Parkinson's disease (PD) is a neurodegenerative disorder characterized by a wide array of motor and non-motor symptoms. It remains unclear whether neurodegeneration in discrete loci gives rise to discrete symptoms, or whether network-wide atrophy gives rise to the unique behavioural and clinical profile associated with PD. Here we apply a data-driven strategy to isolate large-scale, multivariate associations between distributed atrophy patterns and clinical phenotypes in PD. In a sample of N = 229 de novo PD patients, we estimate disease-related atrophy using deformation based morphometry (DBM) of T1 weighted MR images. Using partial least squares (PLS), we identify a network of subcortical and cortical regions whose collective atrophy is associated with a clinical phenotype encompassing motor and non-motor features. Despite the relatively early stage of the disease in the sample, the atrophy pattern encompassed lower brainstem, substantia nigra, basal ganglia and cortical areas, consistent with the Braak hypothesis. In addition, individual variation in this putative atrophy network predicted longitudinal clinical progression in both motor and non-motor symptoms. Altogether, these results demonstrate a pleiotropic mapping between neurodegeneration and the clinical manifestations of PD, and that this mapping can be detected even in de novo patients.

The hippocampus is one of the first brain structures affected by Alzheimer's disease (AD). While many automatic methods for hippocampal segmentation exist, few studies have compared them on the same data. In this study, we compare four fully automated hippocampal segmentation methods in terms of their conformity with manual segmentation and their ability to be used as an AD biomarker in clinical settings. We also apply error correction to the four automatic segmentation methods, and complete a comprehensive validation to investigate differences between the methods. The effect size and classification performance is measured for AD versus normal control (NC) groups and for stable mild cognitive impairment (sMCI) versus progressive mild cognitive impairment (pMCI) groups. Our study shows that the nonlinear patch-based segmentation method with error correction is the most accurate automatic segmentation method and yields the most conformity with manual segmentation (κ=0.894). The largest effect size between AD versus NC and sMCI versus pMCI is produced by FreeSurfer with error correction. We further show that, using only hippocampal volume, age, and sex as features, the area under the receiver operating characteristic curve reaches up to 0.8813 for AD versus NC and 0.6451 for sMCI versus pMCI. However, the automatic segmentation methods are not significantly different in their performance.

Recently, a group of major international experts have completed a comprehensive effort to efficiently define a harmonized protocol for manual hippocampal segmentation that is optimized for Alzheimer's research (known as the EADC-ADNI Harmonized Protocol (the HarP)). This study compares the HarP with one of the widely used hippocampal segmentation protocols (Pruessner, 2000), based on a single automatic segmentation method trained separately with libraries made from each manual segmentation protocol. The automatic segmentation conformity with the corresponding manual segmentation and the ability to capture Alzheimer's disease related hippocampal atrophy on large datasets are measured to compare the manual protocols. In addition to the possibility of harmonizing different procedures of hippocampal segmentation, our results show that using the HarP, the automatic segmentation conformity with manual segmentation is also preserved (Dice's κ=0.88,κ=0.87 for Pruessner and HarP respectively (p = 0.726 for common training library)). Furthermore, the results show that the HarP can capture the Alzheimer's disease related hippocampal volume differences in large datasets. The HarP-derived segmentation shows large effect size (Cohen's d = 1.5883) in separating Alzheimer's Disease patients versus normal controls (AD:NC) and medium effect size (Cohen's d = 0.5747) in separating stable versus progressive Mild Cognitively Impaired patients (sMCI:pMCI). Furthermore, the area under the ROC curve for a LDA classifier trained based on age, sex and HarP-derived hippocampal volume is 0.8858 for AD:NC, and for 0.6677 sMCI:pMCI. These results show that the harmonized protocol-derived labels can be widely used in clinic and research, as a sensitive and accurate way of delineating the hippocampus.

