Automatic quantification of white matter hyperintensities on T2-weighted fluid attenuated inversion recovery magnetic resonance imaging.

White matter hyperintensities (WMH) are areas of increased signal visualized on T2-weighted fluid attenuated inversion recovery (FLAIR) brain magnetic resonance imaging (MRI) sequences. They are typically attributed to small vessel cerebrovascular disease in the context of aging. Among older adults, WMH are associated with risk of cognitive decline and dementia, stroke, and various other health outcomes. There has been increasing interest in incorporating quantitative WMH measurement as outcomes in clinical trials, observational research, and clinical settings. Here, we present a novel, fully automated, unsupervised detection algorithm for WMH segmentation and quantification. The algorithm uses a robust preprocessing pipeline, including brain extraction and a sample-specific mask that incorporates spatial information for automatic false positive reduction, and a half Gaussian mixture model (HGMM). The method was evaluated in 24 participants with varying degrees of WMH (4.9-78.6 cm3) from a community-based study of aging and dementia with dice coefficient, sensitivity, specificity, correlation, and bias relative to the ground truth manual segmentation approach performed by two expert raters. Results were compared with those derived from commonly used available WMH segmentation packages, including SPM lesion probability algorithm (LPA), SPM lesion growing algorithm (LGA), and Brain Intensity AbNormality Classification Algorithm (BIANCA). The HGMM algorithm derived WMH values that had a dice score of 0.87, sensitivity of 0.89, and specificity of 0.99 compared to ground truth. White matter hyperintensity volumes derived with HGMM were strongly correlated with ground truth values (r = 0.97, p = 3.9e-16), with no observable bias (-1.1 [-2.6, 0.44], p-value = 0.16). Our novel algorithm uniquely uses a robust preprocessing pipeline and a half-Gaussian mixture model to segment WMH with high agreement with ground truth for large scale studies of brain aging.

Brain areas with normatively greater cerebral perfusion in early life may be more susceptible to beta amyloid deposition in late life.

BACKGROUND: The amyloid cascade hypothesis characterizes the stereotyped progression of pathological changes in Alzheimer's disease (AD) beginning with beta amyloid deposition, but does not address the reasons for amyloid deposition. Brain areas with relatively higher neuronal activity, metabolic demand, and production of reactive oxygen species in earlier life may have higher beta amyloid deposition in later life. The aim of this study was to investigate early life patterns of perfusion and late life patterns of amyloid deposition to determine the extent to which normative cerebral perfusion predisposes specific regions to future beta amyloid deposition. MATERIALS AND METHODS: One hundred twenty-eight healthy, older human subjects (age: 56-87 years old; 44% women) underwent positron emission tomography (PET) imaging with [11C]PiB for measures of amyloid burden. Cerebral perfusion maps derived from 47 healthy younger adults (age: 22-49; 47%) who had undergone single photon emission computed tomography (SPECT) imaging, were averaged to create a normative template, representative of young, healthy adults. Perfusion and amyloid measures were investigated in 31 cortical regions from the Hammers atlas. We examined the spatial relationship between normative perfusion patterns and amyloid pathophysiology. RESULTS: The pattern of increasing perfusion (temporal lobe < parietal lobe < frontal lobe < insula/cingulate gyrus < occipital lobe; F(4,26) = 7.8, p = 0.0003) in young, healthy adults was not exactly identical to but approximated the pattern of increasing amyloid burden (temporal lobe < occipital lobe < frontal lobe < parietal lobe < insula/cingulate gyrus; F(4,26) = 5.0, p = 0.004) in older adults. However, investigating subregions within cortical lobes provided consistent agreement between ranked normative perfusion patterns and expected Thal staging of amyloid progression in AD (Spearman r = 0.39, p = 0.03). CONCLUSION: Our findings suggest that brain areas with normatively greater perfusion may be more susceptible to amyloid deposition in later life, possibly due to higher metabolic demand, and associated levels of oxidative stress and inflammation.

White Matter Regions With Low Microstructure in Young Adults Spatially Coincide With White Matter Hyperintensities in Older Adults.

Microstructural and macrostructural white matter damage occurs frequently with aging, is associated with negative health outcomes, and can be imaged non-invasively as fractional anisotropy (FA) and white matter hyperintensities (WMH), respectively. The extent to which diminished microstructure precedes or results from macrostructural white matter damage is poorly understood. This study evaluated the hypothesis that white matter areas with normatively lower microstructure in young adults are most susceptible to develop WMH in older adults. Forty-nine younger participants (age = 25.8 ± 2.8 years) underwent diffusion-weighted imaging (DWI), and 557 older participants (age = 73.9 ± 5.7 years) underwent DWI and T2-weighted magnetic resonance imaging (MRI). In older adults, WMH had a mostly periventricular distribution with higher frequency in frontal regions. We found lower FA in areas of frank WMH compared to normal-appearing white matter (NAWM) in older adults. Then, to determine if areas of normatively lower white matter microstructure spatially overlap with areas that frequently develop macrostructural damage in older age, we created a WMH frequency map in which each voxel represented the percentage of older adults with a WMH in that voxel. We found lower normative FA in young adults with regions frequently segmented as WMH in older adults. We conclude that low white matter microstructure is observed in areas of white matter macrostructural damage, but white matter microstructure is also normatively low (i.e., at ages 20-30) in regions with high WMH frequency, prior to white matter macrostructural damage.

