wavelet regularity, and was found to be best for third-order orthogonal spline wavelets. Only about 1.6% of the components were statistically different (p<.05) from noise, constituting a sufficient set to synthesize local image differences by the inverse wavelet transform.<<ETX>>

Fully automated segmentation of cerebrospinal fluid in computed tomography

A method is presented for automated delineation and measurement of cerebrospinal fluid (CSF) regions in computed tomographic (CT) sections. Regions of skull and scalp are removed by using a linear discriminant analysis approach. Beam-hardening artifact is reduced by subtracting from each section the average radial intensity profile, characterized by a polynomial function. Remaining intensity gradients are suppressed by implementing CSF segmentation with a local thresholding technique based on maximum-entropy principles. CSF fractions from 12 regions of interest (ROIs) were measured in 10 patients with alcoholic Korsakoff syndrome and 9 normal volunteers. The same ROIs were also assessed by an interactive segmentation method, which enabled the operator to compensate for beam-hardening distortions by selecting suitable threshold values for each ROI. Both methods identified the same ROIs as displaying statistically significant differences between the two subject groups. However, interactive segmentation underestimated sulcal CSF by 20-70%, which was confirmed by applying both methods to CT scans of an anthropomorphic phantom. Hence, in contrast to interactive thresholding, unsupervised segmentation relies on firmly fixed criteria that reduce the influence of beam-hardening distortions and provide more objective results.

Use of the wavelet transform to investigate differences in brain PET images between patient groups

Suitability of the wavelet transform was studied for the analysis of glucose utilization differences between subject groups as displayed in PET images. To strengthen statistical inference, it was of particular interest investigating the tradeoff between signal localization and image decomposition into uncorrelated components. This tradeoff is shown to be controlled by wavelet regularity, with the optimal compromise attained by third-order orthogonal spline wavelets. Testing of the ensuing wavelet coefficients identified only about 1.5% as statistically different (p < .05) from noise, which then served to resynthesize the difference images by the inverse wavelet transform. The resulting images displayed relatively uniform, noise-free regions of significant differences with, due to the good localization maintained by the wavelets, very little reconstruction artifacts.

Single subject image analysis using the complex general linear model-An application to functional magnetic resonance imaging with multiple inputs

A linear time invariant model is applied to functional fMRI blood flow data. Based on traditional time series analysis, this model assumes that the fMRI stochastic output sequence can be determined by a constant plus a linear filter (hemodynamic response function) of several fixed deterministic inputs and an error term assumed stationary with zero mean. The input function consists of multiple exponential distributed (time delay between images) visual stimuli consisting of negative and erotic images. No a priori assumptions are made about the hemodynamic response function that, in essence, is calculated at each spatial position from the data. The sampling rate for the experiment is 400 ms in order to allow for filtering out higher frequencies associated with the cardiac rate. Since the statistical analysis is carried out in the Fourier domain, temporal correlation problems associated with inference in the time domain are avoided. This formal model easily lends itself to further development based on previously developed statistical techniques.