Improved Automatic Segmentation of White Matter Hyperintensities in MRI Based on Multilevel Lesion Features.

Estimating white matter hyperintensities volume in individuals with stroke using T1-weighted images

White matter hyperintensities (WMHs) are regions of abnormally high intensity on T2-weighted or fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI). Accurate and reproducible automatic segmentation of WMHs is important since WMHs are often seen in the elderly and are associated with various geriatric and psychiatric disorders. We developed a fully automated monospectral segmentation method for WMHs using FLAIR MRIs. Through this method, we introduce an optimal threshold intensity (I O ) for segmenting WMHs, which varies with WMHs volume (V WMH), and we establish the I O -V WMH relationship. Our method showed accurate validations in volumetric and spatial agreements of automatically segmented WMHs compared with manually segmented WMHs for 32 confirmatory images. Bland-Altman values of volumetric agreement were 0.96 ± 8.311ml (bias and 95% confidence interval), and the similarity index of spatial agreement was 0.762 ± 0.127 (mean ± standard deviation). Furthermore, similar validation accuracies were obtained in the images acquired from different scanners. The proposed segmentation method uses only FLAIR MRIs, has the potential to be accurate with images obtained from different scanners, and can be implemented with a fully automated procedure. In our study, validation results were obtained with FLAIR MRIs from only two scanner types. The design of the method may allow its use in large multicenter studies with correct efficiency.

PurposeQuantitative assessment of white matter hyperintensities (WMH) on structural Magnetic Resonance Imaging (MRI) is challenging. It is important to harmonise results from different software tools considering not only the volume but also the signal intensity. Here we propose and evaluate a metric of white matter (WM) damage that addresses this need.MethodsWe obtained WMH and normal-appearing white matter (NAWM) volumes from brain structural MRI from community dwelling older individuals and stroke patients enrolled in three different studies, using two automatic methods followed by manual editing by two to four observers blind to each other. We calculated the average intensity values on brain structural fluid-attenuation inversion recovery (FLAIR) MRI for the NAWM and WMH. The white matter damage metric is calculated as the proportion of WMH in brain tissue weighted by the relative image contrast of the WMH-to-NAWM. The new metric was evaluated using tissue microstructure parameters and visual ratings of small vessel disease burden and WMH: Fazekas score for WMH burden and Prins scale for WMH change.ResultsThe correlation between the WM damage metric and the visual rating scores (Spearman ρ > =0.74, p < 0.0001) was slightly stronger than between the latter and WMH volumes (Spearman ρ > =0.72, p < 0.0001). The repeatability of the WM damage metric was better than WM volume (average median difference between measurements 3.26% (IQR 2.76%) and 5.88% (IQR 5.32%) respectively). The follow-up WM damage was highly related to total Prins score even when adjusted for baseline WM damage (ANCOVA, p < 0.0001), which was not always the case for WMH volume, as total Prins was highly associated with the change in the intense WMH volume (p = 0.0079, increase of 4.42 ml per unit change in total Prins, 95%CI [1.17 7.67]), but not with the change in less-intense, subtle WMH, which determined the volumetric change.ConclusionThe new metric is practical and simple to calculate. It is robust to variations in image processing methods and scanning protocols, and sensitive to subtle and severe white matter damage.

Objectives: To evaluate white matter hyperintensities (WMH) quantification reproducibility from multiple aspects of view and examine the effects of scan–rescan procedure, types of scanner, imaging protocols, scanner software upgrade, and automatic segmentation tools on WMH quantification results using magnetic resonance imaging (MRI).Methods: Six post-stroke subjects (4 males; mean age = 62.8, range = 58–72 years) were scanned and rescanned with both 3D T1-weighted, 2D and 3D T2-weighted fluid-attenuated inversion recovery (T2-FLAIR) MRI across four different MRI scanners within 12 h. Two automated WMH segmentation and quantification tools were used to measure WMH volume based on each MR scan. Robustness was assessed using the coefficient of variation (CV), Dice similarity coefficient (DSC), and intra-class correlation (ICC).Results: Experimental results show that the best reproducibility was achieved by using 3D T2-FLAIR MRI under intra-scanner setting with CV ranging from 2.69 to 2.97%, while the largest variability resulted from comparing WMH volumes measured based on 2D T2-FLAIR MRI with those of 3D T2-FLAIR MRI, with CV values in the range of 15.62%–29.33%. The WMH quantification variability based on 2D MRIs is larger than 3D MRIs due to their large slice thickness. The DSC of WMH segmentation labels between intra-scanner MRIs ranges from 0.63 to 0.77, while that for inter-scanner MRIs is in the range of 0.63–0.65. In addition to image acquisition, the choice of automatic WMH segmentation tool also has a large impact on WMH quantification.Conclusion: WMH reproducibility is one of the primary issues to be considered in multicenter and longitudinal studies. The study provides solid guidance in assisting multicenter and longitudinal study design to achieve meaningful results with enough power.