Linear registration to stereotaxic space is a common first step in many automated image-processing tools for analysis of human brain MRI scans. This step is crucial for the success of the subsequent image-processing steps. Several well-established algorithms are commonly used in the field of neuroimaging for this task, but none have a 100% success rate. Manual assessment of the registration is commonly used as part of quality control. To reduce the burden of this time-consuming step, we propose Deep Automated Registration Qc (DARQ), a fully automatic quality control method based on deep learning that can replace the human rater and accurately perform quality control assessment for stereotaxic registration of T1w brain scans. In a recently published study from our group comparing linear registration methods, we used a database of 9325 MRI scans and 64476 registrations from several publicly available datasets and applied seven linear registration tools to them. In this study, the resulting images that were assessed and labeled by a human rater are used to train a deep neural network to detect cases when registration failed. We further validated the results on an independent dataset of patients with multiple sclerosis, with manual QC labels available (n=1200). In terms of agreement with a manual rater, our automated QC method was able to achieve 89% accuracy and 85% true negative rate (equivalently 15% false positive rate) in detecting scans that should pass quality control in a balanced cross-validation experiments, and 96.1% accuracy and 95.5% true negative rate (or 4.5% FPR) when evaluated in a balanced independent sample, similar to manual QC rater (test-retest accuracy of 93%). The results show that DARQ is robust, fast, accurate, and generalizable in detecting failure in linear stereotaxic registrations and can substantially reduce QC time (by a factor of 20 or more) when processing large datasets.

Linear registration to a standard space is one of the major steps in processing and analyzing magnetic resonance images (MRIs) of the brain. Here we present an overview of linear stereotaxic MRI registration and compare the performance of 5 publicly available and extensively used linear registration techniques in medical image analysis.

The maturational schedule of typical brain development is tightly constrained; deviations from it are associated with cognitive atypicalities, and are potentially predictive of developmental disorders. Previously, we have shown that the white/gray contrast at the inner border of the cortex is a good predictor of chronological age, and is sensitive to aspects of brain development that reflect cognitive performance. Here we extend that work to include the white/gray contrast at the border of subcortical structures. We show that cortical and subcortical contrast together yield better age-predictions than any non-kernel-based method based on a single image-type, and that the residuals of the improved predictions provide new insight into unevenness in cognitive performance. We demonstrate the improvement in age predictions in two large datasets: the NIH Pediatric Data, with 831 scans of typically developing individuals between 4 and 22 years of age; and the Pediatric Imaging, Neurocognition, and Genetics data, with 909 scans of individuals in a similar age-range. Assessment of the relation of the residuals of these age predictions to verbal and performance IQ revealed correlations in opposing directions, and a principal component analysis of the residuals of the model that best fit the contrast data produced components related to either performance IQ or verbal IQ. Performance IQ was associated with the first principle component, reflecting increased cortical contrast, broadly, with almost no subcortical presence; verbal IQ was associated with the second principle component, reflecting reduced contrast in the basal ganglia and increased contrast in the bilateral arcuate fasciculi.

We investigated fluctuations in brain volume throughout the day using statistical modeling of magnetic resonance imaging (MRI) from large populations. We applied fully automated image analysis software to measure the brain parenchymal fraction (BPF), defined as the ratio of the brain parenchymal volume and intracranial volume, thus accounting for variations in head size. The MRI data came from serial scans of multiple sclerosis (MS) patients in clinical trials (n=755, 3269 scans) and from subjects participating in the Alzheimer's Disease Neuroimaging Initiative (ADNI, n=834, 6114 scans). The percent change in BPF was modeled with a linear mixed effect (LME) model, and the model was applied separately to the MS and ADNI datasets. The LME model for the MS datasets included random subject effects (intercept and slope over time) and fixed effects for the time-of-day, time from the baseline scan, and trial, which accounted for trial-related effects (for example, different inclusion criteria and imaging protocol). The model for ADNI additionally included the demographics (baseline age, sex, subject type [normal, mild cognitive impairment, or Alzheimer's disease], and interaction between subject type and time from baseline). There was a statistically significant effect of time-of-day on the BPF change in MS clinical trial datasets (-0.180 per day, that is, 0.180% of intracranial volume, p=0.019) as well as the ADNI dataset (-0.438 per day, that is, 0.438% of intracranial volume, p<0.0001), showing that the brain volume is greater in the morning. Linearly correcting the BPF values with the time-of-day reduced the required sample size to detect a 25% treatment effect (80% power and 0.05 significance level) on change in brain volume from 2 time-points over a period of 1year by 2.6%. Our results have significant implications for future brain volumetric studies, suggesting that there is a potential acquisition time bias that should be randomized or statistically controlled to account for the day-to-day brain volume fluctuations.

