Search

The ventralis intermedius nucleus (Vim) is centrally placed in the dentato-thalamo-cortical pathway (DTCp) and is a key surgical target in the treatment of severe medically refractory tremor. It is not visible on conventional MRI sequences; consequently, stereotactic targeting currently relies on atlas-based coordinates. This fails to capture individual anatomical variability, which may lead to poor long-term clinical efficacy. Probabilistic tractography, combined with known anatomical connectivity, enables localisation of thalamic nuclei at an individual subject level. There are, however, a number of confounds associated with this technique that may influence results. Here we focused on an established method, using probabilistic tractography to reconstruct the DTCp, to identify the connectivity-defined Vim (cd-Vim) in vivo. Using 100 healthy individuals from the Human Connectome Project, our aim was to quantify cd-Vim variability across this population, measure the discrepancy with atlas-defined Vim (ad-Vim), and assess the influence of potential methodological confounds. We found no significant effect of any of the confounds. The mean cd-Vim coordinate was located within 1.88 mm (left) and 2.12 mm (right) of the average midpoint and 3.98 mm (left) and 5.41 mm (right) from the ad-Vim coordinates. cd-Vim location was more variable on the right, which reflects hemispheric asymmetries in the probabilistic DTC reconstructed. The method was reproducible, with no significant cd-Vim location differences in a separate test-retest cohort. The superior cerebellar peduncle was identified as a potential source of artificial variance. This work demonstrates significant individual anatomical variability of the cd-Vim that atlas-based coordinate targeting fails to capture. This variability was not related to any methodological confound tested. Lateralisation of cerebellar functions, such as speech, may contribute to the observed asymmetry. Tractography-based methods seem sensitive to individual anatomical variability that is missed by conventional neurosurgical targeting; these findings may form the basis for translational tools to improve efficacy and reduce side-effects of thalamic surgery for tremor.

The androgen receptor (AR), oestrogen receptor alpha (ESR1) and oestrogen receptor beta (ESR2) play essential roles in mediating the effect of sex hormones on sex differences in the brain. Using Voxel-based morphometry (VBM) and gene sizing in two independent samples (discovery n ​= ​173, replication ​= ​61), we determine the common and unique influences on brain sex differences in grey (GM) and white matter (WM) volume between repeat lengths (n) of microsatellite polymorphisms AR(CAG)n, ESR1(TA)n and ESR2(CA)n. In the hypothalamus, temporal lobes, anterior cingulate cortex, posterior insula and prefrontal cortex, we find increased GM volume with increasing AR(CAG)n across sexes, decreasing ESR1(TA)n across sexes and decreasing ESR2(CA)n in females. Uniquely, AR(CAG)n was positively associated with dorsolateral prefrontal and orbitofrontal GM volume and the anterior corona radiata, left superior fronto-occipital fasciculus, thalamus and internal capsule WM volume. ESR1(TA)n was negatively associated with the left superior corona radiata, left cingulum and left inferior longitudinal fasciculus WM volume uniquely. ESR2(CA)n was negatively associated with right fusiform and posterior cingulate cortex uniquely. We thus describe the neuroanatomical correlates of three microsatellite polymorphisms of steroid hormone receptors and their relationship to sex differences.

Objectives Firstly, to identify subthalamic region stimulation clusters that predict maximum improvement in rigidity, bradykinesia and tremor, or emergence of side-effects; and secondly, to map-out the cortical fingerprint, mediated by the hyperdirect pathways which predict maximum efficacy. Methods High angular resolution diffusion imaging in twenty patients with advanced Parkinson's disease was acquired prior to bilateral subthalamic nucleus deep brain stimulation. All contacts were screened one-year from surgery for efficacy and side-effects at different amplitudes. Voxel-based statistical analysis of volumes of tissue activated models was used to identify significant treatment clusters. Probabilistic tractography was employed to identify cortical connectivity patterns associated with treatment efficacy. Results All patients responded well to treatment (46% mean improvement off medication UPDRS-III [p < 0.0001]) without significant adverse events. Cluster corresponding to maximum improvement in tremor was in the posterior, superior and lateral portion of the nucleus. Clusters corresponding to improvement in bradykinesia and rigidity were nearer the superior border in a further medial and posterior location. The rigidity cluster extended beyond the superior border to the area of the zona incerta and Forel-H2 field. When the clusters where averaged, the coordinates of the area with maximum overall efficacy was X = -10(-9.5), Y = -13(-1) and Z = -7(-3) in MNI(AC-PC) space. Cortical connectivity to primary motor area was predictive of higher improvement in tremor; whilst that to supplementary motor area was predictive of improvement in bradykinesia and rigidity; and connectivity to prefrontal cortex was predictive of improvement in rigidity. Interpretation These findings support the presence of overlapping stimulation sites within the subthalamic nucleus and its superior border, with different cortical connectivity patterns, associated with maximum improvement in tremor, rigidity and bradykinesia.

Neuroimaging data are increasingly being used to predict potential outcomes or groupings, such as clinical severity, drug dose response, and transitional illness states. In these examples, the variable (target) we want to predict is ordinal in nature. Conventional classification schemes assume that the targets are nominal and hence ignore their ranked nature, whereas parametric and/or non-parametric regression models enforce a metric notion of distance between classes. Here, we propose a novel, alternative multivariate approach that overcomes these limitations — whole brain probabilistic ordinal regression using a Gaussian process framework. We applied this technique to two data sets of pharmacological neuroimaging data from healthy volunteers. The first study was designed to investigate the effect of ketamine on brain activity and its subsequent modulation with two compounds — lamotrigine and risperidone. The second study investigates the effect of scopolamine on cerebral blood flow and its modulation using donepezil. We compared ordinal regression to multi-class classification schemes and metric regression. Considering the modulation of ketamine with lamotrigine, we found that ordinal regression significantly outperformed multi-class classification and metric regression in terms of accuracy and mean absolute error. However, for risperidone ordinal regression significantly outperformed metric regression but performed similarly to multi-class classification both in terms of accuracy and mean absolute error. For the scopolamine data set, ordinal regression was found to outperform both multi-class and metric regression techniques considering the regional cerebral blood flow in the anterior cingulate cortex. Ordinal regression was thus the only method that performed well in all cases. Our results indicate the potential of an ordinal regression approach for neuroimaging data while providing a fully probabilistic framework with elegant approaches for model selection., Highlights • Often in neuroimaging the independent variables are ranked or ordered. • Classification and regression models cannot explicitly model an ordinal target. • We present a novel multivariate ordinal regression approach for neuroimaging data. • Our results show that ordinal regression is a powerful method for ranking data.

