Probabilistic Atlas Research Articles

Probabilistic Maps of Visual Topography in Human Cortex

The human visual system contains an array of topographically organized regions. Identifying these regions in individual subjects is a powerful approach to group-level statistical analysis, but this is not always feasible. We addressed this limitation by generating probabilistic maps of visual topographic areas in 2 standardized spaces suitable for use with adult human brains. Using standard fMRI paradigms, we identified 25 topographic maps in a large population of individual subjects (N = 53) and transformed them into either a surface- or volume-based standardized space. Here, we provide a quantitative characterization of the inter-subject variability within and across visual regions, including the likelihood that a given point would be classified as a part of any region (full probability map) and the most probable region for any given point (maximum probability map). By evaluating the topographic organization across the whole of visual cortex, we provide new information about the organization of individual visual field maps and large-scale biases in visual field coverage. Finally, we validate each atlas for use with independent subjects. Overall, the probabilistic atlases quantify the variability of topographic representations in human cortex and provide a useful reference for comparing data across studies that can be transformed into these standard spaces.

Three-dimensional semiautomatic liver segmentation method for non-contrast computed tomography based on a correlation map of locoregional histogram and probabilistic atlas

BackgroundWe sought to evaluate a new regional segmentation method for use with three-dimensional (3D) non-contrast abdominal CT images and to report the preliminary results. MethodsThe proposed method was evaluated in ten cases. Manually segmented areas were used as the gold standard for evaluation. To compare the standard and the extracted liver regions, the degree of coincidence R% was redefined by transforming a volumetric overlap error. We also evaluated the influence of varying the density window size in terms of setting the starting points. ResultsWe confirmed in ten cases that our method could segment the liver region more precisely than the conventional method. A size of window 15 voxels was optimal as the starting point in all cases. ConclusionsWe demonstrated the accuracy of a 3D semiautomatic liver segmentation method for non-contrast CT. This method promises to offer radiologists a time-efficient segmentation aid.

Automated segmentation of breast in 3-D MR images using a robust atlas.

This paper presents a robust atlas-based segmentation (ABS) algorithm for segmentation of the breast boundary in 3-D MR images. The proposed algorithm combines the well-known methodologies of ABS namely probabilistic atlas and atlas selection approaches into a single framework where two configurations are realized. The algorithm uses phase congruency maps to create an atlas which is robust to intensity variations. This allows an atlas derived from images acquired with one MR imaging sequence to be used to segment images acquired with a different MR imaging sequence and eliminates the need for intensity-based registration. Images acquired using a Dixon sequence were used to create an atlas which was used to segment both Dixon images (intra-sequence) and T1-weighted images (inter-sequence). In both cases, highly accurate results were achieved with the median Dice similarity coefficient values of 94% ±4% and 87 ±6.5%, respectively.

LP13: Activity in the brain network for dynamic manipulation of unstable objects is robust to acute tactile deafferentation: an fMRI study

malized CT images were segmented to identify cranial bones using simple thresholding and morphological operations. A signed distance function was first generated for each CT dataset based on its segmented cranial bones. A shape model for cranial bones was then reconstructed by applying the principal component analysis (PCA) on the signed distance functions. The fontanels were identified as gaps between the cranial bones in the shape model. Using a probabilistic atlas, the intracranial tissues (brain and CSF) were segmented from the normalized and averaged MR images. The scalp was then obtained by removing the segmented tissues from the averaged MR images. Finally, the segmented tissues were surface rendered and used to generate a realistic head model. Results: Figure 1 shows the different compartments of the realistic neonatal head model including brain, CSF, scalp, cranial bones and fontanels (Fig. 1a) as well as their 3D reconstruction (Fig. 1b). Conclusion: In this paper we developed a method to create a realistic neonatal head model based on the co-registration of CT and MR images from neonates. The model included the geometry of brain, CSF, scalp, cranial bones and fontanels. With a specific conductivity value assigned to each compartment, the model can be used for EEG/MEG source localization in neonates.

