Regular ArticleHigh-Resolution Random Mesh Algorithms for Creating a Probabilistic 3D Surface Atlas of the Human Brain
Abstract
Striking variations exist, across individuals, in the internal and external geometry of the brain. Such normal variations in the size, orientation, topology, and geometric complexity of cortical and subcortical structures have complicated the problem of quantifying deviations from normal anatomy and of developing standardized neuroanatomical atlases. This paper describes the design, implementation, and results of a technique for creating a three-dimensional (3D) probabilistic surface atlas of the human brain. We have developed, implemented, and tested a new 3D statistical method for assessing structural variations in a database of anatomic images. The algorithm enables the internal surface anatomy of new subjects to be analyzed at an extremely local level. The goal was to quantify subtle and distributed patterns of deviation from normal anatomy by automatically generating detailed probability maps of the anatomy of new subjects. Connected systems of parametric meshes were used to model the internal course of the following structures in both hemispheres: the parieto-occipital sulcus, the anterior and posterior rami of the calcarine sulcus, the cingulate and marginal sulci, and the supracallosal sulcus. These sulci penetrate sufficiently deeply into the brain to introduce an obvious topological decomposition of its volume architecture. A family of surface maps was constructed, encoding statistical properties of local anatomical variation within individual sulci. A probability space of random transformations, based on the theory of Gaussian random fields, was developed to reflect the observed variability in stereotaxic space of the connected system of anatomic surfaces. A complete system of probability density functions was computed, yielding confidence limits on surface variation. The ultimate goal of brain mapping is to provide a framework for integrating functional and anatomical data across many subjects and modalities. This task requires precise quantitative knowledge of the variations in geometry and location of intracerebral structures and critical functional interfaces. The surface mapping and probabilistic techniques presented here provide a basis for the generation of anatomical templates and expert diagnostic systems which retain quantitative information on intersubject variations in brain architecture.
References (0)
Cited by (193)
From sMRI to task-fMRI: A unified geometric deep learning framework for cross-modal brain anatomo-functional mapping
2023, Medical Image AnalysisAchieving predictions of brain functional activation patterns/task-fMRI maps from its underlying anatomy is an important yet challenging problem. Once successful, it will not only open up new ways to understand how brain anatomy influences functional organization of the brain, but also provide new technical support for the clinical use of anatomical information to guide the localization of cortical functional areas. However, due to the non-Euclidean complex architecture of brain anatomy and the inherent low signal-to-noise ratio (SNR) properties of fMRI signals, the key challenge in building such a cross-modal brain anatomo-functional mapping is how to effectively learn the context-aware information of brain anatomy and overcome the interference of noise-containing task-fMRI labels on the learning process. In this work, we propose a Unified Geometric Deep Learning framework (BrainUGDL) to perform the cross-modal brain anatomo-functional mapping task. Considering that both global and local structures of brain anatomy have an impact on brain functions from their respective perspectives, we innovatively propose the novel Global Graph Encoding (GGE) unit and Local Graph Attention (LGA) unit embedded into two parallel branches, focusing on learning the high-level global and local context information, respectively. Specifically, GGE learns the global context information of each mesh vertex by building and encoding global interactions, and LGA learns the local context information of each mesh vertex by selectively aggregating patch structure enhanced features from its spatial neighbors. The information learnt from the two branches is then fused to form a comprehensive representation of brain anatomical features for final brain function predictions. To address the inevitable measurement noise in task-fMRI labels, we further elaborate a novel uncertainty-filtered learning mechanism, which enables BrainUGDL to realize revised learning from the noise-containing labels through the estimated uncertainty. Experiments across seven open task-fMRI datasets from human connectome project (HCP) demonstrate the superiority of BrainUGDL. To our best knowledge, our proposed BrainUGDL is the first to achieve the prediction of individual task-fMRI maps solely based on brain sMRI data.