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.

Brain volume loss measured from magnetic resonance imaging (MRI) is a marker of neurodegeneration and predictor of disability progression in MS, and is commonly used to assess drug efficacy at the group level in clinical trials. Whether measures of brain volume loss could be useful to help guide management of individual patients depends on the relative magnitude of the changes over a given interval to physiological and technical sources of variability.

Brain extraction is an important step in the analysis of brain images. The variability in brain morphology and the difference in intensity characteristics due to imaging sequences make the development of a general purpose brain extraction algorithm challenging. To address this issue, we propose a new robust method (BEaST) dedicated to produce consistent and accurate brain extraction. This method is based on nonlocal segmentation embedded in a multi-resolution framework. A library of 80 priors is semi-automatically constructed from the NIH-sponsored MRI study of normal brain development, the International Consortium for Brain Mapping, and the Alzheimer's Disease Neuroimaging Initiative databases. In testing, a mean Dice similarity coefficient of 0.9834±0.0053 was obtained when performing leave-one-out cross validation selecting only 20 priors from the library. Validation using the online Segmentation Validation Engine resulted in a top ranking position with a mean Dice coefficient of 0.9781±0.0047. Robustness of BEaST is demonstrated on all baseline ADNI data, resulting in a very low failure rate. The segmentation accuracy of the method is better than two widely used publicly available methods and recent state-of-the-art hybrid approaches. BEaST provides results comparable to a recent label fusion approach, while being 40 times faster and requiring a much smaller library of priors.

The objective of this study was to determine which 3D T(1)-weighted acquisition protocol at 3 T is best suited to voxel-based morphometry (VBM), and to characterize the sensitivity of VBM to choice of acquisition. First, image quality of three commonly used protocols, FLASH, MP-RAGE and MDEFT, was evaluated in terms of SNR, CNR, image uniformity and point spread function. These image metrics were estimated from simulations, phantom imaging and human studies. We then performed a VBM study on nine subjects scanned twice using the three protocols to evaluate differences in grey matter (GM) density and scan-rescan variability between the protocols. These results reveal the relative bias and precision of the tissue classification obtained using the different protocols. MDEFT achieved the highest CNR between white and grey matter, and the lowest GM density variability of the three sequences. Each protocol is also characterized by a distinct regional bias in GM density due to the effect of transmission field inhomogeneity on image uniformity combined with spatially variant GM T(1) values and the sequence's T(1) contrast function. The required population sample size estimates to detect a difference in GM density in longitudinal VBM studies, i.e. based only on methodological variance, were lowest for MDEFT. Although MP-RAGE requires more subjects than FLASH, its higher cortical CNR improves the accuracy of the tissue classification results, particularly in the motor cortex. For cross-sectional VBM studies, the variance in morphology across the population is likely to be the primary source of variability in the power analysis.

Precise and accurate quantification of whole-brain atrophy based on magnetic resonance imaging (MRI) data is an important goal in understanding the natural progression of neurodegenerative disorders such as Alzheimer's disease and multiple sclerosis. We found that inconsistent MRI positioning of subjects is common in typically acquired clinical trial data - particularly along the magnet's long (i.e., Z) axis. We also found that, if not corrected for, the gradient distortion effects associated with such Z-shifts can significantly decrease the accuracy and precision of MRI-derived measures of whole-brain atrophy - negative effects that increase in magnitude with (i) increases in the Z-distance between the brains to be compared and (ii) increases in the Z-distance from magnet isocenter of the center of the pair of brains to be compared. These gradient distortion effects can be reduced by accurate subject positioning, and they can also be corrected post hoc with the use of appropriately-generated gradient-distortion correction fields. We used a novel DUPLO-based phantom to develop a spherical-harmonics-based gradient distortion field that was used to (i) correct for observed Z-shift-associated gradient distortion effects on SIENA-generated measures of brain atrophy and (ii) simulate the gradient distortion effects that might be expected with a greater range of Z-shifts than those that we were able to acquire. Our results suggest that consistent alignment to magnet isocenter and/or correcting for the observed effects of gradient distortion should lead to more accurate and precise estimates of brain-related changes and, as a result, to increased statistical power in studies aimed at understanding the natural progression and the effective treatment of neurodegenerative disorders.

