Individual Atlases Research Articles

Spectral Graph Neural Network-Based Multi-Atlas Brain Network Fusion for Major Depressive Disorder Diagnosis.

Major Depressive Disorder (MDD) imposes a substantial burden within the healthcare domain, impacting millions of individuals worldwide. Functional Magnetic Resonance Imaging (fMRI) has emerged as a promising tool for the objective diagnosis of MDD, enabling the investigation of functional connectivity patterns in the brain associated with this disorder. However, most existing methods focus on a single brain atlas, which limits their ability to capture the complex, multi-scale nature of functional brain networks. To address these limitations, we propose a novel multi-atlas fusion method that incorporates early and late fusion in a unified framework. Our method introduces the concept of the holistic Functional Connectivity Network (FCN), which captures both intra-atlas relationships within individual atlases and inter-regional relationships between atlases with different brain parcellation scales. This comprehensive representation enables the identification of potential disease-related patterns associated with MDD in the early stage of our framework. Moreover, by decoding the holistic FCN from various perspectives through multiple spectral Graph Convolutional Neural Networks and fusing their results with decision-level ensembles, we further improve the performance of MDD diagnosis. Our approach is easily implemented with minimal modifications to existing model structures and demonstrates a robust performance across different baseline models. Our method, evaluated on public resting-state fMRI datasets, surpasses the current multi-atlas fusion methods, enhancing the accuracy of MDD diagnosis. The proposed novel multi-atlas fusion framework provides a more reliable MDD diagnostic technique. Experimental results show our approach outperforms both single- and multi-atlas-based methods, demonstrating its effectiveness in advancing MDD diagnosis.

Predicting gene regulatory networks from cell atlases.

Recent single-cell RNA-sequencing atlases have surveyed and identified major cell types across different mouse tissues. Here, we computationally reconstruct gene regulatory networks from three major mouse cell atlases to capture functional regulators critical for cell identity, while accounting for a variety of technical differences, including sampled tissues, sequencing depth, and author assigned cell type labels. Extracting the regulatory crosstalk from mouse atlases, we identify and distinguish global regulons active in multiple cell types from specialised cell type-specific regulons. We demonstrate that regulon activities accurately distinguish individual cell types, despite differences between individual atlases. We generate an integrated network that further uncovers regulon modules with coordinated activities critical for cell types, and validate modules using available experimental data. Inferring regulatory networks during myeloid differentiation from wild-type and Irf8 KO cells, we uncover functional contribution of Irf8 regulon activity and composition towards monocyte lineage. Our analysis provides an avenue to further extract and integrate the regulatory crosstalk from single-cell expression data.

Improved segmentation of cerebellar structures in children

BackgroundConsistent localization of cerebellar cortex in a standard coordinate system is important for functional studies and detection of anatomical alterations in studies of morphometry. To date, no pediatric cerebellar atlas is available. New methodThe probabilistic Cape Town Pediatric Cerebellar Atlas (CAPCA18) was constructed in the age-appropriate National Institute of Health Pediatric Database asymmetric template space using manual tracings of 16 cerebellar compartments in 18 healthy children (9-13 years) from Cape Town, South Africa. The individual atlases of the training subjects were also used to implement multi atlas label fusion using multi atlas majority voting (MAMV) and multi atlas generative model (MAGM) approaches. Segmentation accuracy in 14 test subjects was compared for each method to ‘gold standard’ manual tracings. ResultsSpatial overlap between manual tracings and CAPCA18 automated segmentation was 73% or higher for all lobules in both hemispheres, except VIIb and X. Automated segmentation using MAGM yielded the best segmentation accuracy over all lobules (mean Dice Similarity Coefficient 0.76; range 0.55-0.91; mean Hausdorff distance 0.9mm; range 0.8-2.7mm). Comparison with existing methodsIn all lobules, spatial overlap of CAPCA18 segmentations with manual tracings was similar or higher than those obtained with SUIT (spatially unbiased infra-tentorial template), providing additional evidence of the benefits of an age appropriate atlas. MAGM segmentation accuracy was comparable to values reported recently by Park et al. (Neuroimage 2014;95(1):217) in adults (across all lobules mean DSC=0.73, range 0.40-0.89). ConclusionsCAPCA18 and the associated multi-subject atlases of the training subjects yield improved segmentation of cerebellar structures in children.