Tackling Duhemian Problems: An Alternative to Skepticism of Neuroimaging in Philosophy of Cognitive Science

Duhem’s problem arises especially in scientific contexts where the tools and procedures of measurement and analysis are numerous and complex. Several philosophers of cognitive science have cited its manifestations in fMRI as grounds for skepticism regarding the epistemic value of neuroimaging. To address these Duhemian arguments for skepticism, I offer an alternative approach based on Deborah Mayo’s error-statistical account in which Duhem's problem is more fruitfully approached in terms of error probabilities. This is illustrated in examples such as the use of probabilistic brain atlases, comparison of different preprocessing protocols with respect to their error characteristics, and statistical modeling of fMRI data. These examples demonstrate the ways in which we can better understand and formulate the general methodological problem and direct the way toward alternative approaches to neuroimaging in philosophy of cognitive science, in which we can be more balanced and productive in our scrutiny of the epistemic value of neuroimaging studies.

Evaluation of supervised methods for the classification of major tissues and subcortical structures in multispectral brain magnetic resonance images

This work investigates the capability of supervised classification methods in detecting both major tissues and subcortical structures using multispectral brain magnetic resonance images. First, by means of a realistic digital brain phantom, we investigated the classification performance of various Discriminant Analysis methods, K-Nearest Neighbor and Support Vector Machine. Then, using phantom and real data, we quantitatively assessed the benefits of integrating anatomical information in the classification, in the form of voxels coordinates as additional features to the intensities or tissue probabilistic atlases as priors. In addition we tested the effect of spatial correlations between neighboring voxels and image denoising. For each brain tissue we measured the classification performance in terms of global agreement percentage, false positive and false negative rates and kappa coefficient. The effectiveness of integrating spatial information or a tissue probabilistic atlas has been demonstrated for the aim of accurately classifying brain magnetic resonance images.

Quantifying inter-individual anatomical variability in the subcortex using 7 T structural MRI

Functional magnetic resonance imaging (MRI) data are usually registered into standard anatomical space. However, standard atlases, such as LPBA40, the Harvard–Oxford atlas, FreeSurfer, and the Jülich cytoarchitectonic maps all lack important detailed information about small subcortical structures like the substantia nigra and subthalamic nucleus. Here we introduce a new subcortical probabilistic atlas based on ultra-high resolution in-vivo anatomical imaging from 7T MRI. The atlas includes six important but elusive subcortical nuclei: the striatum, the globus pallidus internal and external segment (GPi/e), the subthalamic nucleus, the substantia nigra, and the red nucleus. With a sample of 30 young subjects and carefully cross-validated delineation protocols, our atlas is able to capture the anatomical variability within healthy populations for each of the included structures at an unprecedented level of detail. All the generated probabilistic atlases are registered to MNI standard space and are publicly available.

Bayesian estimation of probabilistic atlas for tissue segmentation

Automatic anatomical brain image segmentation is still a challenge. In particular, algorithms have to address the partial volume effect (PVE) as well as the variability of the gray level of internal brain structures which may appear closer to gray matter (GM) than white matter (WM). Atlas based segmentation is one solution as it brings prior information. For such tasks, probabilistic atlases are very useful as they take account of the PVE information. In this paper, we provide a detailed analysis of a generative statistical model based on dense deformable templates that represents several tissue types observed in medical images. The inputs are gray level data whereas our atlas is composed of both an estimation of the deformation metric and probability maps of each tissue (called class). This atlas is used to guide the tissue segmentation of new images. Experiments are shown on brain T1 MRI datasets. This method only requires approximate pre-registration, as the latter is done jointly with the segmentation. Note however that an approximate registration is a reasonable pre-requisite given the application.

Multi-organ localization combining global-to-local regression and confidence maps.

We propose a method for fast, accurate and robust localization of several organs in medical images. We generalize global-to-local cascades of regression forests [1] to multiple organs. A first regressor encodes global relationships between organs. Subsequent regressors refine the localization of each organ locally and independently for improved accuracy. We introduce confidence maps, which incorporate information about both the regression vote distribution and the organ shape through probabilistic atlases. They are used within the cascade itself, to better select the test voxels for the second set of regressors, and to provide richer information than the classical bounding boxes thanks to the shape prior. We demonstrate the robustness and accuracy of our approach through a quantitative evaluation on a large database of 130 CT volumes.

