Maria del C. Valdés-Hernández

Blood–Brain Barrier Permeability and Long-Term Clinical and Imaging Outcomes in Cerebral Small Vessel Disease

Background and Purpose— Increased blood–brain barrier (BBB) permeability occurs in cerebral small vessel disease. It is not known if BBB changes predate progression of small vessel disease. Methods— We followed-up patients with nondisabling lacunar or cortical stroke and BBB permeability magnetic resonance imaging after their original stroke. Approximately 3 years later, we assessed functional outcome (Oxford Handicap Score, poor outcome defined as 3–6), recurrent neurological events, and white matter hyperintensity (WMH) progression on magnetic resonance imaging. Results— Among 70 patients with mean age of 68 (SD±11) years, median time to clinical follow-up was 39 months (interquartile range, 30–45) and median Oxford Handicap Score was 2 (interquartile range, 1–3); poor functional outcome was associated with higher baseline WMH score (P<0.001) and increased basal ganglia BBB permeability (P=0.046). Among 48 patients with follow-up magnetic resonance imaging, WMH progression at follow-up was associated with baseline WMH (ANCOVA P<0.0001) and age (ANCOVA P=0.032). Conclusions— Further long-term studies to evaluate the role of BBB dysfunction in progression of small vessel disease are required in studies that are large enough to account for key prognostic influences such as baseline WMH and age.

Circulating Inflammatory Markers Are Associated With Magnetic Resonance Imaging-Visible Perivascular Spaces But Not Directly With White Matter Hyperintensities

Background and Purpose— White matter hyperintensities (WMH) and perivascular spaces (PVS) are features of small vessel disease, found jointly on MRI of older people. Inflammation is a prominent pathological feature of small vessel disease. We examined the association between inflammation, PVS, and WMH in the Lothian Birth Cohort 1936 (N=634). Methods— We measured plasma fibrinogen, C-reactive protein, and interleukin-6 and rated PVS in 3 brain regions. We measured WMH volumetrically and visually using the Fazekas scale. We derived latent variables for PVS, WMH, and Inflammation from measured PVS, WMH, and inflammation markers and modelled associations using structural equation modelling. Results— After accounting for age, sex, stroke, and vascular risk factors, PVS were significantly associated with WMH (&bgr;=0.47; P<0.0001); Inflammation was weakly but significantly associated with PVS (&bgr;=0.12; P=0.048), but not with WMH (&bgr;=0.02; P=NS). Conclusions— Circulating inflammatory markers are weakly associated with MR-visible PVS, but not directly with WMH. Longitudinal studies should examine whether visible PVS predate WMH progression and whether inflammation modulators can prevent small vessel disease.

Plasma Biomarkers of Inflammation, Endothelial Function and Hemostasis in Cerebral Small Vessel Disease.

Background: The cause of lacunar ischemic stroke, a clinical feature of cerebral small vessel disease (SVD), is largely unknown. Inflammation and endothelial dysfunction have been implicated. Plasma biomarkers could provide mechanistic insights but current data are conflicting. White matter hyperintensities (WMHs) are an important imaging biomarker of SVD. It is unknown if plasma biomarkers add predictive capacity beyond age and vascular risk factors in explaining WMH. Methods: We prospectively recruited patients presenting with non-disabling ischemic stroke, classifying them clinically and with the help of MRI as lacunar or cortical. We measured biomarkers of inflammation, endothelial dysfunction and hemostasis for >1 month after stroke and compared biomarker levels between stroke subtypes. We quantitatively calculated WMH. We used multiple linear regression analysis to model WMH as a function of age, sex, hypertension and smoking (the baseline model). We fitted exploratory models using plasma biomarkers as predictor variables to assess model improvement over baseline. Results: We recruited 125 patients. The lacunar group (n = 65) had lower tissue plasminogen activator (t-PA) levels in unadjusted (7.39 vs. 8.59 ng/ml, p = 0.029) and adjusted (p = 0.035) analyses compared with the cortical group (n = 60). There were no significant differences in the other plasma biomarkers. The results for t-PA were consistent with an updated meta-analysis, although the effect remains non-significant (standardized mean difference -0.08 (95% CI -0.25 to 0.09)). The baseline regression model explained 29% of the variance in quantitative WMH (R2 0.289). Inflammatory biomarkers showed minor improvement over baseline (R2 0.291), but the other plasma biomarkers did not improve the baseline model. Conclusion: Plasma t-PA levels appear to differ between lacunar and cortical stroke subtypes, late after stroke, independent of age, sex and vascular risk factors and may reflect endothelial dysfunction. Except for a minor additional predictive effect of inflammatory markers, plasma biomarkers do not relate to WMH severity in this small stroke population.