Spatial normalization, registration, and segmentation techniques for Magnetic Resonance Imaging (MRI) often use a target or template volume to facilitate processing, take advantage of prior information, and define a common coordinate system for analysis. In the neuroimaging literature, the MNI305 Talairach-like coordinate system is often used as a standard template. However, when studying pediatric populations, variation from the adult brain makes the MNI305 suboptimal for processing brain images of children. Morphological changes occurring during development render the use of age-appropriate templates desirable to reduce potential errors and minimize bias during processing of pediatric data. This paper presents the methods used to create unbiased, age-appropriate MRI atlas templates for pediatric studies that represent the average anatomy for the age range of 4.5-18.5 years, while maintaining a high level of anatomical detail and contrast. The creation of anatomical T1-weighted, T2-weighted, and proton density-weighted templates for specific developmentally important age-ranges, used data derived from the largest epidemiological, representative (healthy and normal) sample of the U.S. population, where each subject was carefully screened for medical and psychiatric factors and characterized using established neuropsychological and behavioral assessments. Use of these age-specific templates was evaluated by computing average tissue maps for gray matter, white matter, and cerebrospinal fluid for each specific age range, and by conducting an exemplar voxel-wise deformation-based morphometry study using 66 young (4.5-6.9 years) participants to demonstrate the benefits of using the age-appropriate templates. The public availability of these atlases/templates will facilitate analysis of pediatric MRI data and enable comparison of results between studies in a common standardized space specific to pediatric research.

Several methods exist and are frequently used to quantify grey matter (GM) atrophy in multiple sclerosis (MS). Fundamental to all available techniques is the accurate segmentation of GM in the brain, a difficult task confounded even further by the pathology present in the brains of MS patients. In this paper, we examine the segmentations of six different automated techniques and compare them to a manually defined reference standard. Results demonstrate that, although the algorithms perform similarly to manual segmentations of cortical GM, severe shortcomings are present in the segmentation of deep GM structures. This deficiency is particularly relevant given the current interest in the role of GM in MS and the numerous reports of atrophy in deep GM structures.

Most of human genome is present in two copies (maternal and paternal). However, segments of the genome can be deleted or duplicated, and many of these genomic variations (known as Copy Number Variants) are associated with psychiatric disorders. 16p11.2 copy number variants (breakpoint 4-5) confer high risk for neurodevelopmental disorders and are associated with structural brain alterations of large effect-size. Methods used in previous studies were unable to investigate the onset of these alterations and whether they evolve with age. In this study, we aim at characterizing age-related effects of 16p11.2 copy number variants by analyzing a group with a broad age range including younger individuals. A large normative developmental dataset was used to accurately adjust for effects of age. We normalized volumes of segmented brain regions as well as volumes of each voxel defined by tensor-based morphometry. Results show that the total intracranial volumes, the global gray and white matter volumes are respectively higher and lower in deletion and duplication carriers compared to control subjects at 4.5 years of age. These differences remain stable through childhood, adolescence and adulthood until 23 years of age (range: 0.5 to 1.0 Z-score). Voxel-based results are consistent with previous findings in 16p11.2 copy number variant carriers, including increased volume in the calcarine cortex and insula in deletions, compared to controls, with an inverse effect in duplication carriers (1.0 Z-score). All large effect-size voxel-based differences are present at 4.5 years and seem to remain stable until the age of 23. Our results highlight the stability of a neuroimaging endophenotype over 2 decades during which neurodevelopmental symptoms evolve at a rapid pace.