Differentially disrupted functional connectivity of the subregions of the inferior parietal lobule in Alzheimer’s disease

Recent research on Alzheimer's disease (AD) has shown that the altered structure and function of the inferior parietal lobule (IPL) provides a promising indicator of AD. However, little is known about the functional connectivity of the IPL subregions in AD subjects. In this study, we collected resting-state functional magnetic resonance imaging data from 32 AD patients and 38 healthy controls. We defined seven subregions of the IPL according to probabilistic cytoarchitectonic atlases and mapped the whole-brain resting-state functional connectivity for each subregion. Using hierarchical clustering analysis, we identified three distinct functional connectivity patterns of the IPL subregions: the anterior IPL connected with the sensorimotor network (SMN) and salience network (SN); the central IPL had connectivity with the executive-control network (ECN); and the posterior IPL exhibited connections with the default-mode network (DMN). Compared with the controls, the AD patients demonstrated distinct disruptive patterns of the IPL subregional connectivity with these different networks (SMN, SN, ECN and DMN), which suggests the impairment of the functional integration in the IPL. Notably, we also observed that the IPL subregions showed increased connectivity with the posterior part of the DMN in AD patients, which potentially indicates a compensatory mechanism. Finally, these abnormal IPL functional connectivity changes were closely associated with cognitive performance. Collectively, we show that the subregions of the IPL present distinct functional connectivity patterns with various functional networks that are differentially impaired in AD patients. Our results also suggest that functional disconnection and compensation in the IPL may coexist in AD.

A comparison of three brain atlases for MCI prediction

BackgroundAlthough previous voxel-based studies using features extracted by atlas-based parcellation produced relatively poor performances on the prediction of Alzheimer's disease (AD) in subjects with mild cognitive impairment (MCI), classification performance usually depends on features extracted from the original images by atlas-based parcellation. To establish whether classification performance differs depending on the choice of atlases, support vector machine (SVM)-based classification using different brain atlases was performed. New methodSeventy-seven three-dimensional T1-weighted MRI data sets of subjects with amnestic MCI, including 39 subjects who developed AD (MCI-C) within three years and 38 who did not (MCI-NC), were used for voxel-based morphometry (VBM) analyses and analyzed using SVM-based pattern recognition methods combined with a feature selection method based on the SVM recursive feature elimination (RFE) method. Three brain atlases were used for the feature selections: the Automated Anatomical Labeling (AAL) Atlas, Brodmann's Areas (BA), and the LONI Probabilistic Brain Atlas (LPBA40). ResultsThe VBM analysis showed a significant cluster of gray matter density reduction, located at the left hippocampal region, in MCI-C compared to MCI-NC. The SVM analyses with the SVM-RFE algorithm revealed that the best classification performance was achieved by LPBA40 with 37 selected features, giving an accuracy of 77.9%. The overall performance in LPBA40 was better than that of AAL and BA regardless of the number of selected features. ConclusionsThese results suggest that feature selection is crucial to improve the classification performance in atlas-based analysis and that the choice of atlases is also important.

Construction of an in vivo human spinal cord atlas based on high-resolution MR images at cervical and thoracic levels: preliminary results

Our goal was to build a probabilistic atlas and anatomical template of the human cervical and thoracic spinal cord (SC) that could be used for segmentation algorithm improvement, parametric group studies, and enrichment of biomechanical modelling. High-resolution axial T2*-weighted images were acquired at 3T on 15 healthy volunteers using a multi-echo-gradient-echo sequence (1 slice per vertebral level from C1 to L2). After manual segmentation, linear and affine co-registrations were performed providing either inter-individual morphometric variability maps, or substructure probabilistic maps [CSF, white and grey matter (WM/GM)] and anatomical SC template. The larger inter-individual morphometric variations were observed at the thoraco-lumbar levels and in the posterior GM. Mean SC diameters were in agreement with the literature and higher than post-mortem measurements. A representative SC MR template was generated and values up to 90 and 100% were observed on GM and WM-probability maps. This work provides a probabilistic SC atlas and a template that could offer great potentialities for parametrical MRI analysis (DTI/MTR/fMRI) and group studies, similar to what has already been performed using a brain atlas. It also offers great perspective for biomechanical models usually based on post-mortem or generic data. Further work will consider integration into an automated SC segmentation pipeline.