White matter integrity as a mediator in the relationship between dietary nutrients and cognition in the elderly.

OBJECTIVE: We examined the association of nutrient intake with microstructural white matter integrity, and the role of white matter integrity in the association between nutrient consumption and cognition. METHODS: This cross-sectional analysis included 239 elderly (age ≥ 65 years) participants of a multiethnic cohort. White matter integrity was measured with fractional anisotropy (FA) from diffusion tensor magnetic resonance imaging. Nutrient patterns were derived from principal component analysis based on energy-adjusted intake of 24 selected nutrients. Generalized linear models were used to assess the association between nutrient patterns and mean FA of 26 white matter tracts. Mediation analysis was used to determine whether FA mediates the nutrient-cognition relationship. All models were adjusted for age at time of scan, gender, ethnicity, education, caloric intake, and apolipoprotein genotype. RESULTS: Among the identified 6 nutrient patterns, 1 (nutrient pattern 6, characterized by high intakes of Ω-3 and Ω-6 polyunsaturated fatty acids and vitamin E) was positively associated with FA. Those with the highest tertile of nutrient pattern 6 score had a mean of 0.01 (p = 0.01) higher FA value than those with the lowest tertile, similar to the effect of a 10-year decrease in age (b for age = -0.001, p = 0.01). FA mediated the relationship between nutrient pattern 6 and memory, language, visuospatial and speed/executive function, and mean cognitive scores. INTERPRETATION: Our study suggests that older adults consuming more polyunsaturated fatty acids and vitamin E rich foods had better white matter integrity, and that maintaining white matter microstructural integrity might be a mechanism for the beneficial role of diet on cognition. Ann Neurol 2016;79:1014-1025.

Insight from uncertainty: bootstrap-derived diffusion metrics differentially predict memory function among older adults.

Diffusion tensor imaging suffers from an intrinsic low signal-to-noise ratio. Bootstrap algorithms have been introduced to provide a non-parametric method to estimate the uncertainty of the measured diffusion parameters. To quantify the variability of the principal diffusion direction, bootstrap-derived metrics such as the cone of uncertainty have been proposed. However, bootstrap-derived metrics are not independent of the underlying diffusion profile. A higher mean diffusivity causes a smaller signal-to-noise ratio and, thus, increases the measurement uncertainty. Moreover, the goodness of the tensor model, which relies strongly on the complexity of the underlying diffusion profile, influences bootstrap-derived metrics as well. The presented simulations clearly depict the cone of uncertainty as a function of the underlying diffusion profile. Since the relationship of the cone of uncertainty and common diffusion parameters, such as the mean diffusivity and the fractional anisotropy, is not linear, the cone of uncertainty has a different sensitivity. In vivo analysis of the fornix reveals the cone of uncertainty to be a predictor of memory function among older adults. No significant correlation occurs with the common diffusion parameters. The present work not only demonstrates the cone of uncertainty as a function of the actual diffusion profile, but also discloses the cone of uncertainty as a sensitive predictor of memory function. Future studies should incorporate bootstrap-derived metrics to provide more comprehensive analysis.

BootGraph: probabilistic fiber tractography using bootstrap algorithms and graph theory.

Bootstrap methods have recently been introduced to diffusion-weighted magnetic resonance imaging to estimate the measurement uncertainty of ensuing diffusion parameters directly from the acquired data without the necessity to assume a noise model. These methods have been previously combined with deterministic streamline tractography algorithms to allow for the assessment of connection probabilities in the human brain. Thereby, the local noise induced disturbance in the diffusion data is accumulated additively due to the incremental progression of streamline tractography algorithms. Graph based approaches have been proposed to overcome this drawback of streamline techniques. For this reason, the bootstrap method is in the present work incorporated into a graph setup to derive a new probabilistic fiber tractography method, called BootGraph. The acquired data set is thereby converted into a weighted, undirected graph by defining a vertex in each voxel and edges between adjacent vertices. By means of the cone of uncertainty, which is derived using the wild bootstrap, a weight is thereafter assigned to each edge. Two path finding algorithms are subsequently applied to derive connection probabilities. While the first algorithm is based on the shortest path approach, the second algorithm takes all existing paths between two vertices into consideration. Tracking results are compared to an established algorithm based on the bootstrap method in combination with streamline fiber tractography and to another graph based algorithm. The BootGraph shows a very good performance in crossing situations with respect to false negatives and permits incorporating additional constraints, such as a curvature threshold. By inheriting the advantages of the bootstrap method and graph theory, the BootGraph method provides a computationally efficient and flexible probabilistic tractography setup to compute connection probability maps and virtual fiber pathways without the drawbacks of streamline tractography algorithms or the assumption of a noise distribution. Moreover, the BootGraph can be applied to common DTI data sets without further modifications and shows a high repeatability. Thus, it is very well suited for longitudinal studies and meta-studies based on DTI.

