Validation of a Regression Technique for Segmentation of White Matter Hyperintensities in Alzheimer's Disease

Ea-GANs: Edge-Aware Generative Adversarial Networks for Cross-Modality MR Image Synthesis

Magnetic resonance (MR) imaging is a widely used medical imaging protocol that can be configured to provide different contrasts between the tissues in human body. By setting different scanning parameters, each MR imaging modality reflects the unique visual characteristic of scanned body part, benefiting the subsequent analysis from multiple perspectives. To utilize the complementary information from multiple imaging modalities, cross-modality MR image synthesis has aroused increasing research interest recently. However, most existing methods only focus on minimizing pixel/voxel-wise intensity difference but ignore the textural details of image content structure, which affects the quality of synthesized images. In this paper, we propose edge-aware generative adversarial networks (Ea-GANs) for cross-modality MR image synthesis. Specifically, we integrate edge information, which reflects the textural structure of image content and depicts the boundaries of different objects in images, to reduce this gap. Corresponding to different learning strategies, two frameworks are proposed, i.e., a generator-induced Ea-GAN (gEa-GAN) and a discriminator-induced Ea-GAN (dEa-GAN). The gEa-GAN incorporates the edge information via its generator, while the dEa-GAN further does this from both the generator and the discriminator so that the edge similarity is also adversarially learned. In addition, the proposed Ea-GANs are 3D-based and utilize hierarchical features to capture contextual information. The experimental results demonstrate that the proposed Ea-GANs, especially the dEa-GAN, outperform multiple state-of-the-art methods for cross-modality MR image synthesis in both qualitative and quantitative measures. Moreover, the dEa-GAN also shows excellent generality to generic image synthesis tasks on benchmark datasets about facades, maps, and cityscapes.

Cognitive aging is not created equally: differentiating unique cognitive phenotypes in "normal" adults

Age-related cognitive decline is a public health problem but highly diverse and difficult to predict. We captured nonoverlapping cognitive phenotypes in high-functioning adults and identified baseline factors differentiating trajectories. Three hundred fourteen functionally normal adults (M = 69 y) completed 2+ visits. Participants with sample-based longitudinal slopes in memory or processing speed less than -1 SD were classified as "declining" on that measure; 29 and 50 individuals had slopes less than -1 SD on processing speed or memory, respectively; 2.5% met criteria for both, who were excluded. At baseline, speed decliners demonstrated greater age, inflammation, and cognitive complaints compared with speed-stable adults; memory decliners were more likely to be male and had lower depressive symptoms, gray matter volumes, and white matter hyperintensities compared with memory-stable adults. Baseline speed, TNFα, and cognitive complaints accurately classified 96.3% of future speed decliners; baseline memory, sex, precuneal volume, and white matter hyperintensities accurately classified 88.5% of future memory decliners. There are discrete cognitive aging phenotypes reflecting nonoverlapping vulnerabilities in high-functioning adults. Early markers can predict cognition even within the "normal" spectrum and underscore therapeutic targets.

Review of diffusion MRI studies in chronic white matter diseases

Diffusion MRI studies characterizing the changes in white matter (WM) due to vascular cognitive impairment, which includes all forms of small vessel disease are reviewed. We reviewed the usefulness of diffusion methods in discriminating the affected WM regions and its relation to cognitive impairment. These studies were categorized based on the diffusion MRI techniques used. The most common method was the diffusion tensor imaging, whereas other methods included diffusion weighted imaging, diffusion kurtosis imaging, intravoxel incoherent motion, and studies based on diffusion tractography. The diffusion measures showed correlation with cognitive scores and disease progression, with mean diffusivity being the most robust parameter. Future studies should focus on incorporating multi-compartment and higher order diffusion models, which can handle the presence of multiple and crossing fibers inside a voxel.

White matter in different regions evolves differently during progression to dementia

White matter hyperintensities (WMHs) are common in individuals with mild cognitive impairment (MCI) and Alzheimer's disease. Patients with MCI with high WMH volumes are known to have an increased chance of conversion to Alzheimer's disease compared with those without WMHs. In this article, we assess the differences between patients with MCI that remain stable (N = 413) and those that progress to dementia (N = 178) in terms of WMH volume (as a surrogate of amount of tissue damage) and T1-weighted (T1w) image hypointensity (as a surrogate of severity of tissue damage) in periventricular, deep, and juxtacortical brain regions. Together, lesion volume and T1w hypointensity are used as a surrogate of vascular disease burden. Our results show a significantly greater increase of all regional WMH volumes in the MCI population that converts to dementia (p < 0.001). T1w hypointensity for the juxtacortical WMHs was significantly lower in the converter group (p < 0.0001) and was not affected by age. Conversely, T1w hypointensity in other regions showed a significant decrease with age (p < 0.0001). Within the converters, Time2Conversion was associated with both WMH volume and T1w hypointensity (p < 0.0001), and conversion to dementia was significantly associated with decreased intensity (and not volume) of periventricular and juxtacortical WMHs (p < 0.001). These changes differ according to the WM region, suggesting that different mechanisms affect the juxtacortical area in comparison to deep and periventricular regions in the process of conversion to dementia.