The subthalamic nucleus (STN) is a small, glutamatergic nucleus situated in the diencephalon. A critical component of normal motor function, it has become a key target for deep brain stimulation in the treatment of Parkinson's disease. Animal studies have demonstrated the existence of three functional sub-zones but these have never been shown conclusively in humans. In this work, a data driven method with diffusion weighted imaging demonstrated that three distinct clusters exist within the human STN based on brain connectivity profiles. The STN was successfully sub-parcellated into these regions, demonstrating good correspondence with that described in the animal literature. The local connectivity of each sub-region supported the hypothesis of bilateral limbic, associative and motor regions occupying the anterior, mid and posterior portions of the nucleus respectively. This study is the first to achieve in-vivo, non-invasive anatomical parcellation of the human STN into three anatomical zones within normal diagnostic scan times, which has important future implications for deep brain stimulation surgery., Highlights ► Three distinct sub-regions within the human STN are demonstrated in vivo using DWI. ► Limbic, associative and motor zones are labelled based on the regional connectivity. ► The findings agree with previous results from the animal literature. ► A somatotopic arrangement of STN projections to subcortical structures is shown. ► An overlap between motor STN projections and extra-STN hemiballismus is demonstrated.

This paper describes a general kernel regression approach to predict experimental conditions from activity patterns acquired with functional magnetic resonance image (fMRI). The standard approach is to use classifiers that predict conditions from activity patterns. Our approach involves training different regression machines for each experimental condition, so that a predicted temporal profile is computed for each condition. A decision function is then used to classify the responses from the testing volumes into the corresponding category, by comparing the predicted temporal profile elicited by each event, against a canonical haemodynamic response function. This approach utilizes the temporal information in the fMRI signal and maintains more training samples in order to improve the classification accuracy over an existing strategy. This paper also introduces efficient techniques of temporal compaction, which operate directly on kernel matrices for kernel classification algorithms such as the support vector machine (SVM). Temporal compacting can convert the kernel computed from each fMRI volume directly into the kernel computed from beta-maps, average of volumes or spatial-temporal kernel. The proposed method was applied to three different datasets. The first one is a block-design experiment with three conditions of image stimuli. The method outperformed the SVM classifiers of three different types of temporal compaction in single-subject leave-one-block-out cross-validation. Our method achieved 100% classification accuracy for six of the subjects and an average of 94% accuracy across all 16 subjects, exceeding the best SVM classification result, which was 83% accuracy (p=0.008). The second dataset is also a block-design experiment with two conditions of visual attention (left or right). Our method yielded 96% accuracy and SVM yielded 92% (p=0.005). The third dataset is from a fast event-related experiment with two categories of visual objects. Our method achieved 77% accuracy, compared with 72% using SVM (p=0.0006).

This paper introduces two kernel-based regression schemes to decode or predict brain states from functional brain scans as part of the Pittsburgh Brain Activity Interpretation Competition (PBAIC) 2007, in which our team was awarded first place. Our procedure involved image realignment, spatial smoothing, detrending of low-frequency drifts, and application of multivariate linear and non-linear kernel regression methods: namely kernel ridge regression (KRR) and relevance vector regression (RVR). RVR is based on a Bayesian framework, which automatically determines a sparse solution through maximization of marginal likelihood. KRR is the dual-form formulation of ridge regression, which solves regression problems with high dimensional data in a computationally efficient way. Feature selection based on prior knowledge about human brain function was also used. Post-processing by constrained deconvolution and re-convolution was used to furnish the prediction. This paper also contains a detailed description of how prior knowledge was used to fine tune predictions of specific “feature ratings,” which we believe is one of the key factors in our prediction accuracy. The impact of pre-processing was also evaluated, demonstrating that different pre-processing may lead to significantly different accuracies. Although the original work was aimed at the PBAIC, many techniques described in this paper can be generally applied to any fMRI decoding works to increase the prediction accuracy.

Recent genetic studies have implicated a number of candidate genes in the pathogenesis of Autism Spectrum Disorder (ASD). Polymorphisms of CNTNAP2 (contactin-associated like protein-2), a member of the neurexin family, have already been implicated as a susceptibility gene for autism by at least 3 separate studies. We investigated variation in white and grey matter morphology using structural MRI and diffusion tensor imaging. We compared volumetric differences in white and grey matter and fractional anisotropy values in control subjects characterised by genotype at rs7794745, a single nucleotide polymorphism in CNTNAP2. Homozygotes for the risk allele showed significant reductions in grey and white matter volume and fractional anisotropy in several regions that have already been implicated in ASD, including the cerebellum, fusiform gyrus, occipital and frontal cortices. Male homozygotes for the risk alleles showed greater reductions in grey matter in the right frontal pole and in FA in the right rostral fronto-occipital fasciculus compared to their female counterparts who showed greater reductions in FA of the anterior thalamic radiation. Thus a risk allele for autism results in significant cerebral morphological variation, despite the absence of overt symptoms or behavioural abnormalities. The results are consistent with accumulating evidence of CNTNAP2's function in neuronal development. The finding suggests the possibility that the heterogeneous manifestations of ASD can be aetiologically characterised into distinct subtypes through genetic-morphological analysis.