White matter hyperintensity and stroke lesion segmentation and differentiation using convolutional neural networks

White matter hyperintensities (WMH) are a feature of sporadic small vessel disease also frequently observed in magnetic resonance images (MRI) of healthy elderly subjects. The accurate assessment of WMH burden is of crucial importance for epidemiological studies to determine association between WMHs, cognitive and clinical data; their causes, and the effects of new treatments in randomized trials. The manual delineation of WMHs is a very tedious, costly and time consuming process, that needs to be carried out by an expert annotator (e.g. a trained image analyst or radiologist). The problem of WMH delineation is further complicated by the fact that other pathological features (i.e. stroke lesions) often also appear as hyperintense regions. Recently, several automated methods aiming to tackle the challenges of WMH segmentation have been proposed. Most of these methods have been specifically developed to segment WMH in MRI but cannot differentiate between WMHs and strokes. Other methods, capable of distinguishing between different pathologies in brain MRI, are not designed with simultaneous WMH and stroke segmentation in mind. Therefore, a task specific, reliable, fully automated method that can segment and differentiate between these two pathological manifestations on MRI has not yet been fully identified. In this work we propose to use a convolutional neural network (CNN) that is able to segment hyperintensities and differentiate between WMHs and stroke lesions. Specifically, we aim to distinguish between WMH pathologies from those caused by stroke lesions due to either cortical, large or small subcortical infarcts. The proposed fully convolutional CNN architecture, called uResNet, that comprised an analysis path, that gradually learns low and high level features, followed by a synthesis path, that gradually combines and up-samples the low and high level features into a class likelihood semantic segmentation. Quantitatively, the proposed CNN architecture is shown to outperform other well established and state-of-the-art algorithms in terms of overlap with manual expert annotations. Clinically, the extracted WMH volumes were found to correlate better with the Fazekas visual rating score than competing methods or the expert-annotated volumes. Additionally, a comparison of the associations found between clinical risk-factors and the WMH volumes generated by the proposed method, was found to be in line with the associations found with the expert-annotated volumes.

Exploratory analysis of dietary intake and brain iron accumulation detected using magnetic resonance imaging in older individuals : the Lothian Birth cohort 1936

ContextBrain Iron Deposits (IDs) are associated with neurodegenerative diseases and impaired cognitive function in later life, but their cause is unknown. Animal studies have found evidence of relationships between dietary iron, calorie and cholesterol intake and brain iron accumulation.ObjectivesTo investigate the relationship between iron, calorie, and cholesterol intake, blood indicators of iron status, and brain IDs in humans.Design, Setting and ParticipantsCohort of 1063 community-dwelling older individuals born in 1936 (mean age 72.7years, SD=0.7) with dietary information, results from blood sample analyses and brain imaging data contemporaneously in old age.MeasurementsMagnetic Resonance Imaging was used to assess regional volumes of brain IDs in basal ganglia, brainstem, white matter, thalamus, and cortex/border with the corticomedullary junction, using a fully automatic assessment procedure followed by individual checking/correction where necessary. Haemoglobin, red cell count, haematocrit, mean cell volume, ferritin and transferrin were obtained from blood samples and typical daily intake of iron, calories, and cholesterol were calculated from a validated food-frequency questionnaire.ResultsOverall, 72.8% of the sample that had valid MRI (n=676) had brain IDs. The median total volume of IDs was 40mm3, inter-quartile range (IQR)=196. Basal ganglia IDs (median=35, IQR=159.5 mm3), were found in 70.6% of the sample. IDs in the brainstem were found in 12.9% of the sample, in the cortex in 1.9%, in the white matter in 6.1% and in the thalamus in 1.0%. The median daily intake of calories was 1808.5kcal (IQR=738.5), of cholesterol was 258.5mg (IQR=126.2) and of total iron was 11.7mg (IQR=5). Iron, calorie or cholesterol intake were not directly associated with brain IDs. However, caloric intake was associated with ferritin, an iron storage protein (p=0.01).ConclusionOur results suggest that overall caloric, iron and cholesterol intake are not associated with IDs in brains of healthy older individuals but caloric intake could be associated with iron storage. Further work is required to corroborate our findings on other samples and investigate the underlying mechanisms of brain iron accumulation.

Deep Learning vs. Conventional Machine Learning: Pilot Study of WMH Segmentation in Brain MRI with Absence or Mild Vascular Pathology

In the wake of the use of deep learning algorithms in medical image analysis, we compared performance of deep learning algorithms, namely the deep Boltzmann machine (DBM), convolutional encoder network (CEN) and patch-wise convolutional neural network (patch-CNN), with two conventional machine learning schemes: Support vector machine (SVM) and random forest (RF), for white matter hyperintensities (WMH) segmentation on brain MRI with mild or no vascular pathology. We also compared all these approaches with a method in the Lesion Segmentation Tool public toolbox named lesion growth algorithm (LGA). We used a dataset comprised of 60 MRI data from 20 subjects in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, each scanned once every year during three consecutive years. Spatial agreement score, receiver operating characteristic and precision-recall performance curves, volume disagreement score, agreement with intra-/inter-observer reliability measurements and visual evaluation were used to find the best configuration of each learning algorithm for WMH segmentation. By using optimum threshold values for the probabilistic output from each algorithm to produce binary masks of WMH, we found that SVM and RF produced good results for medium to very large WMH burden but deep learning algorithms performed generally better than conventional ones in most evaluations.