White matter hyperintensities negatively impact white matter structure and relate to cognitive decline in aging. Diffusion tensor imaging detects changes to white matter microstructure, both within the white matter hyperintensity and extending into surrounding (perilesional) normal-appearing white matter. However, diffusion tensor imaging markers are not specific to tissue components, complicating the interpretation of previous microstructural findings. Myelin water imaging is a novel imaging technique that provides specific markers of myelin content (myelin water fraction) and interstitial fluid (geometric mean T2). Here we combined diffusion tensor imaging and myelin water imaging to examine tissue characteristics in white matter hyperintensities and perilesional white matter in 80 individuals (47 older adults and 33 individuals with chronic stroke). To measure perilesional normal-appearing white matter, white matter hyperintensity masks were dilated in 2 mm segments up to 10 mm in distance from the white matter hyperintensity. Fractional anisotropy, mean diffusivity, myelin water fraction, and geometric mean T2 were extracted from white matter hyperintensities and perilesional white matter. We observed a spatial gradient of higher mean diffusivity and geometric mean T2, and lower fractional anisotropy, in the white matter hyperintensity and perilesional white matter. In the chronic stroke group, myelin water fraction was reduced in the white matter hyperintensity but did not show a spatial gradient in perilesional white matter. Across the entire sample, white matter metrics within the white matter hyperintensity related to whole-brain white matter hyperintensity volume; with increasing white matter hyperintensity volume there was increased mean diffusivity and geometric mean T2, and decreased myelin water fraction in the white matter hyperintensity. Normal-appearing white matter adjacent to white matter hyperintensities exhibits characteristics of a transitional stage between healthy white matter and white matter hyperintensities. This effect was observed in markers sensitive to interstitial fluid, but not in myelin water fraction, the specific marker of myelin concentration. Within the white matter hyperintensity, interstitial fluid was higher and myelin concentration was lower in individuals with more severe cerebrovascular disease. Our data suggests white matter hyperintensities have penumbra-like effects in perilesional white matter that specifically reflect increased interstitial fluid, with no changes to myelin concentration. In contrast, within the white matter hyperintensity there are varying levels of demyelination, which vary based on the severity of cerebrovascular disease. Diffusion tensor imaging and myelin imaging may be useful clinical markers to predict white matter hyperintensity formation, and to stage neuronal damage within white matter hyperintensities.

To the Editor: The recent article by DeCarli et al1 addresses a somewhat neglected aspect of white matter hyperintensities (WMH), the significance of their anatomical location. The authors argue that the commonly accepted categorization into deep (DWMH) and periventricular (PVWMH) WMH is arbitrary, because the 2 are very highly correlated, and a spatial analysis does not reveal distinct populations. We think that this conclusion is premature, because the categorization depends on a number of factors. The first limitation of their analysis is that they examined individuals in their 70s who presented to a specialty clinic, suggesting that the white matter lesions in their sample were at an advanced stage. If an analogy is drawn from cerebral atrophy in dementia, regional differences in atrophy that are present in the different subtypes of dementia become less prominent in the later stages. In our study of WMH in middle age (60 to 64 years), the correlation of DWMH and PVWMH was much lower ( r =0.621; …