Machine learning and pattern recognition methods have been used to diagnose Alzheimer's disease (AD) and mild cognitive impairment (MCI) from individual MRI scans. Another application of such methods is to predict clinical scores from individual scans. Using relevance vector regression (RVR), we predicted individuals' performances on established tests from their MRI T1 weighted image in two independent data sets. From Mayo Clinic, 73 probable AD patients and 91 cognitively normal (CN) controls completed the Mini-Mental State Examination (MMSE), Dementia Rating Scale (DRS), and Auditory Verbal Learning Test (AVLT) within 3months of their scan. Baseline MRI's from the Alzheimer's disease Neuroimaging Initiative (ADNI) comprised the other data set; 113 AD, 351 MCI, and 122 CN subjects completed the MMSE and Alzheimer's Disease Assessment Scale—Cognitive subtest (ADAS-cog) and 39 AD, 92 MCI, and 32 CN ADNI subjects completed MMSE, ADAS-cog, and AVLT. Predicted and actual clinical scores were highly correlated for the MMSE, DRS, and ADAS-cog tests (P

Our understanding of how genotype determines phenotype in primary dystonia is limited. Familial young-onset primary dystonia is commonly due to the DYT1 gene mutation. A critical question, given the 30% penetrance of clinical symptoms in DYT1 mutation carriers, is why the same genotype leads to differential clinical expression and whether non-DYT1 adult-onset primary dystonia, with and without family history share pathophysiological mechanisms with DYT1 dystonia. This study examines the relationship between dystonic phenotype and the DYT1 gene mutation by monitoring whole-brain structure using voxel-based morphometry. We acquired magnetic resonance imaging data of symptomatic and asymptomatic DYT1 mutation carriers, of non-DYT1 primary dystonia patients, with and without family history and control subjects with normal DYT1 alleles. By crossing the factors genotype and phenotype we demonstrate a significant interaction in terms of brain anatomy confined to the basal ganglia bilaterally. The explanation for this effect differs according to both gene and dystonia status: non-DYT1 adult-onset dystonia patients and asymptomatic DYT1 carriers have significantly larger basal ganglia compared to healthy subjects and symptomatic DYT1 mutation carriers. There is a significant negative correlation between severity of dystonia and basal ganglia size in DYT1 mutation carriers. We propose that differential pathophysiological and compensatory mechanisms lead to brain structure changes in non-DYT1 primary adult-onset dystonias and DYT1 gene carriers. Given the range of age of onset, there may be differential genetic modulation of brain development that in turn determines clinical expression. Alternatively, a DYT1 gene dependent primary defect of motor circuit development may lead to stress-induced remodelling of the basal ganglia and hence dystonia.

In functional brain mapping, pattern recognition methods allow detecting multivoxel patterns of brain activation which are informative with respect to a subject's perceptual or cognitive state. The sensitivity of these methods, however, is greatly reduced when the proportion of voxels that convey the discriminative information is small compared to the total number of measured voxels. To reduce this dimensionality problem, previous studies employed univariate voxel selection or region-of-interest-based strategies as a preceding step to the application of machine learning algorithms. Here we employ a strategy for classifying functional imaging data based on a multivariate feature selection algorithm, Recursive Feature Elimination (RFE) that uses the training algorithm (support vector machine) recursively to eliminate irrelevant voxels and estimate informative spatial patterns. Generalization performances on test data increases while features/voxels are pruned based on their discrimination ability. In this article we evaluate RFE in terms of sensitivity of discriminative maps (Receiver Operative Characteristic analysis) and generalization performances and compare it to previously used univariate voxel selection strategies based on activation and discrimination measures. Using simulated fMRI data, we show that the recursive approach is suitable for mapping discriminative patterns and that the combination of an initial univariate activation-based (F-test) reduction of voxels and multivariate recursive feature elimination produces the best results, especially when differences between conditions have a low contrast-to-noise ratio. Furthermore, we apply our method to high resolution (2 × 2 × 2mm3) data from an auditory fMRI experiment in which subjects were stimulated with sounds from four different categories. With these real data, our recursive algorithm proves able to detect and accurately classify multivoxel spatial patterns, highlighting the role of the superior temporal gyrus in encoding the information of sound categories. In line with the simulation results, our method outperforms univariate statistical analysis and statistical learning without feature selection.

We describe a Bayesian scheme to analyze images, which uses spatial priors encoded by a diffusion kernel, based on a weighted graph Laplacian. This provides a general framework to formulate a spatial model, whose parameters can be optimized. The application we have in mind is a spatiotemporal model for imaging data. We illustrate the method on a random effects analysis of fMRI contrast images from multiple subjects; this simplifies exposition of the model and enables a clear description of its salient features. Typically, imaging data are smoothed using a fixed Gaussian kernel as a pre-processing step before applying a mass-univariate statistical model (e.g., a general linear model) to provide images of parameter estimates. An alternative is to include smoothness in a multivariate statistical model (Penny, W.D., Trujillo-Barreto, N.J., Friston, K.J., 2005. Bayesian fMRI time series analysis with spatial priors. Neuroimage 24, 350–362). The advantage of the latter is that each parameter field is smoothed automatically, according to a measure of uncertainty, given the data. In this work, we investigate the use of diffusion kernels to encode spatial correlations among parameter estimates. Nonlinear diffusion has a long history in image processing; in particular, flows that depend on local image geometry (Romeny, B.M.T., 1994. Geometry-driven Diffusion in Computer Vision. Kluwer Academic Publishers) can be used as adaptive filters. This can furnish a non-stationary smoothing process that preserves features, which would otherwise be lost with a fixed Gaussian kernel. We describe a Bayesian framework that incorporates non-stationary, adaptive smoothing into a generative model to extract spatial features in parameter estimates. Critically, this means adaptive smoothing becomes an integral part of estimation and inference. We illustrate the method using synthetic and real fMRI data.

A key component of group analyses of neuroimaging data is precise and valid spatial normalization (i.e., inter-subject image registration). When patients have structural brain lesions, such as a stroke, this process can be confounded by the lack of correspondence between the subject and standardized template images. Current procedures for dealing with this problem include regularizing the estimate of warping parameters used to match lesioned brains to the template, or “cost function masking”; both these solutions have significant drawbacks. We report three experiments that identify the best spatial normalization for structurally damaged brains and establish whether differences among normalizations have a significant effect on inferences about functional activations. Our novel protocols evaluate the effects of different normalization solutions and can be applied easily to any neuroimaging study. This has important implications for users of both structural and functional imaging techniques in the study of patients with structural brain damage.