A critical analysis of neuroanatomical software protocols reveals clinically relevant differences in parcellation schemes

ABSTRACT A high replicability in region‐of‐interest (ROI) morphometric or ROI‐based connectivity analyses is essential for such methods to provide biomarkers of good health or disease. In this article, we focus on package design, and more specifically on cortical parcellation protocols, for novel insight into their contribution to inter‐package differences. A critical analysis of cortical parcellation protocols from FreeSurfer, BrainSuite, BrainVISA and BrainGyrusMapping revealed major limitations. Details of reference populations are generally missing, cortical variability is not always explicitly accounted for and, more importantly, definition of gyral borders can be inconsistent. We recommend that in the package selection process end users incorporate protocol suitability for the ROIs under investigation, with these particular points in mind, as inter‐package differences are likely to be significant and the source of incompatibility between studies’ results. HIGHLIGHTSInvestigation of 4 packages’ parcellation protocols and their implications.Variablity and lack of protocol explicitness regarding population atlas details.Variablity and lack of protocol explicitness regarding landmarks and gyral borders.Inconsistency in handling cortical variability.Large variations between software protocols underpinning lack of reproducibility.

Evaluation of Four Supervised Learning Schemes in White Matter Hyperintensities Segmentation in Absence or Mild Presence of Vascular Pathology

We investigated the performance of four popular supervised learning algorithms in medical image analysis for white matter hyperintensities segmentation in brain MRI with mild or no vascular pathology. The algorithms evaluated in this study are support vector machine (SVM), random forest (RF), deep Boltzmann machine (DBM) and convolution encoder network (CEN). We compared these algorithms with two methods in the Lesion Segmentation Tool (LST) public toolbox which are lesion growth algorithm (LGA) and lesion prediction algorithm (LPA). We used a dataset comprised of 60 MRI data from 20 subjects from the ADNI database, each scanned once in three consecutive years. In this study, CEN produced the best Dice similarity coefficient (DSC): mean value 0.44. All algorithms struggled to produce good DSC due to the very small WMH burden (i.e., smaller than 1,500 \(\text {mm}^3\)). LST-LGA, LST-LPA, SVM, RF and DBM produced mean DSC scores ranging from 0.17 to 0.34.

Automatic Irregular Texture Detection in Brain MRI Without Human Supervision.

We propose a novel approach named one-time sampling irregularity age map (OTS-IAM) to detect any irregular texture in FLAIR brain MRI without any human supervision or interaction. In this study, we show that OTS-IAM is able to detect FLAIR’s brain tissue irregularities (i.e. hyperintensities) without any manual labelling. One-time sampling (OTS) scheme is proposed in this study to speed up the computation. The proposed OTS-IAM implementation on GPU successfully speeds up IAM’s computation by more than 17 times. We compared the performance of OTS-IAM with two unsupervised methods for hyperintensities’ detection; the original IAM and the Lesion Growth Algorithm from public toolbox Lesion Segmentation Toolbox (LST-LGA), and two conventional supervised machine learning algorithms; support vector machine (SVM) and random forest (RF). Furthermore, we also compared OTS-IAM’s performance with three supervised deep neural networks algorithms; Deep Boltzmann machine (DBM), convolutional encoder network (CEN) and 2D convolutional neural network (2D Patch-CNN). Based on our experiments, OTS-IAM outperformed LST-LGA, SVM, RF and DBM while it was on par with CEN.

Hierarchical Complexity of the Adult Human Structural Connectome

The structural network of the human brain has a rich topology which many have sought to characterise using standard network science measures and concepts. However, this characterisation remains incomplete and the non-obvious features of this topology have confounded attempts to model it constructively. This calls for new perspectives. Hierarchical complexity is an emerging paradigm of complex network topology based on the observation that complex systems are composed of hierarchies within which the roles of hierarchically equivalent nodes display highly variable connectivity patterns. Here we test the hierarchical complexity of the human structural connectomes of a group of seventy-nine healthy adults. Binary connectomes are found to be more hierarchically complex than three null models — random graphs, random geometric graphs, and edge-randomised connectomes. This presents important new insights into the structure of the human brain, indicating a rich variety of connectivity patterns within hierarchically equivalent nodes. That random models fail to show such behaviour suggests that the generative mechanisms of brain structure may even insist on such dissimilarity. This also provides the strongest evidence to date in support of the hierarchical complexity paradigm of complex brain networks — both ordered and random systems are inherently more predictable. Dividing the connectomes into four tiers based on degree magnitudes indicates that the most complex nodes are neither those with the highest nor lowest degrees but are instead found in the third and second tiers. Spatial mapping of the brain regions in each hierarchical tier reveals consistency with the current anatomical, functional and neuropsychological knowledge of the human brain. The most complex tier (tier 3) involves regions believed to bridge high-order cognitive (tier 1) and low-order sensorimotor processing (tier 2), revealing a strikingly large diversity of connectivity patterns elicited in the integration of these processes.