Multiatlas segmentation
2019, Handbook of Medical Image Computing and Computer Assisted InterventionAdvancing radiology towards quantitative interpretation relies on innovating with image processing algorithms for context navigation (e.g., “Does this image contain a tumor?”), object localization (”Where is the spleen located?”), and segmentation (e.g., “Which voxels in the image belong to the hippocampus?”). In this chapter, we focus on the most invasive of these questions – segmentation. An algorithm that yields a segmentation of a medical image assigns each voxel in a target medical image to an object class (either deterministically or probabilistically). The results of a segmentation process can be used to address a wide variety of clinical problems from volumetry and shape analysis to radiomics and image-guided interventions. Historically, human efforts for capturing medical image segmentations (e.g., labelers or, herein, raters) have been remarkably resource intensive and subject to high degrees of interrater variability. Hence, medical images paired with their segmentations (so-called maps, charts, or, herein, atlases) are treasured resources. Creating approaches to learn algorithms based on an available and/or feasible collection of atlases has been at the forefront of medical image computing research for the better part of a quarter century. An intuitive and effective family of methods for using existing atlases to create a new algorithm is known as multiatlas segmentation. The core idea underlying multiatlas segmentation is that differences in anatomy as seen through medical images are relatively small and can be compensated through image registration. Hence, the segmentation of an atlas registered to an unseen target represents a reasonable estimation of the true segmentation for that target. When the registration process is repeated for multiple atlases, one would achieve multiple reasonable estimates of a segmentation. The field of multiatlas segmentation centers on capturing and resolving the uncertainty associated with multiple segmentation estimates and optimizing the preprocessing, registration, information integration, and postprocessing steps needed to address complexities of specific anatomies, imaging modalities, and clinical context. The intent of this chapter is to provide historical context for interpretation of modern multiatlas segmentation methods while offering a consistent notation to compare and contrast the evolving literature. Notably, while we highlight salient innovations both from practical and theoretical perspectives, this chapter does not seek to provide a comprehensive literature review.
Cortical Brain Abnormalities in 4474 Individuals With Schizophrenia and 5098 Control Subjects via the Enhancing Neuro Imaging Genetics Through Meta Analysis (ENIGMA) Consortium
2018, Biological PsychiatryThe profile of cortical neuroanatomical abnormalities in schizophrenia is not fully understood, despite hundreds of published structural brain imaging studies. This study presents the first meta-analysis of cortical thickness and surface area abnormalities in schizophrenia conducted by the ENIGMA (Enhancing Neuro Imaging Genetics through Meta Analysis) Schizophrenia Working Group.
The study included data from 4474 individuals with schizophrenia (mean age, 32.3 years; range, 11–78 years; 66% male) and 5098 healthy volunteers (mean age, 32.8 years; range, 10–87 years; 53% male) assessed with standardized methods at 39 centers worldwide.
Compared with healthy volunteers, individuals with schizophrenia have widespread thinner cortex (left/right hemisphere: Cohen’s d = −0.530/−0.516) and smaller surface area (left/right hemisphere: Cohen’s d = −0.251/−0.254), with the largest effect sizes for both in frontal and temporal lobe regions. Regional group differences in cortical thickness remained significant when statistically controlling for global cortical thickness, suggesting regional specificity. In contrast, effects for cortical surface area appear global. Case-control, negative, cortical thickness effect sizes were two to three times larger in individuals receiving antipsychotic medication relative to unmedicated individuals. Negative correlations between age and bilateral temporal pole thickness were stronger in individuals with schizophrenia than in healthy volunteers. Regional cortical thickness showed significant negative correlations with normalized medication dose, symptom severity, and duration of illness and positive correlations with age at onset.
The findings indicate that the ENIGMA meta-analysis approach can achieve robust findings in clinical neuroscience studies; also, medication effects should be taken into account in future genetic association studies of cortical thickness in schizophrenia.
Tests of cortical parcellation based on white matter connectivity using diffusion tensor imaging
2018, NeuroImageThe cerebral cortex is conventionally divided into a number of domains based on cytoarchitectural features. Diffusion tensor imaging (DTI) enables noninvasive parcellation of the cortex based on white matter connectivity patterns. However, the correspondence between DTI-connectivity-based and cytoarchitectural parcellation has not been systematically established. In this study, we compared histological parcellation of New World monkey neocortex to DTI- connectivity-based classification and clustering in the same brains. First, we used supervised classification to parcellate parieto-frontal cortex based on DTI tractograms and the cytoarchitectural prior (obtained using Nissl staining). We performed both within and across sample classification, showing reasonable classification performance in both conditions. Second, we used unsupervised clustering to parcellate the cortex and compared the clusters to the cytoarchitectonic standard. We then explored the similarities and differences with several post-hoc analyses, highlighting underlying principles that drive the DTI-connectivity-based parcellation. The differences in parcellation between DTI-connectivity and Nissl histology probably represent both DTI's bias toward easily-tracked bundles and true differences between cytoarchitectural and connectivity defined domains. DTI tractograms appear to cluster more according to functional networks, rather than mapping directly onto cytoarchitectonic domains. Our results show that caution should be used when DTI-tractography classification, based on data from another brain, is used as a surrogate for cytoarchitectural parcellation.