Registration of 3D fetal neurosonography and MRI.

We propose a method for registration of 3D fetal brain ultrasound with a reconstructed magnetic resonance fetal brain volume. This method, for the first time, allows the alignment of models of the fetal brain built from magnetic resonance images with 3D fetal brain ultrasound, opening possibilities to develop new, prior information based image analysis methods for 3D fetal neurosonography. The reconstructed magnetic resonance volume is first segmented using a probabilistic atlas and a pseudo ultrasound image volume is simulated from the segmentation. This pseudo ultrasound image is then affinely aligned with clinical ultrasound fetal brain volumes using a robust block-matching approach that can deal with intensity artefacts and missing features in the ultrasound images. A qualitative and quantitative evaluation demonstrates good performance of the method for our application, in comparison with other tested approaches. The intensity average of 27 ultrasound images co-aligned with the pseudo ultrasound template shows good correlation with anatomy of the fetal brain as seen in the reconstructed magnetic resonance image.

Overlapping and distinct gray and white matter abnormalities in schizophrenia and bipolar I disorder

Schizophrenia and bipolar disorder may share common neurobiological mechanisms, but few studies have directly compared gray and white matter structure in these disorders. We used diffusion-weighted magnetic resonance imaging and a region of interest based analysis to identify overlapping and distinct gray and white matter abnormalities in 35 patients with schizophrenia and 20 patients with bipolar I disorder in comparison to 56 healthy volunteers. We examined fractional anisotropy within the white matter and mean diffusivity within the gray matter in 42 regions of interest defined on a probabilistic atlas following non-linear registration of the images to atlas space. Patients with schizophrenia had significantly lower fractional anisotropy in temporal (superior temporal and parahippocampal) and occipital (superior and middle occipital) white matter compared to patients with bipolar disorder and healthy volunteers. By contrast, both patient groups demonstrated significantly higher mean diffusivity in frontal (inferior frontal and lateral orbitofrontal) and temporal (superior temporal and parahippocampal) gray matter compared to healthy volunteers, but did not differ from each other. Our study implicates overlapping gray matter frontal and temporal lobe structural alterations in the neurobiology of schizophrenia and bipolar I disorder, but suggests that temporal and occipital lobe white matter deficits may be an additional risk factor for schizophrenia. Our findings may have relevance for future diagnostic classification systems and the identification of susceptibility genes for these disorders.

Unsupervised Segmentation, Clustering, and Groupwise Registration of Heterogeneous Populations of Brain MR Images

Population analysis of brain morphology from magnetic resonance images contributes to the study and understanding of neurological diseases. Such analysis typically involves segmentation of a large set of images and comparisons of these segmentations between relevant subgroups of images (e.g., "normal" versus "diseased"). The images of each subgroup are usually selected in advance in a supervised way based on clinical knowledge. Their segmentations are typically guided by one or more available atlases, assumed to be suitable for the images at hand. We present a data-driven probabilistic framework that simultaneously performs atlas-guided segmentation of a heterogeneous set of brain MR images and clusters the images in homogeneous subgroups, while constructing separate probabilistic atlases for each cluster to guide the segmentation. The main benefits of integrating segmentation, clustering and atlas construction in a single framework are that: 1) our method can handle images of a heterogeneous group of subjects and automatically identifies homogeneous subgroups in an unsupervised way with minimal prior knowledge, 2) the subgroups are formed by automatical detection of the relevant morphological features based on the segmentation, 3) the atlases used by our method are constructed from the images themselves and optimally adapted for guiding the segmentation of each subgroup, and 4) the probabilistic atlases represent the morphological pattern that is specific for each subgroup and expose the groupwise differences between different subgroups. We demonstrate the feasibility of the proposed framework and evaluate its performance with respect to image segmentation, clustering and atlas construction on simulated and real data sets including the publicly available BrainWeb and ADNI data. It is shown that combined segmentation and atlas construction leads to improved segmentation accuracy. Furthermore, it is demonstrated that the clusters generated by our unsupervised framework largely coincide with the clinically determined subgroups in case of disease-specific differences in brain morphology and that the differences between the cluster-specific atlases are in agreement with the expected disease-specific patterns, indicating that our method is capable of detecting the different modes in a population. Our method can thus be seen as a comprehensive image-driven population analysis framework that can contribute to the detection of novel subgroups and distinctive image features, potentially leading to new insights in the brain development and disease.