Evaluation of Hippocampal Volume Based on Various Inversion Time in Normal Adults by Manual Tracing and Automated Segmentation Methods

Magnetic resonance imaging (MRI) enables non-invasive examinations of the Purpose: To investigate the value of image post-processing software (FreeSurfer, IBASPM [individual brain atlases using statistical parametric mapping software]) and inversion time (TI) in volumetric analyses of the hippocampus and to identify differences in comparison with manual tracing. Materials and Methods: Brain images from 12 normal adults were acquired using magnetization prepared rapid acquisition gradient echo (MPRAGE) with a slice thickness of 1.3 mm and TI of 800, 900, 1000, and 1100 ms. Hippocampal volumes were measured using FreeSurfer, IBASPM and manual tracing. Statistical differences were examined using correlation analyses accounting for spatial interpretations percent volume overlap and percent volume difference. Results: FreeSurfer revealed a maximum percent volume overlap and maximum percent volume difference at TI = 800 ms (77.1 ± 2.9%) and TI = 1100 ms (13.1 ± 2.1%), respectively. The respective values for IBASPM were TI = 1100 ms (55.3 ± 9.1%) and TI = 800 ms (43.1 ± 10.7%). FreeSurfer presented a higher correlation than IBASPM but it was not statistically significant. Conclusion: FreeSurfer performed better in volumetric determination than IBASPM. Given the subjective nature of manual tracing, automated image acquisition and analysis image is accurate and preferable.

Encoding atlases by randomized classification forests for efficient multi-atlas label propagation.

We propose a method for multi-atlas label propagation (MALP) based on encoding the individual atlases by randomized classification forests. Most current approaches perform a non-linear registration between all atlases and the target image, followed by a sophisticated fusion scheme. While these approaches can achieve high accuracy, in general they do so at high computational cost. This might negatively affect the scalability to large databases and experimentation. To tackle this issue, we propose to use a small and deep classification forest to encode each atlas individually in reference to an aligned probabilistic atlas, resulting in an Atlas Forest (AF). Our classifier-based encoding differs from current MALP approaches, which represent each point in the atlas either directly as a single image/label value pair, or by a set of corresponding patches. At test time, each AF produces one probabilistic label estimate, and their fusion is done by averaging. Our scheme performs only one registration per target image, achieves good results with a simple fusion scheme, and allows for efficient experimentation. In contrast to standard forest schemes, in which each tree would be trained on all atlases, our approach retains the advantages of the standard MALP framework. The target-specific selection of atlases remains possible, and incorporation of new scans is straightforward without retraining. The evaluation on four different databases shows accuracy within the range of the state of the art at a significantly lower running time.

Through diffusion tensor magnetic resonance imaging to evaluate the original properties of neural pathways of patients with partial seizures and secondary generalization by individual anatomic reference atlas.

To investigate white matter (WM) abnormalities in neocortical epilepsy, we extract supratentorial WM parameters from raw tensor magnetic resonance images (MRI) with automated region-of-interest (ROI) registrations. Sixteen patients having neocortical seizures with secondarily generalised convulsions and 16 age-matched normal subjects were imaged with high-resolution and diffusion tensor MRIs. Automated demarcation of supratentorial fibers was accomplished with personalized fiber-labeled atlases. From the individual atlases, we observed significant elevation of mean diffusivity (MD) in fornix (cres)/stria terminalis (FX/ST) and sagittal stratum (SS) and a significant difference in fractional anisotropy (FA) among FX/ST, SS, posterior limb of the internal capsule (PLIC), and posterior thalamic radiation (PTR). For patients with early-onset epilepsy, the diffusivities of the SS and the retrolenticular part of the internal capsule were significantly elevated, and the anisotropies of the FX/ST and SS were significantly decreased. In the drug-resistant subgroup, the MDs of SS and PTR and the FAs of SS and PLIC were significantly different. Onset age was positively correlated with increases in FAs of the genu of the corpus callosum. Patients with neocortical seizures and secondary generalisation had microstructural anomalies in WM. The changes in WM are relevant to early onset, progression, and severity of epilepsy.

From the Jar to the World Wide Web: Designing a Public Digital Library for the Human Brain

The development of the Digital Brain Library (DBL), envisaged as a permanent archive that describes the spectrum of normal brain maturation, aging and neurological disease, is described. The DBL is composed of individual brain atlases, constructed by the combined techniques of magnetic resonance imaging (MRI) and computer-controlled microscopy to show the brain at multiple levels of resolution, in 2-D and 3-D. The archive contains quantitative parameters that are linked to the results of neuropsychological and attitudinal tests as well as biographical records obtained from the individuals who donated the brain to the collection. Data exploration is supported by statistical tools and web technologies that maximize the utilization of the collection at different levels of expertise.