This note derives the variational free energy under the Laplace approximation, with a focus on accounting for additional model complexity induced by increasing the number of model parameters. This is relevant when using the free energy as an approximation to the log-evidence in Bayesian model averaging and selection. By setting restricted maximum likelihood (ReML) in the larger context of variational learning and expectation maximisation (EM), we show how the ReML objective function can be adjusted to provide an approximation to the log-evidence for a particular model. This means ReML can be used for model selection, specifically to select or compare models with different covariance components. This is useful in the context of hierarchical models because it enables a principled selection of priors that, under simple hyperpriors, can be used for automatic model selection and relevance determination (ARD). Deriving the ReML objective function, from basic variational principles, discloses the simple relationships among Variational Bayes, EM and ReML. Furthermore, we show that EM is formally identical to a full variational treatment when the precisions are linear in the hyperparameters. Finally, we also consider, briefly, dynamic models and how these inform the regularisation of free energy ascent schemes, like EM and ReML.

A probabilistic framework is presented that enables image registration, tissue classification, and bias correction to be combined within the same generative model. A derivation of a log-likelihood objective function for the unified model is provided. The model is based on a mixture of Gaussians and is extended to incorporate a smooth intensity variation and nonlinear registration with tissue probability maps. A strategy for optimising the model parameters is described, along with the requisite partial derivatives of the objective function.

In this paper we address the assumptions about the distribution of errors made by voxel-based morphometry. Voxel-based morphometry (VBM) uses the general linear model to construct parametric statistical tests. In order for these statistics to be valid, a small number of assumptions must hold. A key assumption is that the model's error terms are normally distributed. This is usually ensured through the Central Limit Theorem by smoothing the data. However, there is increasing interest in using minimal smoothing (in order to sensitize the analysis to regional differences at a small spatial scale). The validity of such analyses is investigated. In brief, our results indicate that nonnormality in the error terms can be an issue in VBM. However, in balanced designs, provided the data are smoothed with a 4-mm FWHM kernel, nonnormality is sufficiently attenuated to render the tests valid. Unbalanced designs appear to be less robust to violations of normality: a significant number of false positives arise at a smoothing of 4 and 8 mm when comparing a single subject to a group. This is despite the fact that conventional group comparisons appear to be robust, remaining valid even with no smoothing. The implications of the results for researchers using voxel-based morphometry are discussed.

We compared voxel-based morphometry (VBM) with independent accurate region-of-interest (ROI) measurements of temporal lobe structures in order to validate the usefulness of this fully automated and unbiased technique in Alzheimer's disease (AD) and semantic dementia (SD). In AD, ROI analyses appear more sensitive to volume loss in the amygdalae, whereas VBM analyses appear more sensitive to right middle temporal gyrus and regional hippocampal volume loss. In SD, ROI analyses appear more sensitive to left middle and inferior temporal gyrus volume loss, whereas VBM appears more sensitive to regional hippocampal volume loss. In addition the significance of volume reductions was generally less in VBM owing to more stringent corrections for multiple comparisons. In conclusion, the automated technique detects a general trend of atrophy similar to that of expertly labeled ROI measurements in AD and SD, although there are discrepancies in the ranking of severity and in the significance of volume reductions that are more marked in AD.

In Friston et al. ((2002) Neuroimage 16: 465-483) we introduced empirical Bayes as a potentially useful way to estimate and make inferences about effects in hierarchical models. In this paper we present a series of models that exemplify the diversity of problems that can be addressed within this framework. In hierarchical linear observation models, both classical and empirical Bayesian approaches can be framed in terms of covariance component estimation (e.g., variance partitioning). To illustrate the use of the expectation-maximization (EM) algorithm in covariance component estimation we focus first on two important problems in fMRI: nonsphericity induced by (i) serial or temporal correlations among errors and (ii) variance components caused by the hierarchical nature of multisubject studies. In hierarchical observation models, variance components at higher levels can be used as constraints on the parameter estimates of lower levels. This enables the use of parametric empirical Bayesian (PEB) estimators, as distinct from classical maximum likelihood (ML) estimates. We develop this distinction to address: (i) The difference between response estimates based on ML and the conditional means from a Bayesian approach and the implications for estimates of intersubject variability. (ii) The relationship between fixed- and random-effect analyses. (iii) The specificity and sensitivity of Bayesian inference and, finally, (iv) the relative importance of the number of scans and subjects. The forgoing is concerned with within- and between-subject variability in multisubject hierarchical fMRI studies. In the second half of this paper we turn to Bayesian inference at the first (within-voxel) level, using PET data to show how priors can be derived from the (between-voxel) distribution of activations over the brain. This application uses exactly the same ideas and formalism but, in this instance, the second level is provided by observations over voxels as opposed to subjects. The ensuing posterior probability maps (PPMs) have enhanced anatomical precision and greater face validity, in relation to underlying anatomy. Furthermore, in comparison to conventional SPMs they are not confounded by the multiple comparison problem that, in a classical context, dictates high thresholds and low sensitivity. We conclude with some general comments on Bayesian approaches to image analysis and on some unresolved issues.

A well-recognized problem with the echo-planar imaging (EPI) technique most commonly used for functional magnetic resonance imaging (fMRI) studies is geometric distortion caused by magnetic field inhomogeneity. This makes it difficult to achieve an accurate registration between a functional activation map calculated from an EPI time series and an undistorted, high resolution anatomical image. A correction method based on mapping the spatial distribution of field inhomogeneities can be used to reduce these distortions. This approach is attractive in its simplicity but requires postprocessing to improve the robustness of the acquired field map and reduce any secondary artifacts. Furthermore, the distribution of the internal magnetic field throughout the head is position dependent resulting in an interaction between distortion and head motion. Therefore, a single field map may not be sufficient to correct for the distortions throughout a whole fMRI time series. In this paper we present a quantitative evaluation of image distortion correction for fMRI at 2T. We assess (i) methods for the acquisition and calculation of field maps, (ii) the effect of image distortion correction on the coregistration between anatomical and functional images, and (iii) the interaction between distortion and head motion, assessing the feasibility of using field maps to reduce this effect. We propose that field maps with acceptable noise levels can be generated easily using a dual echo-time EPI sequence and demonstrate the importance of distortion correction for anatomical coregistration, even for small distortions. Using a dual echo-time series to generate a unique field map at each time point, we characterize the interaction between head motion and geometric distortion. However, we suggest that the variance between successively measured field maps introduces additional unwanted variance in the voxel time-series and is therefore not adequate to correct for time-varying distortions.