Background: White Matter Hyperintensities (WMH) are thought to represent end-stage white matter injury. Increasing WMH burden is associated with greater risk of stroke, infarct growth, and poor outcomes after stroke. Understanding the microstructural changes of normal appearing white matter (NAWM) that predate WMH development is therefore clinically relevant as it could identify targets for therapeutic intervention to reduce stroke risk and improve post-stroke outcomes. Diffusion tensor imaging-based axial (AD) and radial diffusivity (RD) can assess white matter axonal integrity and myelin status, respectively. In a cohort of patients with acute ischemic stroke (AIS), we measured NAWM and WMH AD and RD to elucidate the white matter microstructural changes associated with WMH burden. Methods: Brain MRI with diffusion tensor imaging sequences was acquired within 48 hours of admission on consecutive AIS patients. WMH volume (WMHv) was measured in a semi-automated manner. Median fractional anisotropy, mean diffusivity, RD, and AD values were calculated within NAWM and WMH in the hemisphere contralateral to the acute lesion. Linear regression analysis was performed to evaluate predictors of WMH. Level of significance was set at P &lt; 0.05 for all analyses. Results: In 319 AIS patients, mean age was 67 +/- 15.9 years. Median WMHv was 6.19 cm3 (IQR 3.0-12.6 cm 3 ). Mean AD was significantly increased in WMH compared to NAWM (WMH: 1.29 vs. NAWM: 1.16 x 10 -3 , mm 2 /s; P &lt; 0.0001). In multivariable linear regression, age (β = 0.18, P = 0.017) and NAWM AD (β = 42.3, P = 0.012) were independent predictors of WMHv. Diffusivity anisotropy metrics of WMH were not associated with WMHv. Conclusions: In patients with AIS, NAWM AD is an independent predictor of WMHv. Our findings suggest that loss of NAWM axonal integrity contributes to WMH development.

Associations between accelerometer-derived sedentary behavior and physical activity with white matter hyperintensities in middle-aged to older adults.

To assess the effect of weightlessness and the respective roles of CSF and vascular fluid on changes in white matter hyperintensity (WMH) burden in astronauts. We analyzed prespaceflight and postspaceflight brain MRI scans from 17 astronauts, 10 who flew a long-duration mission on the International Space Station (ISS) and 7 who flew a short-duration mission on the Space Shuttle. Automated analysis methods were used to determine preflight to postflight changes in periventricular and deep WMH, CSF, and brain tissue volumes in fluid-attenuated inversion recovery and high-resolution 3-dimensional T1-weighted imaging. Differences between cohorts and associations between individual measures were assessed. The short-term reversibility of the identified preflight to postflight changes was tested in a subcohort of 5 long-duration astronauts who had a second postflight MRI scan 1 month after the first postflight scan. Significant preflight to postflight changes were measured only in the long-duration cohort and included only the periventricular WMH and ventricular CSF volumes. Changes in deep WMH and brain tissue volumes were not significant in either cohort. The increase in periventricular WMH volume was significantly associated with an increase in ventricular CSF volume (ρ = 0.63, p = 0.008). A partial reversal of these increases was observed in the long-duration subcohort with a 1-month follow-up scan. Long-duration exposure to microgravity is associated with an increase in periventricular WMH in astronauts. This increase was linked to an increase in ventricular CSF volume documented in ISS astronauts. There was no associated change in or abnormal levels of WMH volumes in deep white matter as reported in U-2 high-altitude pilots.

Accurate and reliable detection of white matter hyperintensities and their volume quantification can provide valuable clinical information to assess neurologic disease progression. In this work, a stacked generalization ensemble of orthogonal 3D convolutional neural networks, StackGen-Net, is explored for improving automated detection of white matter hyperintensities in 3D T2-FLAIR images. Individual convolutional neural networks in StackGen-Net were trained on 2.5D patches from orthogonal reformatting of 3D-FLAIR (n = 21) to yield white matter hyperintensity posteriors. A meta convolutional neural network was trained to learn the functional mapping from orthogonal white matter hyperintensity posteriors to the final white matter hyperintensity prediction. The impact of training data and architecture choices on white matter hyperintensity segmentation performance was systematically evaluated on a test cohort (n = 9). The segmentation performance of StackGen-Net was compared with state-of-the-art convolutional neural network techniques on an independent test cohort from the Alzheimer's Disease Neuroimaging Initiative-3 (n = 20). StackGen-Net outperformed individual convolutional neural networks in the ensemble and their combination using averaging or majority voting. In a comparison with state-of-the-art white matter hyperintensity segmentation techniques, StackGen-Net achieved a significantly higher Dice score (0.76 [SD, 0.08], F1-lesion (0.74 [SD, 0.13]), and area under precision-recall curve (0.84 [SD, 0.09]), and the lowest absolute volume difference (13.3% [SD, 9.1%]). StackGen-Net performance in Dice scores (median = 0.74) did not significantly differ (P = .22) from interobserver (median = 0.73) variability between 2 experienced neuroradiologists. We found no significant difference (P = .15) in white matter hyperintensity lesion volumes from StackGen-Net predictions and ground truth annotations. A stacked generalization of convolutional neural networks, utilizing multiplanar lesion information using 2.5D spatial context, greatly improved the segmentation performance of StackGen-Net compared with traditional ensemble techniques and some state-of-the-art deep learning models for 3D-FLAIR.

