Search

The paper presents a method for learning multimodal classifiers from datasets in which not all subjects have data from all modalities. Usually, subjects with a severe form of pathology are the ones failing to satisfactorily complete the study, especially when it consists of multiple imaging modalities. A classifier capable of handling subjects with unequal numbers of modalities prevents discarding any subjects, as is traditionally done, thereby broadening the scope of the classifier to more severe pathology. It also allows design of the classifier to include as much of the available information as possible and facilitates testing of subjects with missing modalities over the constructed classifier. The presented method employs an ensemble based approach where several subsets of complete data are formed and trained using individual classifiers., The output from these classifiers is fused using a weighted aggregation step giving an optimal probabilistic score for each subject. The method is applied to a spatio-temporal dataset for autism spectrum disorders (ASD) (96 patients with ASD and 42 typically developing controls) that consists of functional features from magnetoencephalography (MEG) and structural connectivity features from diffusion tensor imaging (DTI). A clear distinction between ASD and controls is obtained with an average 5-fold accuracy of 83.3% and testing accuracy of 88.4%. The fusion classifier performance is superior to the classification achieved using single modalities as well as multimodal classifier using only complete data (78.3%). The presented multimodal classifier framework is applicable to all modality combinations.

This paper presents a paradigm for generating a quantifiable marker of pathology that supports diagnosis and provides a potential biomarker of neuropsychiatric disorders, such as autism spectrum disorder (ASD). This is achieved by creating high-dimensional nonlinear pattern classifiers using support vector machines (SVM), that learn the underlying pattern of pathology using numerous atlas-based regional features extracted from diffusion tensor imaging (DTI) data. These classifiers, in addition to providing insight into the group separation between patients and controls, are applicable on a single subject basis and have the potential to aid in diagnosis by assigning a probabilistic abnormality score to each subject that quantifies the degree of pathology and can be used in combination with other clinical scores to aid in diagnostic decision. They also produce a ranking of regions that contribute most to the group classification and separation, thereby providing a neurobiological insight into the pathology. As an illustrative application of the general framework for creating diffusion based abnormality classifiers we create classifiers for a dataset consisting of 45 children with ASD (mean age 10.5 ± 2.5 yr) as compared to 30 typically developing (TD) controls (mean age 10.3 ± 2.5 yr). Based on the abnormality scores, a distinction between the ASD population and TD controls was achieved with 80% leave one out (LOO) cross-validation accuracy with high significance of p<0.001, ~84% specificity and ~74% sensitivity. Regions that contributed to this abnormality score involved fractional anisotropy (FA) differences mainly in right occipital regions as well as in left superior longitudinal fasciculus, external and internal capsule while mean diffusivity (MD) discriminates were observed primarily in right occipital gyrus and right temporal white matter., (Copyright © 2011 Elsevier Inc. All rights reserved.)

The paper presents a method for creating abnormality classifiers from high angular resolution diffusion imaging (HARDI) data. We utilized the fiber orientation distribution (FOD) diffusion model to represent the local WM architecture of each subject. The FOD images are then spatially normalized to a common template using a non-linear registration technique. Regions of homogeneous white matter architecture (ROIs) are determined by applying a parcellation algorithm to the population average FOD image. Orientation invariant features of each ROI's mean FOD are determined and concatenated into a feature vector to represent each subject. Principal component analysis (PCA) was used for dimensionality reduction and a linear support vector machine (SVM) classifier is trained on the PCA coefficients. The classifier assigns each test subject a probabilistic score indicating the likelihood of belonging to the patient group. The method was validated using a 5 fold validation scheme on a population containing autism spectrum disorder (ASD) patients and typically developing (TD) controls. A clear distinction between ASD patients and controls was obtained with a 77% accuracy.

This work presents an automated method for partitioning neuronal white matter (WM) into regions of interest with uniform WM architecture. These regions can then be used to replace atlas-derived regions for any subsequent statistical analysis. The fiber orientation distribution function is used as a model of WM architecture resulting in a voxel similarity function sensitive to both fiber orientations and configurations. The method utilizes the normalized cuts algorithm to partition WM voxels based on this similarity function along with a connected component labeling algorithm to ensure spatial compactness. We illustrate the algorithms ability to discern regions based on both orientation and complexity through its application to a simulated fiber crossing and an in-vivo dataset.