Construction of 4D high-definition cortical surface atlases of infants: Methods and applications
2015, Medical Image AnalysisIn neuroimaging, cortical surface atlases play a fundamental role for spatial normalization, analysis, visualization, and comparison of results across individuals and different studies. However, existing cortical surface atlases created for adults are not suitable for infant brains during the first two postnatal years, which is the most dynamic period of postnatal structural and functional development of the highly-folded cerebral cortex. Therefore, spatiotemporal cortical surface atlases for infant brains are highly desired yet still lacking for accurate mapping of early dynamic brain development. To bridge this significant gap, leveraging our infant-dedicated computational pipeline for cortical surface-based analysis and the unique longitudinal infant MRI dataset acquired in our research center, in this paper, we construct the first spatiotemporal (4D) high-definition cortical surface atlases for the dynamic developing infant cortical structures at seven time points, including 1, 3, 6, 9, 12, 18, and 24 months of age, based on 202 serial MRI scans from 35 healthy infants. For this purpose, we develop a novel method to ensure the longitudinal consistency and unbiasedness to any specific subject and age in our 4D infant cortical surface atlases. Specifically, we first compute the within-subject mean cortical folding by unbiased groupwise registration of longitudinal cortical surfaces of each infant. Then we establish longitudinally-consistent and unbiased inter-subject cortical correspondences by groupwise registration of the geometric features of within-subject mean cortical folding across all infants. Our 4D surface atlases capture both longitudinally-consistent dynamic mean shape changes and the individual variability of cortical folding during early brain development. Experimental results on two independent infant MRI datasets show that using our 4D infant cortical surface atlases as templates leads to significantly improved accuracy for spatial normalization of cortical surfaces across infant individuals, in comparison to the infant surface atlases constructed without longitudinal consistency and also the FreeSurfer adult surface atlas. Moreover, based on our 4D infant surface atlases, for the first time, we reveal the spatially-detailed, region-specific correlation patterns of the dynamic cortical developmental trajectories between different cortical regions during early brain development.
Hippocampal and amygdalar local structural differences in elderly patients with schizophrenia
2015, American Journal of Geriatric PsychiatryMorphological abnormalities have been reported for the hippocampi and amygdalae in young schizophrenia patients, but very little is known about the pattern of abnormalities in elderly schizophrenia patients. Here we investigated local structural differences in the hippocampi and amygdalae of elderly schizophrenia patients compared with healthy elderly subjects. We also related these differences to clinical symptom severity.
20 schizophrenia patients (mean age: 67.4 ± 6.2 years; Mini-Mental State Exam: 22.8 ± 4.4) and 20 healthy elderly subjects (70.3 ± 7.5 years; 29.0 ± 1.1) underwent high resolution magnetic resonance imaging of the brain. The Radial Atrophy Mapping technique was used to reconstruct the 3D shape of the amygdala and the hippocampus. Local differences in tissue reductions were computed between groups and permutation tests were run to correct for multiple comparisons, in statistical maps thresholded at p = 0.05.
Significant tissue reduction was observed bilaterally in the amygdala and hippocampus of schizophrenia patients. The basolateral-ventral-medial amygdalar nucleus showed the greatest involvement, with over 30% local tissue reduction. The centro-medial, cortical, and lateral nuclei were also atrophic in patients. The hippocampus showed significant tissue loss in the medio-caudal and antero-lateral aspects of CA1, and in medial section of its left head (pre- and para-subiculum). In the left amygdala and hippocampus, local tissue volumes were significantly correlated with negative symptoms.
Tissue loss and altered morphology were found in elderly schizophrenia patients. Tissue loss mapped to amygdalo-hippocampal subregions known to have bidirectional and specific connections with frontal cortical and limbic structures and was related to clinical severity.