A new automatic model-based segmentation scheme that combines level set shape modeling and active appearance modeling (AAM) is presented. Since different MR image contrasts can yield complementary information, multi-contrast images can be incorporated into the active appearance modeling to improve segmentation performance. During active appearance modeling, the weighting of each contrast is optimized to account for the potentially varying contribution of each image while optimizing the model parameters that correspond to the shape and appearance eigen-images in order to minimize the difference between the multi-contrast test images and the ones synthesized from the shape and appearance modeling. As appearance-based modeling techniques are dependent on the initial alignment of training data, we compare (i) linear alignment of whole brain, (ii) linear alignment of a local volume of interest and (iii) non-linear alignment of a local volume of interest. The proposed segmentation scheme can be used to segment human hippocampi (HC) and amygdalae (AG), which have weak intensity contrast with their background in MRI. The experiments demonstrate that non-linear alignment of training data yields the best results and that multimodal segmentation using T1-weighted, T2-weighted and proton density-weighted images yields better segmentation results than any single contrast. In a four-fold cross validation with eighty young normal subjects, the method yields a mean Dice к of 0.87 with intraclass correlation coefficient (ICC) of 0.946 for HC and a mean Dice к of 0.81 with ICC of 0.924 for AG between manual and automatic labels.

Three-dimensional atlases of subcortical brain structures are valuable tools to reference anatomy in neuroscience and neurology. For instance, they can be used to study the position and shape of the three most common deep brain stimulation (DBS) targets, the subthalamic nucleus (STN), internal part of the pallidum (GPi) and ventral intermediate nucleus of the thalamus (VIM) in spatial relationship to DBS electrodes. Here, we present a composite atlas based on manual segmentations of a multimodal high resolution brain template, histology and structural connectivity. In a first step, four key structures were defined on the template itself using a combination of multispectral image analysis and manual segmentation. Second, these structures were used as anchor points to coregister a detailed histological atlas into standard space. Results show that this approach significantly improved coregistration accuracy over previously published methods. Finally, a sub-segmentation of STN and GPi into functional zones was achieved based on structural connectivity. The result is a composite atlas that defines key nuclei on the template itself, fills the gaps between them using histology and further subdivides them using structural connectivity. We show that the atlas can be used to segment DBS targets in single subjects, yielding more accurate results compared to priorly published atlases. The atlas will be made publicly available and constitutes a resource to study DBS electrode localizations in combination with modern neuroimaging methods.

The human cerebellum plays an essential role in motor control, is involved in cognitive function (i.e., attention, working memory, and language), and helps to regulate emotional responses. Quantitative in-vivo assessment of the cerebellum is important in the study of several neurological diseases including cerebellar ataxia, autism, and schizophrenia. Different structural subdivisions of the cerebellum have been shown to correlate with differing pathologies. To further understand these pathologies, it is helpful to automatically parcellate the cerebellum at the highest fidelity possible. In this paper, we coordinated with colleagues around the world to evaluate automated cerebellum parcellation algorithms on two clinical cohorts showing that the cerebellum can be parcellated to a high accuracy by newer methods. We characterize these various methods at four hierarchical levels: coarse (i.e., whole cerebellum and gross structures), lobe, subdivisions of the vermis, and the lobules. Due to the number of labels, the hierarchy of labels, the number of algorithms, and the two cohorts, we have restricted our analyses to the Dice measure of overlap. Under these conditions, machine learning based methods provide a collection of strategies that are efficient and deliver parcellations of a high standard across both cohorts, surpassing previous work in the area. In conjunction with the rank-sum computation, we identified an overall winning method.