Magnetic resonance imaging (MRI) is often used by electrophysiologists to target specific brain regions for placement of microelectrodes. However, the effectiveness of this technique has been limited by few methods to quantify in three dimensions the relative locations of brain structures, recording chambers and microelectrode trajectories. Here we present such a method. After surgical implantation, recording chambers are fitted with a plastic cylinder that is filled with a high-contrast agent to aid in the segmentation of the cylinder from brain matter in an MRI volume. The resulting images of the filled cylinder correspond to a virtual cylinder that is projected along its long axis - parallel to the trajectories of microelectrodes advanced through the recording chamber - through the three-dimensional image of the brain. This technique, which does not require a stereotaxic coordinate system, can be used to quantify the coverage of an implanted recording chamber relative to anatomical landmarks at any depth or orientation. We have used this technique in conjunction with Caret [Van Essen DC, Drury HA, Dickson J, Harwell J, Hanlon D, Anderson CH. An integrated software suite for surface-based analyses of cerebral cortex. J Am Med Inform Assoc 2001;8:443-59] and AFNI [Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res 1996;29:162-73] brain-mapping software to successfully localize several regions of macaque cortex, including the middle temporal area, the lateral intraparietal area and the frontal eye field, and one subcortical structure, the locus coeruleus, for electrophysiological recordings.

In this work, a novel method for determining the principal directions (maxima) of the diffusion orientation distribution function (ODF) is proposed. We represent the ODF as a symmetric high-order Cartesian tensor restricted to the unit sphere and show that the extrema of the ODF are solutions to a system of polynomial equations whose coefficients are polynomial functions of the tensor elements. In addition to demonstrating the ability of our methods to identify the principal directions in real data, we show that this method correctly identifies the principal directions under a range of noise levels. We also propose the use of the principal curvatures of the graph of the ODF function as a measure of the degree of diffusion anisotropy in that direction. We present simulated results illustrating the relationship between the mean principal curvature, measured at the maxima, and the fractional anisotropy of the underlying diffusion tensor.

Rationale and Objectives: Physiological noise in blood oxygen level-dependent functional magnetic resonance imaging (BOLD fMRI) has been shown to have characteristics similar to the BOLD signal itself, suggesting that it may have a vascular dependence. In this study, we evaluated the influence of physiological noise in fMRI as revealed by the differences in vasculature sensitivity of gradient-echo echo-planar imaging (GE-EPI) and spin-echo EPI (SE-EPI)., Materials and Methods: The contribution of physiological noise to the fMRI signal during activation of the visual cortex was assessed by comparing its temporal characteristics with respect to echo time (TE), using both GE-EPI and SE-EPI. The correlation of the noise in fMRI with apparent diffusion coefficient (ADC) and the number of components required to describe its variance, as determined by principal-component analysis (PCA), were also assessed., Results: The SE-EPI data were less affected by a TE-dependence of noise, in contrast to the apparent physiological noise in GE-EPI. Voxel-wise analysis revealed that total apparent noise increased as ADC values increased, and the relationship was different for GE-EPI and SE-EPI. PCA revealed that while the number of components characterizing the noise in SE-EPI data increased in a TE-dependent manner, approaching that of white noise at long echo time, the number of components from GE-EPI data was TE-independent., Conclusions: The difference in sensitivities to physiological noise between SE-EPI and GE-EPI suggests that extravascular BOLD processes around draining veins contribute significantly to physiological noise in BOLD fMRI, and the suppression of this noise component may enhance SE-EPI BOLD sensitivity at higher fields.