White matter hyperintensities are linked to future cognitive decline in de novo Parkinson's disease patients

White Matter Hyperintensities (WMHs) are associated with cognitive decline in aging and Alzheimer's disease. However, the pathogenesis of cognitive decline in Parkinson's disease (PD) is not as clearly related to vascular causes, and therefore the role of WMHs as a marker of small-vessel disease (SVD) in PD is less clear. Currently, SVD in PD is assessed and treated independently of the disease. However, if WMH as the major MRI sign of SVD has a higher impact on cognitive decline in PD patients than in healthy controls, vascular pathology needs to be assessed and treated with a higher priority in this population. Here we investigate whether the presence of WMHs leads to increased cognitive decline in de novo PD, and if these effects relate to cortical atrophy. WMHs and cortical thickness were measured in de novo PD patients and age-matched controls (N_PD = 365, N_Control = 174) from Parkinson's Progression Markers Initiative (PPMI) to study the relationship between baseline WMHs, future cognitive decline (follow-up: 4.09 ± 1.14 years) and cortical atrophy (follow-up: 1.05 ± 0.10 years). PD subjects with high baseline WMH loads had significantly greater cognitive decline than i) PD subjects with low WMH load, and ii) control subjects with high WMH load. Furthermore, in PD subjects, high WMH load resulted in more cortical thinning in the right frontal lobe. Theses results show that the presence of WMHs in de novo PD patients predicts greater future cognitive decline and cortical atrophy than in normal aging.

Validation of T1w-based segmentations of white matter hyperintensity volumes in large-scale datasets of aging

INTRODUCTION

Fluid-attenuated Inversion Recovery (FLAIR) and dual T2w and proton density (PD) magnetic resonance images (MRIs) are considered to be the optimum sequences for detecting white matter hyperintensities (WMHs) in aging and Alzheimer's disease populations. However, many existing large multisite studies forgo their acquisition in favor of other MRI sequences due to economic and time constraints.

METHODS

In this article, we have investigated whether FLAIR and T2w/PD sequences are necessary to detect WMHs in Alzheimer's and aging studies, compared to using only T1w images. Using a previously validated automated tool based on a Random Forests classifier, WMHs were segmented for the baseline visits of subjects from ADC, ADNI1, and ADNI2/GO studies with and without T2w/PD and FLAIR information. The obtained WMH loads (WMHLs) in different lobes were then correlated with manually segmented WMHLs, each other, age, cognitive, and clinical measures to assess the strength of the correlations with and without using T2w/PD and FLAIR information.

RESULTS

The WMHLs obtained from T1w-Only segmentations correlated with the manual WMHLs (ADNI1: r = .743, p < .001, ADNI2/GO: r = .904, p < .001), segmentations obtained from T1w + T2w + PD for ADNI1 (r = .888, p < .001) and T1w + FLAIR for ADNI2/GO (r = .969, p < .001), age (ADNI1: r = .391, p < .001, ADNI2/GO: r = .466, p < .001), and ADAS13 (ADNI1: r = .227, p < .001, ADNI2/GO: r = .190, p < 0.001), and NPI (ADNI1: r = .290, p < .001, ADNI2/GO: r = 0.144, p < .001), controlling for age.

CONCLUSION

Our results suggest that while T2w/PD and FLAIR provide more accurate estimates of the true WMHLs, T1w-Only segmentations can still provide estimates that hold strong correlations with the actual WMHLs, age, and performance on various cognitive/clinical scales, giving added value to datasets where T2w/PD or FLAIR are not available.