Recently in this journal, Alkemade and Forstmann again challenged the evidence for a tripartite organisation to the subthalamic nucleus (STN) (Alkemade & Forstmann 2014). Additionally, they raised specific issues with the earlier published results using 3T MRI to perform in vivo diffusion weighted imaging (DWI) based segmentation of the STN (Lambert et al. 2012). Their comments reveal a common misconception related to the underlying methodologies used, which we clarify in this reply, in addition to highlighting how their current conclusions are synonymous with our original paper. The ongoing debate, instigated by the controversies surrounding STN parcellation, raises important implications for the assumptions and methodologies employed in mapping functional brain anatomy, both in vivo and ex vivo, and reveals a fundamental emergent problem with the current techniques. These issues are reviewed, and potential strategies that could be developed to manage them in the future are discussed further.

Total intracranial volume (TIV/ICV) is an important covariate for volumetric analyses of the brain and brain regions, especially in the study of neurodegenerative diseases, where it can provide a proxy of maximum pre-morbid brain volume. The gold-standard method is manual delineation of brain scans, but this requires careful work by trained operators. We evaluated Statistical Parametric Mapping 12 (SPM12) automated segmentation for TIV measurement in place of manual segmentation and also compared it with SPM8 and FreeSurfer 5.3.0. For T1-weighted MRI acquired from 288 participants in a multi-centre clinical trial in Alzheimer's disease we find a high correlation between SPM12 TIV and manual TIV (R2 = 0.940, 95% Confidence Interval (0.924, 0.953)), with a small mean difference (SPM12 40.4 ± 35.4 ml lower than manual, amounting to 2.8% of the overall mean TIV in the study). The correlation with manual measurements (the key aspect when using TIV as a covariate) for SPM12 was significantly higher (p, Graphical abstract, Highlights • 288 T1 MRI from multiple scanners were manually segmented for intracranial volume. • We compare SPM12 with the current methods of estimating intracranial volume. • SPM12 shows a very high correlation with manual measures and little bias. • Newer automated volume measures are more accurate controls for head size variation.

We used voxel-based morphometry (VBM) to examine human brain asymmetry and the effects of sex and handedness on brain structure in 465 normal adults. We observed significant asymmetry of cerebral grey and white matter in the occipital, frontal, and temporal lobes (petalia), including Heschl's gyrus, planum temporale (PT) and the hippocampal formation. Males demonstrated increased leftward asymmetry within Heschl's gyrus and PT compared to females. There was no significant interaction between asymmetry and handedness and no main effect of handedness. There was a significant main effect of sex on brain morphology, even after accounting for the larger global volumes of grey and white matter in males. Females had increased grey matter volume adjacent to the depths of both central sulci and the left superior temporal sulcus, in right Heschl's gyrus and PT, in right inferior frontal and frontomarginal gyri and in the cingulate gyrus. Females had significantly increased grey matter concentration extensively and relatively symmetrically in the cortical mantle, parahippocampal gyri, and in the banks of the cingulate and calcarine sulci. Males had increased grey matter volume bilaterally in the mesial temporal lobes, entorhinal and perirhinal cortex, and in the anterior lobes of the cerebellum, but no regions of increased grey matter concentration.

In this paper, we present a post hoc method for identifying regions where functional MRI data are subject to signal reduction that may compromise sensitivity to activations. The motivation for developing this technique derives from recent language studies that showed responses in the inferior temporal lobe that could be detected by PET but not by fMRI. Reduced signal is due mostly to susceptibility artifacts and can be identified by comparing the T2* images (which are subject to susceptibility artifacts) with T2 images (which are not). However, T2 images are not usually acquired in fMRI studies. Therefore, we propose that areas with reduced signal can be identified by comparing T2* images that are corrected for nonuniformity with the original uncorrected images. The technique provides a voxel-wise characterization of signal reduction that pertains to the particular data that enter into a statistical model. It requires only the functional data and can thus be applied post hoc and without any additional scans.

At its simplest, voxel-based morphometry (VBM) involves a voxel-wise comparison of the local concentration of gray matter between two groups of subjects. The procedure is relatively straightforward and involves spatially normalizing high-resolution images from all the subjects in the study into the same stereotactic space. This is followed by segmenting the gray matter from the spatially normalized images and smoothing the gray-matter segments. Voxel-wise parametric statistical tests which compare the smoothed gray-matter images from the two groups are performed. Corrections for multiple comparisons are made using the theory of Gaussian random fields. This paper describes the steps involved in VBM, with particular emphasis on segmenting gray matter from MR images with nonuniformity artifact. We provide evaluations of the assumptions that underpin the method, including the accuracy of the segmentation and the assumptions made about the statistical distribution of the data.