Construction and Evaluation of Quantitative Small-Animal PET Probabilistic Atlases for [18F]FDG and [18F]FECT Functional Mapping of the Mouse Brain

Automated voxel-based or pre-defined volume-of-interest (VOI) analysis of small-animal PET data in mice is necessary for optimal information usage as the number of available resolution elements is limited. We have mapped metabolic ([18F]FDG) and dopamine transporter ([18F]FECT) small-animal PET data onto a 3D Magnetic Resonance Microscopy (MRM) mouse brain template and aligned them in space to the Paxinos co-ordinate system. In this way, ligand-specific templates for sensitive analysis and accurate anatomical localization were created. Next, using a pre-defined VOI approach, test-retest and intersubject variability of various quantification methods were evaluated. Also, the feasibility of mouse brain statistical parametric mapping (SPM) was explored for [18F]FDG and [18F]FECT imaging of 6-hydroxydopamine-lesioned (6-OHDA) mice.MethodsTwenty-three adult C57BL6 mice were scanned with [18F]FDG and [18F]FECT. Registrations and affine spatial normalizations were performed using SPM8. [18F]FDG data were quantified using (1) an image-derived-input function obtained from the liver (cMRglc), using (2) standardized uptake values (SUVglc) corrected for blood glucose levels and by (3) normalizing counts to the whole-brain uptake. Parametric [18F]FECT binding images were constructed by reference to the cerebellum. Registration accuracy was determined using random simulated misalignments and vectorial mismatch determination.ResultsRegistration accuracy was between 0.21–1.11 mm. Regional intersubject variabilities of cMRglc ranged from 15.4% to 19.2%, while test-retest values were between 5.0% and 13.0%. For [18F]FECT uptake in the caudate-putamen, these values were 13.0% and 10.3%, respectively. Regional values of cMRglc positively correlated to SUVglc measured within the 45–60 min time frame (spearman r = 0.71). Next, SPM analysis of 6-OHDA-lesioned mice showed hypometabolism in the bilateral caudate-putamen and cerebellum, and an unilateral striatal decrease in DAT availability.ConclusionMRM-based small-animal PET templates facilitate accurate assessment and spatial localization of mouse brain function using VOI or voxel-based analysis. Regional intersubject- and test-retest variations indicate that for these targets accuracy comparable to humans can be achieved.

A multi-atlas based method for automated anatomical Macaca fascicularis brain MRI segmentation and PET kinetic extraction