Structural brain abnormalities are associated with late-life major depression, with numerous studies reporting increased white matter hyperintensities (WMH) and reduced cortical/subcortical grey matter volumes. There is strong evidence linking vascular disease to WMH, but limited evidence on its association with subcortical volumes. To investigate the relationship of orthostatic blood pressure changes to WMH and subcortical grey matter volumes in late-life depression. Thirty-eight people with depression and a similarly aged comparison group (n = 30) underwent fluid attenuated inversion recovery (FLAIR) and T(1)-weighted magnetic resonance imaging as well as systematic orthostatic blood pressure assessments. Volumetric estimates of WMH and subcortical grey matter were obtained for each participant and the relationship to blood pressure drop on active stand was examined. An association between orthostatic systolic blood pressure drop and WMH volumes in temporal and parietal regions was found in the depression group (age-corrected partial correlation r' = 0.31-0.35, P<0.05). Subcortical volumes were not related to blood pressure changes or WMH volumes in either group. We found evidence for an association between the degree of orthostatic systolic blood pressure drop and WMH volume in the depression group. Since blood pressure drops lead to WMH in animals our findings suggest systolic blood pressure drops may be a factor contributing to these lesions in late-life depression.

In cognitively normal (CN) elderly individuals, white matter hyperintensities (WMH) are commonly viewed as a marker of cerebral small vessel disease (SVD). SVD is due to exposure to systemic vascular injury processes associated with highly prevalent vascular risk factors (VRFs) such as hypertension, high cholesterol, and diabetes. However, cerebral amyloid accumulation is also prevalent in this population and is associated with WMH accrual. Therefore, we examined the independent associations of amyloid burden and VRFs with WMH burden in CN elderly individuals with low to moderate vascular risk. Participants (n = 150) in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) received fluid attenuated inversion recovery (FLAIR) MRI at study entry. Total WMH volume was calculated from FLAIR images co-registered with structural MRI. Amyloid burden was determined by cerebrospinal fluid Aβ1-42 levels. Clinical histories of VRFs, as well as current measurements of vascular status, were recorded during a baseline clinical evaluation. We tested ridge regression models for independent associations and interactions of elevated blood pressure (BP) and amyloid to total WMH volume. We found that greater amyloid burden and a clinical history of hypertension were independently associated with greater WMH volume. In addition, elevated BP modified the association between amyloid and WMH, such that those with either current or past evidence of elevated BP had greater WMH volumes at a given burden of amyloid. These findings are consistent with the hypothesis that cerebral amyloid accumulation and VRFs are independently associated with clinically latent white matter damage represented by WMHs. The potential contribution of amyloid to WMHs should be further explored, even among elderly individuals without cognitive impairment and with limited VRF exposure.

We present the application of limited one-time sampling irregularity map (LOTS-IM): a fully automatic unsupervised approach to extract brain tissue irregularities in magnetic resonance images (MRI), for quantitatively assessing white matter hyperintensities (WMH) of presumed vascular origin, and multiple sclerosis (MS) lesions and their progression. LOTS-IM generates an irregularity map (IM) that represents all voxels as irregularity values with respect to the ones considered ”normal”. Unlike probability values, IM represents both regular and irregular regions in the brain based on the original MRI's texture information. We evaluated and compared the use of IM for WMH and MS lesions segmentation on T2-FLAIR MRI with the state-of-the-art unsupervised lesions’ segmentation method, Lesion Growth Algorithm from the public toolbox Lesion Segmentation Toolbox (LST-LGA), with several well established conventional supervised machine learning schemes and with state-of-the-art supervised deep learning methods for WMH segmentation. In our experiments, LOTS-IM outperformed unsupervised method LST-LGA on WMH segmentation, both in performance and processing speed, thanks to the limited one-time sampling scheme and its implementation on GPU. Our method also outperformed supervised conventional machine learning algorithms (i.e., support vector machine (SVM) and random forest (RF)) and deep learning algorithms (i.e., deep Boltzmann machine (DBM) and convolutional encoder network (CEN)), while yielding comparable results to the convolutional neural network schemes that rank top of the algorithms developed up to date for this purpose (i.e., UResNet and UNet). LOTS-IM also performed well on MS lesions segmentation, performing similar to LST-LGA. On the other hand, the high sensitivity of IM on depicting signal change deems suitable for assessing MS progression, although care must be taken with signal changes not reflective of a true pathology.