Quantitative evaluation of MRI in patients with epilepsy can give more information than qualitative assessment. Previously developed volume-of-interest-based methods identified subtle widespread structural changes in the neocortex beyond the visualized lesions in patients with malformations of cortical development (MCD) and hippocampal sclerosis (HS) and also in MRI-negative patients with juvenile myoclonic epilepsy (JME). This study evaluates a voxel-based automated analysis of structural MRI in epilepsy. After fully automated segmentation of cerebral gray matter from structural T1-weighted, high-resolution MRI scans, we applied the automated and objective technique of statistical parametric mapping (SPM) to the analysis of gray matter of 35 control subjects, 10 patients with partial seizures and MCD, 10 patients with left temporal lobe epilepsy (TLE) and HS, 10 patients with left TLE and normal MR quantitation of the hippocampus, and 20 patients with JME. At a corrected threshold of P < 0.05, significant abnormalities were found in 3/35 controls; in all 10 patients with MCD, 6 of whom had additional lesions beyond the margins of the visualized abnormalities; in 2/10 TLE patients with HS; in 2/10 MRI-negative TLE; and in 4/20 JME patients. Group comparisons between control subjects and HS patients identified the affected left temporal lobe with an increase in gray matter in the posterior temporal lobe, but did not identify hippocampal atrophy. The group of MRI-negative TLE patients showed no abnormalities compared with control subjects. Group comparison between control subjects and JME patients identified a mesial frontal increase in gray matter. The SPM-based voxel-by-voxel comparison of gray matter distribution identified MCD and abnormalities beyond the visualized lesion in individual MCD patients. The method did not reliably identify HS in individual patients or identify abnormalities in individual MRI-negative patients with TLE or JME in a proportion larger than the chance findings in the control group. Using group comparisons, structural abnormalities in the neocortical gray matter of patients with TLE and HS were lateralized to the affected temporal lobe. In patients with JME as a group, an increase in gray matter was localized to the mesial frontal area, corroborating earlier quantitative MRI findings.

This paper concerns the experimental design and statistical models employed by fMRI activation studies which block presentation of linguistic stimuli. In particular, we note that the relationship between the timing of stimulus presentation and data acquisition can have a substantial impact on the ability to detect activations in critical language areas, even when the stimuli are presented in blocks. Using a blocked word rhyming paradigm and repeated investigations on a single subject, activation was observed in Broca's area (left inferior frontal cortex) and Wernicke's area (left posterior temporoparietal cortex) when (i) the timing of data acquisition was distributed throughout the peristimulus time and (ii) an event-related analysis was used to model the phasic nature of the hemodynamic response within each block of repeated word stimuli. In contrast, when the timing of data acquisition relative to stimulus presentation was fixed, activation was detected in Broca's area but not consistently in Wernicke's area. Our results indicate that phasic responses to stimuli occur even in a blocked design and that the sampling and proper modeling of these responses can have profound effects on their detection. Specifically, distributed sampling over peristimulus time is essential in order to detect small activations particularly when they are transient. These findings are likely to generalize to the detection of transient signals in any cognitive paradigm.

This paper is about warping a brain image from one subject (the object image) so that it matches another (the template image). A high-dimensional model is used, whereby a finite element approach is employed to estimate translations at the location of each voxel in the template image. Bayesian statistics are used to obtain a maximum a posteriori (MAP) estimate of the deformation field. The validity of any registration method is largely based upon the constraints or, in this instance, priors incorporated into the model describing the transformations. In this approach we assume that the priors should have some form of symmetry, in that priors describing the probability distribution of the deformations should be identical to those for the inverses (i.e., warping brain A to brain B should not be different probabilistically from warping B to A). The fundamental assumption is that the probability of stretching a voxel by a factor of n is considered to be the same as the probability of shrinking n voxels by a factor of n(-1). In the Bayesian framework adopted here, the priors are assumed to have a Gibbs form, where the Gibbs potential is a penalty function that embodies this symmetry. The penalty function of choice is based upon the singular values of the Jacobian having a lognormal distribution. This enforces a continuous one-to-one mapping. A gradient descent algorithm is presented that incorporates the above priors in order to obtain a MAP estimate of the deformations. We demonstrate this approach for the two-dimensional case, but the principles can be extended to three dimensions. A number of examples are given to demonstrate how the method works.

The first step in the spatial normalization of brain images is usually to determine the affine transformation that best maps the image to a template image in a standard space. We have developed a rapid and automatic method for performing this registration, which uses a Bayesian scheme to incorporate prior knowledge of the variability in the shape and size of heads. We compared affine registrations with and without incorporating the prior knowledge. We found that the affine transformations derived using the Bayesian scheme are much more robust and that the rate of convergence is greater.

Coregistration of functional PET and T1-weighted MR images is a necessary step for combining functional information from PET images with anatomical information in MR images. Several coregistration algorithms have been published and are used in functional brain imaging studies. In this paper, we present a comparison and cross validation of the two most widely used coregistration routines (Friston et al., 1995, Hum. Brain Map. 2: 165-189; Woods et al., 1993, J. Comput. Assisted Tomogr: 17: 536-546). Several transformations were applied to high-resolution anatomical MR images to generate simulated PET images so that the exact (rigid body) transformations between each MR image and its associated simulated PET images were known. The estimation error of a coregistration in relation to the known transformation allows a comparison of the performance of different coregistration routines. Under the assumption that the simulated PET images embody the salient features of real PET images with respect to coregistration, this study shows that the routines examined reliably solve the MRI to PET coregistration problem.

This paper presents a very selective review of some of the approaches for multivariate modelling of inter-subject variability among brain images. It focusses on applying probabilistic kernel-based pattern recognition approaches to pre-processed anatomical MRI, with the aim of most accurately modelling the difference between populations of subjects. Some of the principles underlying the pattern recognition approaches of Gaussian process classification and regression are briefly described, although the reader is advised to look elsewhere for full implementational details. Kernel pattern recognition methods require matrices that encode the degree of similarity between the images of each pair of subjects. This review focusses on similarity measures derived from the relative shapes of the subjects' brains. Pre-processing is viewed as generative modelling of anatomical variability, and there is a special emphasis on the diffeomorphic image registration framework, which provides a very parsimonious representation of relative shapes. Although the review is largely methodological, excessive mathematical notation is avoided as far as possible, as the paper attempts to convey a more intuitive understanding of various concepts. The paper should be of interest to readers wishing to apply pattern recognition methods to MRI data, with the aim of clinical diagnosis or biomarker development. It also tries to explain that the best models are those that most accurately predict, so similar approaches should also be relevant to basic science. Knowledge of some basic linear algebra and probability theory should make the review easier to follow, although it may still have something to offer to those readers whose mathematics may be more limited.

This note presents a framework for generating tissue probability maps that represent the average shape of a number of subjects' brain images. The procedure is formulated as finding maximum a posteriori estimates within a probabilistic generative model. Estimating the parameters involves alternating between estimating the deformations that match tissue class images of individual subjects to template, and updating the template according to the latest estimates of the deformations. A multinomial matching criterion is used, such that multiple tissue class images (e.g. grey and white matter) are registered simultaneously with the current template estimate. In order to generalise the resulting template to a broader range of subjects, a template blurriness prior is included within the model.