The Longitudinal Trajectory of Default Mode Network Connectivity in Healthy Older Adults Varies As a Function of Age and Is Associated with Changes in Episodic Memory and Processing Speed

The default mode network (DMN) supports memory functioning and may be sensitive to preclinical Alzheimer's pathology. Little is known, however, about the longitudinal trajectory of this network's intrinsic functional connectivity (FC). In this study, we evaluated longitudinal FC in 111 cognitively normal older human adults (ages 49-87, 46 women/65 men), 92 of whom had at least three task-free fMRI scans (n = 353 total scans). Whole-brain FC and three DMN subnetworks were assessed: (1) within-DMN, (2) between anterior and posterior DMN, and (3) between medial temporal lobe network and posterior DMN. Linear mixed-effects models demonstrated significant baseline age × time interactions, indicating a nonlinear trajectory. There was a trend toward increasing FC between ages 50-66 and significantly accelerating declines after age 74. A similar interaction was observed for whole-brain FC. APOE status did not predict baseline connectivity or change in connectivity. After adjusting for network volume, changes in within-DMN connectivity were specifically associated with changes in episodic memory and processing speed but not working memory or executive functions. The relationship with processing speed was attenuated after covarying for white matter hyperintensities (WMH) and whole-brain FC, whereas within-DMN connectivity remained associated with memory above and beyond WMH and whole-brain FC. Whole-brain and DMN FC exhibit a nonlinear trajectory, with more rapid declines in older age and possibly increases in connectivity early in the aging process. Within-DMN connectivity is a marker of episodic memory performance even among cognitively healthy older adults.SIGNIFICANCE STATEMENT Default mode network and whole-brain connectivity, measured using task-free fMRI, changed nonlinearly as a function of age, with some suggestion of early increases in connectivity. For the first time, longitudinal changes in DMN connectivity were shown to correlate with changes in episodic memory, whereas volume changes in relevant brain regions did not. This relationship was not accounted for by white matter hyperintensities or mean whole-brain connectivity. Functional connectivity may be an early biomarker of changes in aging but should be used with caution given its nonmonotonic nature, which could complicate interpretation. Future studies investigating longitudinal network changes should consider whole-brain changes in connectivity.

Performance comparison of 10 different classification techniques in segmenting white matter hyperintensities in aging

INTRODUCTION

White matter hyperintensities (WMHs) are areas of abnormal signal on magnetic resonance images (MRIs) that characterize various types of histopathological lesions. The load and location of WMHs are important clinical measures that may indicate the presence of small vessel disease in aging and Alzheimer's disease (AD) patients. Manually segmenting WMHs is time consuming and prone to inter-rater and intra-rater variabilities. Automated tools that can accurately and robustly detect these lesions can be used to measure the vascular burden in individuals with AD or the elderly population in general. Many WMH segmentation techniques use a classifier in combination with a set of intensity and location features to segment WMHs, however, the optimal choice of classifier is unknown.

METHODS

We compare 10 different linear and nonlinear classification techniques to identify WMHs from MRI data. Each classifier is trained and optimized based on a set of features obtained from co-registered MR images containing spatial location and intensity information. We further assess the performance of the classifiers using different combinations of MRI contrast information. The performances of the different classifiers were compared on three heterogeneous multi-site datasets, including images acquired with different scanners and different scan-parameters. These included data from the ADC study from University of California Davis, the NACC database and the ADNI study. The classifiers (naïve Bayes, logistic regression, decision trees, random forests, support vector machines, k-nearest neighbors, bagging, and boosting) were evaluated using a variety of voxel-wise and volumetric similarity measures such as Dice Kappa similarity index (SI), Intra-Class Correlation (ICC), and sensitivity as well as computational burden and processing times. These investigations enable meaningful comparisons between the performances of different classifiers to determine the most suitable classifiers for segmentation of WMHs. In the spirit of open-source science, we also make available a fully automated tool for segmentation of WMHs with pre-trained classifiers for all these techniques.

RESULTS

Random Forests yielded the best performance among all classifiers with mean Dice Kappa (SI) of 0.66±0.17 and ICC=0.99 for the ADC dataset (using T1w, T2w, PD, and FLAIR scans), SI=0.72±0.10, ICC=0.93 for the NACC dataset (using T1w and FLAIR scans), SI=0.66±0.23, ICC=0.94 for ADNI1 dataset (using T1w, T2w, and PD scans) and SI=0.72±0.19, ICC=0.96 for ADNI2/GO dataset (using T1w and FLAIR scans). Not using the T2w/PD information did not change the performance of the Random Forest classifier (SI=0.66±0.17, ICC=0.99). However, not using FLAIR information in the ADC dataset significantly decreased the Dice Kappa, but the volumetric correlation did not drastically change (SI=0.47±0.21, ICC=0.95).

CONCLUSION

Our investigations showed that with appropriate features, most off-the-shelf classifiers are able to accurately detect WMHs in presence of FLAIR scan information, while Random Forests had the best performance across all datasets. However, we observed that the performances of most linear classifiers and some nonlinear classifiers drastically decline in absence of FLAIR information, with Random Forest still retaining the best performance.