MRI templates and digital atlases are needed for automated and reproducible quantitative analysis of non-human primate PET studies. Segmenting brain images via multiple atlases outperforms single-atlas labelling in humans. We present a set of atlases manually delineated on brain MRI scans of the monkey Macaca fascicularis. We use this multi-atlas dataset to evaluate two automated methods in terms of accuracy, robustness and reliability in segmenting brain structures on MRI and extracting regional PET measures. Twelve individual Macaca fascicularis high-resolution 3DT1 MR images were acquired. Four individual atlases were created by manually drawing 42 anatomical structures, including cortical and sub-cortical structures, white matter regions, and ventricles. To create the MRI template, we first chose one MRI to define a reference space, and then performed a two-step iterative procedure: affine registration of individual MRIs to the reference MRI, followed by averaging of the twelve resampled MRIs. Automated segmentation in native space was obtained in two ways: 1) Maximum probability atlases were created by decision fusion of two to four individual atlases in the reference space, and transformation back into the individual native space (MAXPROB)(.) 2) One to four individual atlases were registered directly to the individual native space, and combined by decision fusion (PROPAG). Accuracy was evaluated by computing the Dice similarity index and the volume difference. The robustness and reproducibility of PET regional measurements obtained via automated segmentation was evaluated on four co-registered MRI/PET datasets, which included test-retest data. Dice indices were always over 0.7 and reached maximal values of 0.9 for PROPAG with all four individual atlases. There was no significant mean volume bias. The standard deviation of the bias decreased significantly when increasing the number of individual atlases. MAXPROB performed better when increasing the number of atlases used. When all four atlases were used for the MAXPROB creation, the accuracy of morphometric segmentation approached that of the PROPAG method. PET measures extracted either via automatic methods or via the manually defined regions were strongly correlated, with no significant regional differences between methods. Intra-class correlation coefficients for test-retest data were over 0.87. Compared to single atlas extractions, multi-atlas methods improve the accuracy of region definition. They also perform comparably to manually defined regions for PET quantification. Multiple atlases of Macaca fascicularis brains are now available and allow reproducible and simplified analyses.

Probabilistic atlases for face and biological motion perception: an analysis of their reliability and overlap.

Neuroimaging research has identified several category-selective regions in visual cortex that respond most strongly when viewing an exemplar image from a preferred category, such as faces. Recent studies, however, have suggested a more complex pattern of activation that has been heretofore unrecognized, e.g., the presence of additional patches of activation to faces beyond the well-studied fusiform face area, and the activation of ostensible face selective regions by animate motion of non-biological forms. Here, we characterize the spatial pattern of brain activity evoked by viewing faces or biological motion in large fMRI samples (N>120). We create probabilistic atlases for both face and biological motion activation, and directly compare their spatial patterns of activation. Our findings support the suggestion that the fusiform face area is composed of at least two separable foci of activation. The face-evoked response in the fusiform and nearby ventral temporal cortex has good reliability across runs; however, we found surprisingly high variability in lateral brain regions by faces, and for all brain regions by biological motion, which had an overall much lower effect size. We found that faces and biological motion evoke substantially overlapping activation distributions in both ventral and lateral occipitotemporal cortices. The peaks of activation for these different categories within these overlapping regions were close but distinct.

Editorial introductions

Editorial introductions

Identification of a strategic brain network underlying processing speed deficits in vascular cognitive impairment

Patients with vascular cognitive impairment (VCI) commonly exhibit deficits in processing speed. This has been attributed to a disruption of frontal-subcortical neuronal circuits by ischemic lesions, but the exact mechanisms and underlying anatomical structures are poorly understood. We set out to identify a strategic brain network for processing speed by applying graph-based data-mining techniques to MRI lesion maps from patients with small vessel disease. We studied 235 patients with CADASIL, a genetic small vessel disease causing pure VCI. Using a probabilistic atlas in standard space we first determined the regional volumes of white matter hyperintensities (WMH) and lacunar lesions (LL) within major white matter tracts. Conditional dependencies between the regional lesion volumes and processing speed were then examined using Bayesian network analysis. Exploratory analysis identified a network of five imaging variables as the best determinant of processing speed. The network included LL in the left anterior thalamic radiation and the left cingulum as well as WMH in the left forceps minor, the left parahippocampal white matter and the left corticospinal tract. Together these variables explained 34% of the total variance in the processing speed score. Structural equation modeling confirmed the findings obtained from the Bayesian models. In summary, using graph-based models we identified a strategic brain network having the highest predictive value for processing speed in our cohort of patients with pure small vessel disease. Our findings confirm and extend previous results showing a role of frontal-subcortical neuronal circuits, in particular dorsolateral prefrontal and cingulate circuits, in VCI.