This paper introduces a multivariate Bayesian (MVB) scheme to decode or recognise brain states from neuroimages. It resolves the ill-posed many-to-one mapping, from voxel values or data features to a target variable, using a parametric empirical or hierarchical Bayesian model. This model is inverted using standard variational techniques, in this case expectation maximisation, to furnish the model evidence and the conditional density of the model's parameters. This allows one to compare different models or hypotheses about the mapping from functional or structural anatomy to perceptual and behavioural consequences (or their deficits). We frame this approach in terms of decoding measured brain states to predict or classify outcomes using the rhetoric established in pattern classification of neuroimaging data. However, the aim of MVB is not to predict (because the outcomes are known) but to enable inference on different models of structure-function mappings; such as distributed and sparse representations. This allows one to optimise the model itself and produce predictions that outperform standard pattern classification approaches, like support vector machines.Technically, the model inversion and inference uses the same empirical Bayesian procedures developed for ill-posed inverse problems (e.g., source reconstruction in EEG). However, the MVB scheme used here extends this approach to include a greedy search for sparse solutions. It reduces the problem to the same form used in Gaussian process modelling, which affords a generic and efficient scheme for model optimisation and evaluating model evidence. We illustrate MVB using simulated and real data, with a special focus on model comparison; where models can differ in the form of the mapping (i.e., neuronal representation) within one region, or in the (combination of) regions per se. (C) 2007 Elsevier Inc. All rights reserved.

This paper describes DARTEL, which is an algorithm for diffeomorphic image registration. It is implemented for both 2D and 3D image registration and has been formulated to include an option for estimating inverse consistent deformations. Nonlinear registration is considered as a local optimisation problem, which is solved using a Levenberg-Marquardt strategy. The necessary matrix solutions are obtained in reasonable time using a multigrid method. A constant Eulerian velocity framework is used, which allows a rapid scaling and squaring method to be used in the computations. DARTEL has been applied to intersubject registration of 471 whole brain images, and the resulting deformations were evaluated in terms of how well they encode the shape information necessary to separate male and female subjects and to predict the ages of the subjects.

The analysis of functional magnetic resonance imaging (fMRI) time-series data can provide information not only about task-related activity, but also about the connectivity (functional or effective) among regions and the influences of behavioral or physiologic states on that connectivity. Similar analyses have been performed in other imaging modalities, such as positron emission tomography. However, fMRI is unique because the information about the underlying neuronal activity is filtered or convolved with a hemodynamic response function. Previous studies of regional connectivity in fMRI have overlooked this convolution and have assumed that the observed hemodynamic response approximates the neuronal response. In this article, this assumption is revisited using estimates of underlying neuronal activity. These estimates use a parametric empirical Bayes formulation for hemodynamic deconvolution.

Pure autonomic failure (PAF) is characterized by an acquired, selective, peripheral denervation of the autonomic nervous system. Patients with PAF fail to generate bodily states of arousal via the autonomic nervous system in response to physical or cognitive effort. We used voxel-based morphometry to test the hypothesis that changes in the morphology of brain regions involved in autonomic control would arise as a consequence to the longstanding absence of peripheral autonomic responses in PAF patients. Optimized voxel-based morphometry of structural magnetic resonance scans was used to test for regional differences in grey and white matter in 15 PAF patients and matched controls. There were no group differences observed in global measures of grey matter, white matter, or cerebrospinal fluid (CSF). We identified morphometric differences reflecting regional decreases in grey matter volume and concentration in anterior cingulate and insular cortices in PAF patients relative to controls. Morphometric differences in brainstem and subcortical regions did not reach statistical significance. Our findings suggest that peripheral autonomic denervation is associated with grey matter loss in cortical regions encompassing areas that we have previously shown are functionally involved in generation and representation of bodily states of autonomic arousal. The nature of these changes cannot be determined from morphometric analysis alone, but we suggest that they reflect experience-dependent change consequent upon loss of afferent input to brain regions involved in representation of autonomic states.

We investigated the accuracy of spatial basis function normalization using anatomical landmarks to determine how precisely homologous regions are colocalized. We examined precision in terms of: (1) the number of nonlinear basis functions used by the normalization procedure; (2) the degree of (Bayesian) regularization; and (3) the effect of substituting different templates and how this interacted with the number of basis functions. The face validity of spatial normalization was assessed as a function of these parameters, using the colocalization of homologous landmarks in a test sample of 20 normally developing children and 5 children with bilateral hippocampal pathology. Our results suggest that when optimal normalization parameters are used, anatomical landmarks in the medial temporal lobes are colocalized to within a standard deviation of about 1 mm. When suboptimal parameters are used this standard deviation can increase up to 3 mm. Interestingly the optimal parameters are those that provide a rather constrained normalization as opposed to those that optimize intensity matching at the expense of rendering the warps “unlikely.” The implications of our results, for users of voxel-based morphometry, are discussed.

Diffusion tensor imaging is often performed by acquiring a series of diffusion-weighted spin-echo echo-planar images with different direction diffusion gradients. A problem of echo-planar images is the geometrical distortions that obtain near junctions between tissues of differing magnetic susceptibility. This results in distorted diffusion-tensor maps. To resolve this we suggest acquiring two images for each diffusion gradient; one with bottom-up and one with top-down traversal of k-space in the phase-encode direction. This achieves the simultaneous goals of providing information on the underlying displacement field and intensity maps with adequate spatial sampling density even in distorted areas. The resulting DT maps exhibit considerably higher geometric fidelity, as assessed by comparison to an image volume acquired using a conventional 3D MR technique.