A unified framework for cross-modality multi-atlas segmentation of brain MRI

Multi-atlas label fusion is a powerful image segmentation strategy that is becoming increasingly popular in medical imaging. A standard label fusion algorithm relies on independently computed pairwise registrations between individual atlases and the (target) image to be segmented. These registrations are then used to propagate the atlas labels to the target space and fuse them into a single final segmentation. Such label fusion schemes commonly rely on the similarity between intensity values of the atlases and target scan, which is often problematic in medical imaging - in particular, when the atlases and target images are obtained via different sensor types or imaging protocols. In this paper, we present a generative probabilistic model that yields an algorithm for solving the atlas-to-target registrations and label fusion steps simultaneously. The proposed model does not directly rely on the similarity of image intensities. Instead, it exploits the consistency of voxel intensities within the target scan to drive the registration and label fusion, hence the atlases and target image can be of different modalities. Furthermore, the framework models the joint warp of all the atlases, introducing interdependence between the registrations. We use variational expectation maximization and the Demons registration framework in order to efficiently identify the most probable segmentation and registrations. We use two sets of experiments to illustrate the approach, where proton density (PD) MRI atlases are used to segment T1-weighted brain scans and vice versa. Our results clearly demonstrate the accuracy gain due to exploiting within-target intensity consistency and integrating registration into label fusion.

A Magnetic Resonance Image Based Atlas of the Rabbit Brain for Automatic Parcellation

Rabbit brain has been used in several works for the analysis of neurodevelopment. However, there are not specific digital rabbit brain atlases that allow an automatic identification of brain regions, which is a crucial step for various neuroimage analyses, and, instead, manual delineation of areas of interest must be performed in order to evaluate a specific structure. For this reason, we propose an atlas of the rabbit brain based on magnetic resonance imaging, including both structural and diffusion weighted, that can be used for the automatic parcellation of the rabbit brain. Ten individual atlases, as well as an average template and probabilistic maps of the anatomical regions were built. In addition, an example of automatic segmentation based on this atlas is described.

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.

Modeling Regional Sensitivity of Radiation Pneumonitis using Multi-institutional EUD Atlases from Phase I Dose Escalation Protocols for Non-small Cell Lung Cancer

Materials/Methods: Doses ranged from 57.6 - 90 Gy in 1.8 - 2 Gy fractions (MSKCC) and 60.75 - 94.5 Gy in 2.25 Gy fractions (NKI). RP was scored prospectively as requiring steroids or worse toxicity at 6 months. Seventy-eight and 86 patients met followup requirements, 10 and 14 with RP, respectively. EUD was calculated from DVHs normalized to 2 Gy fractions (linear quadratic model witha/b = 3 Gy) for whole, contralateral (contra), ipsilateral (ipsi), superior (sup), and inferior (inf) lungs; the volume effect parameter ranged from log10a = -1.0 to +1.0 in steps of 0.1. The atlas EUD grid was 0.5 Gy. All 78 MSKCC patients had dose volume histograms (DVH)for all 5 lung regions; 86, 53, 53, 49, and 49 NKI patientshad correspondingDVHs. Logistic regression was used in each lung region to find significant (p # 0.05) EUD responses for RP. Consistency was checked between both individual and combined atlases, and atlas based and direct EUD analysis. Results: EUD responses from the combined data from whole, ipsi, and inf lung regions gave the best models, and were significantly correlated with RP when log10a\0.2, 0.0, 0.6, respectively. Weak significant correlation with contra lung EUD occurred for log10a\-0.8. There was no significant correlation for sup lungs. Qualitative dependence of model p values on log10a was similar for individual and combined atlases, implying consistency of the individual data sets. Combining atlases decreased p values of the whole lung, ipsi, and inf lungs. For contra lung, only combined data gave significant correlation. Atlas and direct EUD analysis results were consistent, indicating minimal information loss from using atlases. Conclusions: EUDs of the whole, ipsi, and inf lung are significantly correlated with RP. Sensitivity of the different lung regions is similar in individual and combined data sets. Combining EUD atlases from two institutions improved modeling of the relationship of EUD with RP. Calculations based on the atlas or directly on EUD data are consistent. Supported in part by NIH R01CA129182-01A2.

Statistical neuroanatomy of the human inferior frontal gyrus and probabilistic atlas in a standard stereotaxic space.