A continuous-wave (cw) near-infrared spectroscopy (NIRS) instrument has been developed to noninvasively quantify fetal cerebral blood oxygen saturation (StO2). A linear Green's function formulism was used to analytically solve the photon diffusion equation and extract the time-varying fetal tissue oxy- and deoxy-hemoglobin concentrations from the NIR measurements. Here we explored the accuracy with which this instrument can be expected to perform over a range of fetal hypoxic states. We investigated the dependence of this accuracy on the accuracy of the reference optical properties chosen based on the literature. The fetal oxygenation of a pregnant ewe model was altered via maternal aortic occlusion. The NIR cw instrument was placed on the maternal abdomen directly above the fetal head, continuously acquiring diffuse optical measurements. Blood was sampled periodically from the fetus to obtain fetal arterial saturation (SaO2) measurements from blood gas analysis. The NIR StO2 values were compared with the fetal SaO2 measurements. Variations in the NIR results due to uncertainty in the reference optical properties were relatively small within the fetal SaO2 range of 30 to 80%. Under hypoxic conditions, however, the variability of the NIR StO2 calculations with changes in the assumed reference properties became more significant.

The application of T1 in the rotating frame (T1rho) to functional MRI in humans was studied at 3 T. Increases in neural activity increased parenchymal T1rho. Modeling suggested that cerebral blood volume mediated this increase. A pulse sequence named spin-locked echo planar imaging (SLEPI) that produces both T1rho and T2* contrast was developed and used in a visual functional MRI (fMRI)experiment. Spin-locked contrast significantly augments the T2* blood oxygen level-dependent (BOLD) contrast in this sequence. The total functional contrast generated by the SLEPI sequence (1.31%) was 54% larger than the contrast (0.85%) obtained from a conventional gradient-echo EPI sequence using echo times of 30 ms. Analysis of image SNR revealed that the spin-locked preparation period of the sequence produced negligible signal loss from static dephasing effects. The SLEPI sequence appears to be an attractive alternative to conventional BOLD fMRI, particularly when long echo times are undesirable, such as when studying prefrontal cortex or ventral regions, where static susceptibility gradients often degrade T2*-weighted images., ((c) 2005 Wiley-Liss, Inc.)

The temporal resolving power of blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) at 3T was investigated in the visual and auditory cortices of the human brain. By using controlled temporal delays and selective visual hemifield stimulation, regions with similar (left vs. right occipital cortex) and different (occipital cortex vs. auditory cortex) vascular architectures were compared. Estimates of the time-to-peak (TTP) of the BOLD hemodynamic response function (hrf) were obtained using a spin echo (SE) sequence and compared to those acquired using a traditional gradient echo (GE) sequence. The hrf TTP in the visual cortex was found to be 4.73 s and 4.21 s for GE and SE, respectively. The auditory cortex response was significantly delayed, with TTPs of 4.95 s and 4.51 s for GE and SE, respectively. The GE response was able to resolve visual stimuli separated by 250 ms, whereas SE could resolve stimuli 500 ms apart. Apparent-diffusion-coefficient (ADC) compartmentalization of the BOLD signal was applied to restrict the vascular sensitivity of the SE and GE sequences. Limiting the response to voxels with ADCs < 0.8 x 10(-3) mm(2)/s improved the temporal resolving power of GE and SE BOLD to 125 ms and 250 ms, respectively.

Recent theoretical and experimental work has suggested that spin echo (SE) functional MRI (fMRI) has improved localization of neural activity compared to gradient echo (GE) fMRI at high field strengths, albeit with a decrease in blood oxygenation level-dependent (BOLD) contrast. The present study investigated spatial and temporal variations in GE and SE fMRI at 3 T in response to a brief visual stimulus. The results demonstrate that SE BOLD contrast reaches its maximum amplitude more quickly than does GE contrast at long echo times. We have called this metric the peak hemodynamic activation time (PHAT). Because BOLD changes in response to increased neuronal activity occur earlier in the microvasculature and then later propagate into the venous compartment, these results provide further evidence that SE-based BOLD contrast provides superior localization to the site of activation at 3 T. Spatial overlay of SE and GE PHAT maps onto structural images reveal markedly different spatial profiles and further support the interpretation that shorter peak times correlate to improved spatial sensitivity.