This paper reviews hierarchical observation models, used in functional neuroimaging, in a Bayesian light. It emphasizes the common ground shared by classical and Bayesian methods to show that conventional analyses of neuroimaging data can be usefully extended within an empirical Bayesian framework. In particular we formulate the procedures used in conventional data analysis in terms of hierarchical linear models and establish a connection between classical inference and parametric empirical Bayes (PEB) through covariance component estimation. This estimation is based on an expectation maximization or EM algorithm. The key point is that hierarchical models not only provide for appropriate inference at the highest level but that one can revisit lower levels suitably equipped to make Bayesian inferences. Bayesian inferences eschew many of the difficulties encountered with classical inference and characterize brain responses in a way that is more directly predicated on what one is interested in. The motivation for Bayesian approaches is reviewed and the theoretical background is presented in a way that relates to conventional methods, in particular restricted maximum likelihood (ReML). This paper is a technical and theoretical prelude to subsequent papers that deal with applications of the theory to a range of important issues in neuroimaging. These issues include; (i) Estimating nonsphericity or variance components in fMRI time-series that can arise from serial correlations within subject, or are induced by multisubject (i.e., hierarchical) studies. (ii) Spatiotemporal Bayesian models for imaging data, in which voxels-specific effects are constrained by responses in other voxels. (iii) Bayesian estimation of nonlinear models of hemodynamic responses and (iv) principled ways of mixing structural and functional priors in EEG source reconstruction. Although diverse, all these estimation problems are accommodated by the PEB framework described in this paper.

This article has been written in response to Dr. Fred L. Bookstein's article entitled '"Voxel-Based Morphometry" Should Not Be Used with Imperfectly Registered Images' in this issue of NeuroImage. We will address three main issues: (i) Dr. Bookstein appears to have misunderstood the objective of voxel-based morphometry (VBM) and the nature of the continuum we referred to. (ii) We agree with him when he states that findings from VBM can pertain to systematic registration errors during spatial normalization. (iii) His argument about voxelwise tests on smooth data holds in the absence of error variance, but is of no consequence when using actual data. We first review the tenets of VBM, paying particular attention to the relationship between VBM and tensor-based morphometry. The last two sections of this response deal with the specific concerns raised by Dr. Bookstein.

Primates can give behavioral responses on the basis of arbitrary, context-dependent rules. When sensory instructions and behavioral responses are associated by arbitrary rules, these rules need to be learned. This study investigates the temporal dynamics of functional segregation at the basis of visuomotor associative learning in humans, isolating specific learning-related changes in neurovascular activity across the whole brain. We have used fMRI to measure human brain activity during performance of two tasks requiring the association of visual patterns with motor responses. Both tasks were learned by trial and error, either before (visuomotor control) or during (visuomotor learning) the scanning session. Epochs of tasks performance ( approximately 30 s) were alternated with a baseline period over the whole scanning session ( approximately 50 min). We have assessed both linear and nonlinear modulations in the differential signal between tasks, independently from overall task differences. The performance indices of the visuomotor learning task smoothly converged onto the values of a steady-state control condition, according to nonlinear timecourses. Specific visuomotor learning-related activity has been found over a distributed cortical network, centred on a temporo-prefrontal circuit. These cortical time-modulated activities were supported early in learning by the hippocampal/parahippocampal complex, and late in learning by the basal ganglia system. These findings suggest the inferior temporal and the ventral prefrontal cortex are critical neural nodes for integrating perceptual information with executive processes.

Voxel-based-morphometry (VBM) is a whole-brain, unbiased technique for characterizing regional cerebral volume and tissue concentration differences in structural magnetic resonance images. We describe an optimized method of VBM to examine the effects of age on grey and white matter and CSF in 465 normal adults. Global grey matter volume decreased linearly with age, with a significantly steeper decline in males. Local areas of accelerated loss were observed bilaterally in the insula, superior parietal gyri, central sulci, and cingulate sulci. Areas exhibiting little or no age effect (relative preservation) were noted in the amygdala, hippocampi, and entorhinal cortex. Global white matter did not decline with age, but local areas of relative accelerated loss and preservation were seen. There was no interaction of age with sex for regionally specific effects. These results corroborate previous reports and indicate that VBM is a useful technique for studying structural brain correlates of ageing through life in humans.

In studies of patients with focal brain lesions, it is often useful to coregister an image of the patient's brain to that of another subject or a standard template. We refer to this process as spatial normalization. Spatial normalization can improve the presentation and analysis of lesion location in neuropsychological studies; it can also allow other data, for example from functional imaging, to be compared to data from other patients or normal controls. In functional imaging, the standard procedure for spatial normalization is to use an automated algorithm, which minimizes a measure of difference between image and template, based on image intensity values. These algorithms usually optimize both linear (translations, rotations, zooms, and shears) and nonlinear transforms. In the presence of a focal lesion, automated algorithms attempt to reduce image mismatch between template and image at the site of the lesion. This can lead to significant inappropriate image distortion, especially when nonlinear transforms are used. One solution is to use cost-function masking—masking the areas used in the calculation of image difference—to exclude the area of the lesion, so that the lesion does not bias the transformations. We introduce and evaluate this technique using normalizations of a selection of brains with focal lesions and normal brains with simulated lesions. Our results suggest that cost-function masking is superior to the standard approach to this problem, which is affine-only normalization; we propose that cost-function masking should be used routinely for normalizations of brains with focal lesions.

The global activity is an important confound when analyzing PET data in that its inclusion in the statistical model can substantially reduce error variance and increase sensitivity. However, by defining global activity as the average over all voxels one introduces a bias that is collinear with experimental factors. This leads to an underestimation of true activations and the introduction of artefactual deactivations. We propose a novel estimator for the global activity based on the notion of finding a maximally nonlocal mode in a multivariate characterization of the data, while maximizing the locality of the remaining modes. The approach uses singular value decomposition (SVD) to find a provisional set of modes, which are subsequently rotated such that a metric based on the above heuristic is maximized. This metric is a version of the stochastic sign change (SSC) criterion that has been used previously for normalizing medical images with focal defects. The estimator was evaluated on simulated and real functional imaging (PET) data. The simulations show that the bias of the global mean, introduced by focal activations, is reduced by 80--90% with the new estimator. Comparison with a previous unbiased estimator, using the empirical data, yielded similar results. The advantage of the new estimator is that it is not informed of experimental design and relies only on general assumptions regarding the nature of the signal.