BackgroundRecent genetic association studies suggested the polygenic contribution to white matter hyperintensities (WMH). The aim of this study was to provide a genetic characterization of WMH in middle‐aged cognitively unimpaired individuals at higher risk of Alzheimer’s disease (AD) by investigating whether the genetic predisposition to specific complex diseases is associated with WMH and whether these effects are sex‐specific and dependent on core AD CSF biomarker levels.MethodA total of 497 cognitively unimpaired middle‐aged participants at risk of AD were included (63.2% of women). A subgroup of 288 individuals also had available information on CSF biomarkers measured using the Elecsys ® and the exploratory Roche NeuroToolKit immunoassays. We classified individuals in groups defined by their biomarker profile (A/T framework). Genetic predisposition was calculated through polygenic risk scores (PRS), estimated from the most recent genome‐wide association meta‐analyses for cerebrovascular/vascular, neurodegenerative, and AD‐related biomarkers. WMH volumes were quantified through fluid‐attenuated inversion recovery (FLAIR) images obtained from magnetic resonance scans. We used non‐parametric Spearman’s rank tests to assess the association between global WMH volumes and PRSs (nominal p‐value &lt; 0.05). Models stratified by sex and A/T group were assessed to discern WMH sex‐specific genetic differences and to discriminate genetic profiles associated with WMH in the earliest stages of the Alzheimer’s continuum. ResultGenetic predisposition to small vessel disease was positively correlated with volumes of WMH only in women [Figure 1]. Furthermore, only for A+T‐ individuals, we found that genetic predisposition to higher CSF interleukin 6 levels and cerebral infarction was positively correlated with WMH volumes. Moreover, genetic predisposition to abnormal CSF p‐tau levels and AD was negatively correlated with volumes of WMH. We did not find significant results neither in A+T+ nor in A‐T‐ individuals.ConclusionSex differences were found suggesting better genetic predictability of cerebrovascular‐PRSs to assess WMH in women. Results by A/T group suggested that asymptomatic individuals that are amyloid‐β‐positive but tau‐negative (A+T‐) have several differential genetic patterns associated with WMH that were not significant in the other groups. Overall, these findings may lead to sex differences and different biological mechanisms affecting WMH in the Alzheimer's continuum.

Cerebral small vessel disease is the most common cause of vascular cognitive impairment. It typically manifests with lacunar infarcts and ischaemic white matter lesions. However, little is known about how these lesions relate to the cognitive symptoms. Previous studies have found a poor correlation between the burden of ischaemic lesions and cognitive symptoms, thus leaving much of the variance in cognitive performance unexplained. The objective of the current study was to investigate the relationship between the location of subcortical ischaemic lesions and cognitive symptoms in small vessel disease. We applied a voxel-based lesion-symptom mapping approach to data from 215 patients with CADASIL, a genetically defined small vessel disease with mutations in the NOTCH3 gene. All patients were examined by magnetic resonance imaging and comprehensive neuropsychological testing. Lacunar lesions and white matter lesions were segmented on three-dimensional T(1) and fluid-attenuated inversion recovery sequences, respectively. One hundred and forty-five subjects had a total of 854 lacunar lesions (range 1-13 per individual). The normalized volume of white matter hyperintensities ranged from 0.0425% to 21.5% of the intracranial cavity. Significant clusters for cognitive performance were detected for both lacunar lesions and white matter hyperintensities. The most prominent results were obtained on a compound score for processing speed, the predominantly affected cognitive domain in this group of patients. Strategic locations included the anterior parts of the thalamus, the genu and anterior limb of the internal capsule, the anterior corona radiata and the genu of the corpus callosum. By combining the lesion-symptom mapping data with information from a probabilistic white matter atlas we found that the majority of the processing speed clusters projected on the anterior thalamic radiation and the forceps minor. In multivariate models that included demographic parameters, brain atrophy and the volume of ischaemic lesions, regional volumes of lacunar lesions and white matter hyperintensities in the anterior thalamic radiation predicted performance in processing speed tasks, whereas there was no independent contribution of the global volume of ischaemic lesions. These observations emphasize the importance of lesion location for both lacunar and ischaemic white matter lesions. Our findings further highlight the anterior thalamic radiation as a major anatomical structure impacting on processing speed. Together these findings provide strong support for a central role of frontal-subcortical circuits in cerebral small vessel disease and vascular cognitive impairment.