We manually defined the inferior frontal gyrus (IFG) on high-resolution MRIs in native space in 30 healthy subjects (15 female, median age 31 years; 15 male, median age 30 years), resulting in 30 individual atlases. Using standard software (SPM99), these were spatially transformed to a widely used stereotaxic space (MNI/ICBM 152) to create probabilistic maps. In native space, the total IFG volume was on average 5%, and the gray matter (GM) portion 12% larger in women (not significant). Expressed as a percentage of ipsilateral frontal lobe volume (i.e., correcting for brain size), the IFG was an average of 20%, and the GM portion of the IFG 27%, larger in women (P < 0.005). Correcting for total lobar volume yielded the same result. No asymmetry was found in IFG volumes. There were significant positional differences between the right and left IFGs, with the right IFG being further lateral in both native and stereotaxic space. Variability was similar on the left and right, but more pronounced anteriorly and superiorly. We show differences in IFG volume, composition, and position between sexes and between hemispheres. Applications include probabilistic determination of location in group studies, automatic labeling of new scans, and detection of anatomical abnormalities in patients.

Mindboggle: Automated brain labeling with multiple atlases

BackgroundTo make inferences about brain structures or activity across multiple individuals, one first needs to determine the structural correspondences across their image data. We have recently developed Mindboggle as a fully automated, feature-matching approach to assign anatomical labels to cortical structures and activity in human brain MRI data. Label assignment is based on structural correspondences between labeled atlases and unlabeled image data, where an atlas consists of a set of labels manually assigned to a single brain image. In the present work, we study the influence of using variable numbers of individual atlases to nonlinearly label human brain image data.MethodsEach brain image voxel of each of 20 human subjects is assigned a label by each of the remaining 19 atlases using Mindboggle. The most common label is selected and is given a confidence rating based on the number of atlases that assigned that label. The automatically assigned labels for each subject brain are compared with the manual labels for that subject (its atlas). Unlike recent approaches that transform subject data to a labeled, probabilistic atlas space (constructed from a database of atlases), Mindboggle labels a subject by each atlas in a database independently.ResultsWhen Mindboggle labels a human subject's brain image with at least four atlases, the resulting label agreement with coregistered manual labels is significantly higher than when only a single atlas is used. Different numbers of atlases provide significantly higher label agreements for individual brain regions.ConclusionIncreasing the number of reference brains used to automatically label a human subject brain improves labeling accuracy with respect to manually assigned labels. Mindboggle software can provide confidence measures for labels based on probabilistic assignment of labels and could be applied to large databases of brain images.

Three-dimensional maximum probability atlas of the human brain, with particular reference to the temporal lobe.

Probabilistic atlases of neuroanatomy are more representative of population anatomy than single brain atlases. They allow anatomical labeling of the results of group studies in stereotaxic space, automated anatomical labeling of individual brain imaging datasets, and the statistical assessment of normal ranges for structure volumes and extents. No such manually constructed atlas is currently available for the frequently studied group of young adults. We studied 20 normal subjects (10 women, median age 31 years) with high-resolution magnetic resonance imaging (MRI) scanning. Images were nonuniformity corrected and reoriented along both the anterior-posterior commissure (AC-PC) line horizontally and the midsagittal plane sagittally. Building on our previous work, we have expanded and refined existing algorithms for the subdivision of MRI datasets into anatomical structures. The resulting algorithm is presented in the Appendix. Forty-nine structures were interactively defined as three-dimensional volumes-of-interest (VOIs). The resulting 20 individual atlases were spatially transformed (normalized) into standard stereotaxic space, using SPM99 software and the MNI/ICBM 152 template. We evaluated volume data for all structures both in native space and after spatial normalization, and used the normalized superimposed atlases to create a maximum probability map in stereotaxic space, which retains quantitative information regarding inter-subject variability. Its potential applications range from the automatic labeling of new scans to the detection of anatomical abnormalities in patients. Further data can be extracted from the atlas for the detailed analysis of individual structures.

Selected Atlases of Cardiovascular Nuclear Medicine

Each volume in the Atlases of Clinical Nuclear Medicine covers one anatomic region or system. Extensively illustrated with superb quality images, each atlas reveals the spectrum of normal scintigraphic findings as well as examples of both common and unusual conditions. Detailed figure legends describe the findings within each image, and most discuss the image's important teaching point. The text, which is descriptive yet concise, covers such topics as procedure technique, dosimetry, physiology, and scan interpretation. The Atlases of Clinical Nuclear Medicine fulfill equally the needs of practicing radiologists, nuclear medicine physicians, and residents seeking to utilize this helpful diagnostic modality more effectively. Selected Atlases of Cardiovascular Nuclear Medicine is comprised of five individual atlases in the areas of Myocardial Perfusion Images, Artifacts in 201 Thallium Spect Cardiac Perfusion Imaging, Artifacts in 99m Technetium Sestamibi SPECT Cardiac Perfusion Imaging, Cardiac Positron Emission Tomography, and Indium-111 Antimyosin Imaging of Myocardial Necrosis.