D. Louis Collins

Unbiased Average Age-Appropriate Atlases for Pediatric Studies

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

The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)

In this paper we report the set-up and results of the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) organized in conjunction with the MICCAI 2012 and 2013 conferences. Twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low- and high-grade glioma patients - manually annotated by up to four raters - and to 65 comparable scans generated using tumor image simulation software. Quantitative evaluations revealed considerable disagreement between the human raters in segmenting various tumor sub-regions (Dice scores in the range 74%-85%), illustrating the difficulty of this task. We found that different algorithms worked best for different sub-regions (reaching performance comparable to human inter-rater variability), but that no single algorithm ranked in the top for all sub-regions simultaneously. Fusing several good algorithms using a hierarchical majority vote yielded segmentations that consistently ranked above all individual algorithms, indicating remaining opportunities for further methodological improvements. The BRATS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource.

Patch-based Segmentation using Expert Priors: Application to Hippocampus and Ventricle Segmentation

Quantitative magnetic resonance analysis often requires accurate, robust, and reliable automatic extraction of anatomical structures. Recently, template-warping methods incorporating a label fusion strategy have demonstrated high accuracy in segmenting cerebral structures. In this study, we propose a novel patch-based method using expert manual segmentations as priors to achieve this task. Inspired by recent work in image denoising, the proposed nonlocal patch-based label fusion produces accurate and robust segmentation. Validation with two different datasets is presented. In our experiments, the hippocampi of 80 healthy subjects and the lateral ventricles of 80 patients with Alzheimers disease were segmented. The influence on segmentation accuracy of different parameters such as patch size and number of training subjects was also studied. A comparison with an appearance-based method and a template-based method was also carried out. The highest median kappa index values obtained with the proposed method were 0.884 for hippocampus segmentation and 0.959 for lateral ventricle segmentation.

Adaptive Non-Local Means Denoising of MR Images with Spatially Varying Noise Levels

To adapt the so‐called nonlocal means filter to deal with magnetic resonance (MR) images with spatially varying noise levels (for both Gaussian and Rician distributed noise).

Focal Gray Matter Changes in Schizophrenia across the Course of the Illness: A 5-Year Follow-Up Study

Recent volumetric magnetic resonance imaging (MRI) studies have suggested brain volume changes in schizophrenia to be progressive in nature. Whether this is a global process or some brain areas are more affected than others is not known. In a 5-year longitudinal study, MRI whole brain scans were obtained from 96 patients with schizophrenia and 113 matched healthy comparison subjects. Changes over time in focal gray and white matter were measured with voxel-based morphometry throughout the brain. Over the 5-year interval, excessive decreases in gray matter density were found in patients in the left superior frontal area (Brodmann areas 9/10), left superior temporal gyrus (Brodmann area 42), right caudate nucleus, and right thalamus as compared to healthy individuals. Excessive gray matter density decrease in the superior frontal gray matter was related to increased number of hospitalizations, whereas a higher cumulative dose of clozapine and olanzapine during the scan interval was related to lesser decreases in this area. In conclusion, gray matter density loss occurs across the course of the illness in schizophrenia, predominantly in left frontal and temporal cortices. Moreover, the progression in left frontal density loss appears to be related to an increased number of psychotic episodes, with atypical antipsychotic medication attenuating these changes.

Automated cortical thickness measurements from MRI can accurately separate Alzheimer's patients from normal elderly controls.

We investigated the potential of fully automated measurements of cortical thickness to reproduce the clinical diagnosis in Alzheimers Disease (AD) using 19 patients and 17 healthy controls. Thickness maps were analyzed using three different discriminant techniques to separate patients from controls. All analyses were performed using leave-one-out cross-validation to avoid overtraining of the discriminants. The results show regionally variant patterns of discrimination ability, with over 90% accuracy obtained in the medial temporal lobes and other limbic structures. Multivariate discriminant analysis produced 100% accuracy with six different combinations, all involving the parahippocampal gyrus. We therefore propose automated measurements of cortical thickness as a tool to improve the clinical diagnosis of probable AD, as well as a research method to gain unique insight into the etiology of cortical pathology in the disease.

BEaST: brain extraction based on nonlocal segmentation technique.

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

Tuning and comparing spatial normalization methods

Spatial normalization is a key process in cross-sectional studies of brain structure and function using MRI, fMRI, PET and other imaging techniques. A wide range of 2D surface and 3D image deformation algorithms have been developed, all of which involve design choices that are subject to debate. Moreover, most have numerical parameters whose value must be specified by the user. This paper proposes a principled method for evaluating design choices and choosing parameter values. This method can also be used to compare competing spatial normalization algorithms. We demonstrate the method through a performance analysis of a nonaffine registration algorithm for 3D images and a registration algorithm for 2D cortical surfaces.

Warping of a computerized 3-D atlas to match brain image volumes for quantitative neuroanatomical and functional analysis

We describe the implementation, experience and preliminary results obtained with a 3-D computerized brain atlas for topographical and functional analysis of brain sub-regions. A volume-of-interest (VOI) atlas was produced by manual contouring on 64 adjacent 2 mm-thick MRI slices to yield 60 brain structures in each hemisphere which could be adjusted, originally by global affine transformation or local interactive adjustments, to match individual MRI datasets. We have now added a non-linear deformation (warp) capability (Bookstein, 1989) into the procedure for fitting the atlas to the brain data. Specific target points are identified in both atlas and MRI spaces which define a continuous 3-D warp transformation that maps the atlas on to the individual brain image. The procedure was used to fit MRI brain image volumes from 16 young normal volunteers. Regional volume and positional variability were determined, the latter in such a way as to assess the extent to which previous linear models of brain anatomical variability fail to account for the true variation among normal individuals. Using a linear model for atlas deformation yielded 3-D fits of the MRI data which, when pooled across subjects and brain regions, left a residual mis-match of 6 - 7 mm as compared to the non-linear model. The results indicate a substantial component of morphometric variability is not accounted for by linear scaling. This has profound implications for applications which employ stereotactic coordinate systems which map individual brains into a common reference frame: quantitative neuroradiology, stereotactic neurosurgery and cognitive mapping of normal brain function with PET. In the latter case, the combination of a non-linear deformation algorithm would allow for accurate measurement of individual anatomic variations and the inclusion of such variations in inter-subject averaging methodologies used for cognitive mapping with PET.

Symmetric atlasing and model based segmentation: an application to the hippocampus in older adults

In model-based segmentation, automated region identification is achieved via registration of novel data to a pre-determined model. The desired structure is typically generated via manual tracing within this model. When model-based segmentation is applied to human cortical data, problems arise if left-right comparisons are desired. The asymmetry of the human cortex requires that both left and right models of a structure be composed in order to effectively segment the desired structures. Paradoxically, defining a model in both hemi-spheres carries a likelihood of introducing bias to one of the structures. This paper describes a novel technique for creating a symmetric average model in which both hemispheres are equally represented and thus left-right comparison is possible. This work is an extension of that proposed by Guimond et al. Hippocampal segmentation is used as a test-case in a cohort of 118 normal eld-erly subjects and results are compared with expert manual tracing.