Recent functional connectivity studies have demonstrated that, in resting humans, activity in a dorsally-situated neocortical network is inversely associated with activity in the amygdalae. Similarly, in human neuroimaging studies, aspects of emotion regulation have been associated with increased activity in dorsolateral, dorsomedial, orbital and ventromedial prefrontal regions, as well as concomitant decreases in amygdalar activity. These findings indicate the presence of two countervailing systems in the human brain that are reciprocally related: a dorsally-situated cognitive control network, and a ventrally-situated limbic network. We investigated the extent to which this functional reciprocity between limbic and dorsal neocortical regions is recapitulated from a purely structural standpoint. Specifically, we hypothesized that amygdalar volume would be related to cerebral cortical thickness in cortical regions implicated in aspects of emotion regulation. In 297 typically developing youths (162 females, 135 males; 572 MRIs), the relationship between cortical thickness and amygdalar volume was characterized. Amygdalar volume was found to be inversely associated with thickness in bilateral dorsolateral and dorsomedial prefrontal, inferior parietal, as well as bilateral orbital and ventromedial prefrontal cortices. Our findings are in line with previous work demonstrating that a predominantly dorsally-centered neocortical network is reciprocally related to core limbic structures such as the amygdalae. Future research may benefit from investigating the extent to which such cortical-limbic morphometric relations are qualified by the presence of mood and anxiety psychopathology.

Structural MRI is widely used for investigating brain atrophy in many neurodegenerative disorders, with several research groups developing and publishing techniques to provide quantitative assessments of this longitudinal change. Often techniques are compared through computation of required sample size estimates for future clinical trials. However interpretation of such comparisons is rendered complex because, despite using the same publicly available cohorts, the various techniques have been assessed with different data exclusions and different statistical analysis models. We created the MIRIAD atrophy challenge in order to test various capabilities of atrophy measurement techniques. The data consisted of 69 subjects (46 Alzheimer's disease, 23 control) who were scanned multiple (up to twelve) times at nine visits over a follow-up period of one to two years, resulting in 708 total image sets. Nine participating groups from 6 countries completed the challenge by providing volumetric measurements of key structures (whole brain, lateral ventricle, left and right hippocampi) for each dataset and atrophy measurements of these structures for each time point pair (both forward and backward) of a given subject. From these results, we formally compared techniques using exactly the same dataset. First, we assessed the repeatability of each technique using rates obtained from short intervals where no measurable atrophy is expected. For those measures that provided direct measures of atrophy between pairs of images, we also assessed symmetry and transitivity. Then, we performed a statistical analysis in a consistent manner using linear mixed effect models. The models, one for repeated measures of volume made at multiple time-points and a second for repeated "direct" measures of change in brain volume, appropriately allowed for the correlation between measures made on the same subject and were shown to fit the data well. From these models, we obtained estimates of the distribution of atrophy rates in the Alzheimer's disease (AD) and control groups and of required sample sizes to detect a 25% treatment effect, in relation to healthy ageing, with 95% significance and 80% power over follow-up periods of 6, 12, and 24months. Uncertainty in these estimates, and head-to-head comparisons between techniques, were carried out using the bootstrap. The lateral ventricles provided the most stable measurements, followed by the brain. The hippocampi had much more variability across participants, likely because of differences in segmentation protocol and less distinct boundaries. Most methods showed no indication of bias based on the short-term interval results, and direct measures provided good consistency in terms of symmetry and transitivity. The resulting annualized rates of change derived from the model ranged from, for whole brain: -1.4% to -2.2% (AD) and -0.35% to -0.67% (control), for ventricles: 4.6% to 10.2% (AD) and 1.2% to 3.4% (control), and for hippocampi: -1.5% to -7.0% (AD) and -0.4% to -1.4% (control). There were large and statistically significant differences in the sample size requirements between many of the techniques. The lowest sample sizes for each of these structures, for a trial with a 12month follow-up period, were 242 (95% CI: 154 to 422) for whole brain, 168 (95% CI: 112 to 282) for ventricles, 190 (95% CI: 146 to 268) for left hippocampi, and 158 (95% CI: 116 to 228) for right hippocampi. This analysis represents one of the most extensive statistical comparisons of a large number of different atrophy measurement techniques from around the globe. The challenge data will remain online and publicly available so that other groups can assess their methods.