In vivo precision of bootstrap algorithms applied to diffusion tensor imaging data.

PURPOSE: To determine the precision for in vivo applications of model and non-model-based bootstrap algorithms for estimating the measurement uncertainty of diffusion parameters derived from diffusion tensor imaging data. MATERIALS AND METHODS: Four different bootstrap methods were applied to diffusion datasets acquired during 10 repeated imaging sessions. Measurement uncertainty was derived in eight manually selected regions of interest and in the entire brain white matter and gray matter. The precision of the bootstrap methods was analyzed using coefficients of variation and intra-class correlation coefficients. Comprehensive simulations were performed to validate the results. RESULTS: All bootstrap algorithms showed similar precision which slightly varied in dependence of the selected region of interest. The averaged coefficient of variation in the selected regions of interest was 13.81%, 12.35%, and 17.93% with respect to the apparent diffusion coefficient, the fractional anisotropy value, and the cone of uncertainty, respectively. The repeated measurements showed a very high similarity with intraclass-correlation coefficients larger than 0.96. The simulations confirmed most of the in vivo findings. CONCLUSION: All investigated bootstrap methods perform with a similar, high precision in deriving the measurement uncertainty of diffusion parameters. Thus, the time-efficient model-based bootstrap approaches should be the method of choice in clinical practice.

Optimizing the functional diffusion map using Monte Carlo simulations.

PURPOSE: To optimize the diagnostic accuracy of the functional diffusion map for monitoring tumor treatment response in cancer patients. MATERIALS AND METHODS: Using Monte Carlo simulations, measurement precision of the apparent diffusion coefficient (ADC), and particularly accuracy of threshold determination from healthy reference tissue, are evaluated by investigating the repeatability limit of the ADC as a function of different degrees of diffusion weighting of the sequence. Phantom and in-vivo experiments are performed to verify and illustrate the results of the simulations. RESULTS: While diagnostic accuracy of the functional diffusion map is hardly diminished by differing values of the T(2) relaxation time in tumor and reference tissue, it is shown to be impaired by differing ADCs, resulting in erroneously determined segmentation thresholds. This problem can be addressed by decreasing the maximum b-factor and increasing the number of signal averages at the maximum b-factor or, alternatively, the number of b-factors while favoring schemes with higher b-factors. Phantom experiments confirm the results of the simulations. In-vivo data are presented to illustrate the effect of sequence optimization on the diagnostic accuracy of the functional diffusion map. CONCLUSION: The present work demonstrates that the diagnostic accuracy of the functional diffusion map can be impaired by inaccurate segmentation thresholds and derives means for its optimization that will increase the fidelity of future clinical studies.

Optimizing signal-to-noise ratio of high-resolution parallel single-shot diffusion-weighted echo-planar imaging at ultrahigh field strengths.

The potential signal-to-noise ratio (SNR) gain at ultrahigh field strengths offers the promise of higher image resolution in single-shot diffusion-weighted echo-planar imaging the challenge being reduced T(2) and T(2) * relaxation times and increased B(0) inhomogeneity which lead to geometric distortions and image blurring. These can be addressed using parallel imaging (PI) methods for which a greater range of feasible reduction factors has been predicted at ultrahigh field strengths-the tradeoff being an associated SNR loss. Using comprehensive simulations, the SNR of high-resolution diffusion-weighted echo-planar imaging in combination with spin-echo and stimulated-echo acquisition is explored at 7 T and compared to 3 T. To this end, PI performance is simulated for coil arrays with a variable number of circular coil elements. Beyond that, simulations of the point spread function are performed to investigate the actual image resolution. When higher PI reduction factors are applied at 7 T to address increased image distortions, high-resolution imaging benefits SNR-wise only at relatively low PI reduction factors. On the contrary, it features generally higher image resolutions than at 3 T due to smaller point spread functions. The SNR simulations are confirmed by phantom experiments. Finally, high-resolution in vivo images of a healthy volunteer are presented which demonstrate the feasibility of higher PI reduction factors at 7 T in practice.