Image analysis – Applications – 3-d or stereo imaging analysis
Reexamination Certificate
1999-09-14
2003-07-08
Ahmed, Samir (Department: 2623)
Image analysis
Applications
3-d or stereo imaging analysis
C128S922000, C378S004000, C600S410000
Reexamination Certificate
active
06591004
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention describes a method for (i) generating probabilistic maps of the structure of cerebral cortex from neuroimaging data; (ii) using these maps to reconstruct explicit surface representations of the cerebral cortex; (iii) generating segmented representations of selected subcortical neuroanatomical structures and using these to exclude subcortical portions of the reconstructed surface; (iv) using a related set of strategies to characterize other aspects of cerebral structure such as the location of gyral and sulcal landmarks; and (v) extending the approach to the modeling of other structures besides cerebral cortex. The method is called SURE-FIT (SUrface REconstruction by Filtering and Image Transformations), because it relies heavily on a suite of filtering operations and transformations of image data to extract the information needed to reliably represent the shape of the cortical sheet.
2. Description of the Prior Art
Automatically identifying complex objects represented in 2-D or 3-D image data and determining their precise shape is an important but difficult problem in many areas of science, engineering, and technology. In many instances, the challenge is to reconstruct structures whose precise shape is not known, but which conform to strong constraints regarding their local geometric characteristics.
One problem of this type involves the mammalian cerebral cortex. The cerebral cortex is a thin sheet of tissue (gray matter) that is folded into a complex pattern of convolutions in humans and many other species. For a variety of purposes in both basic neuroscience research and clinical investigations, it is desirable to generate three-dimensional surface reconstructions that represent the shape of the cortical sheet. Relevant areas of application include experimental studies of the structure, function, and development of the cortex in humans and laboratory animals, plus clinical efforts to understand, diagnose, and treat neurological diseases, mental disorders, and injuries that involve the cerebral cortex.
The need for automated cortical surface reconstruction methods has grown rapidly with the advent of modern neuroimaging methods. Magnetic resonance imaging is particularly important, as it can noninvasively reveal the detailed pattern of cortical folds in individual subjects, and it also allows visualization of brain function on a scale comparable to that of cortical thickness.
A related problem involves the automated identification of the many subcortical nuclei and other neuroanatomical structures contained in the interior of the brain. These structures have a variety of complex shapes; some are heterogeneous in their material composition (of gray matter, white matter, and CSF); and some have common boundaries with several different structures. This makes it difficult to establish a consistent set of criteria for reliably segmenting any given structure. Most subcortical structures have a relatively consistent location in relation to standard neuroanatomical landmarks, which can be a valuable aid for segmentation.
This document makes no attempt to survey the extensive literature on segmentation and surface reconstruction in general, or even that relating to cerebral cortex and subcortical structures in particular. However, it is widely recognized that currently available computerized methods for reconstructing the shape of the cortex have major limitations in their accuracy and fidelity when dealing with the noisy images typically obtained with current neuroimaging methods. The SURE-FIT method offers a number of conceptual and practical advantages as an improved method for reconstructing and modeling the cerebral cortex and associated subcortical structures.
SUMMARY OF THE INVENTION
SURE-FIT is designed to operate on gray-scale volumetric imaging data as its primary input. Two common sources of relevant data are structural MRI and images of the cut face of the brain taken during histological sectioning.
SURE-FIT can produce a variety of volumetric (voxel-based) representations and surface representations that are useful individually or in various combinations. Surface representations include an initial surface representation that is constrained to lie within the inner and outer boundaries of the cortical sheet; representations of the inner, middle, and outer surfaces of the cortex; a representation of the radial axis along which these surfaces are linked, a representation of location within the cortical sheet in a three-dimensional coordinate system that respects the natural topology and structure of the cortex. Volume representations include both probabilistic (gray-scale) and deterministic (classified) maps of gray matter, subcortical white matter, and other structures of interest; plus vector-field measures of the location and orientation of the inner and outer boundaries of cortical gray matter and of the radial axis of the cortical sheet.
SURE-FIT emphasizes a combination of mathematical filters and transformations. that are designed to be near-optimal for extracting relevant structural information, based on known characteristics of the underlying anatomy and of the imaging process (i.e., priors in the Bayesian probabilistic sense). The use of filters and transformations per se for image segmentation and tissue classification is not new. The power of the SURE-FIT approach, as well as its novelty, derives from the particular choices of mathematical operations and their systematic application in order to efficiently utilize a large fraction of the relevant data contained in structural images. SURE-FIT also includes a family of shape-changing operations such as dilation, erosion, shifting, and sculpting that are applied to segmented (binary) volumes. When applied in appropriate combinations to appropriate intermediate volumes, these operations allow accurate segmentation of major subcortical structures.
In one form, the method of the invention is for reconstructing surfaces and analyzing surface and volume representations of the shape of an object corresponding to image data, in which the object has been modeled as one or more physically distinct compartments. The method comprises the following steps. Characteristics of a compartmental model are specified in terms of the material types contained in each distinct compartment as defined by the image data and in terms of the nature of compartmental boundaries as defined by the image data. An image model is specified that includes image intensity functions for each material type and for each boundary type based on the specified characteristics. Gradient functions are specified that characterize boundary types and some compartmental regions based on the specified characteristics. A set is generated of probabilistic volume representations of the location of different compartments and of the location and orientation of compartmental boundaries based on the image intensity functions and the gradient functions. A set of segmented (binary) volumes is generated that represent structures in the vicinity of said object, particularly those adjoining its perimeter, in order to identify and subsequently exclude said adjoining structures from the surface reconstruction.
In another form, the invention comprises a method for analyzing and visualizing the volumes of compartments enclosed by explicit surfaces comprising the following steps. For a topologically closed surface of the explicit surface, whether a voxel is inside, outside, or intersected by the closed surface is determined. For each voxel intersected by the surface, the fractional occupancy of the voxel by the region enclosed by the surface is determined. The total volume enclosed by the surface is determined by summing the fractional occupancy values, including those contained entirely within the surface. The total volume is visualized by scaling the voxel intensity according to the fractional occupancy.
In another form, the invention comprises a method for reconstructing the shape and identifying objects in 2-di
Anderson Charles H.
Drury Heather A.
VanEssen David C.
Ahmed Samir
Bhatnagar Anand
Senniger Powers Leavitt & Roedel
Washington University
LandOfFree
Sure-fit: an automated method for modeling the shape of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Sure-fit: an automated method for modeling the shape of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Sure-fit: an automated method for modeling the shape of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3041570