Image analysis – Applications – Biomedical applications
Reexamination Certificate
2002-10-02
2004-05-04
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
Biomedical applications
C382S154000, C382S173000
Reexamination Certificate
active
06731782
ABSTRACT:
FIELD OF THE INVENTION
The present invention is directed to a method and system for detection and delineation of abnormal regions in volumetric image sets and is more particularly directed to such a method and system in which the detection and delineation are automated through the use of spectral and spatial information.
DESCRIPTION OF RELATED ART
Numerous disease processes are characterized by the development of lesions that are structurally and compositionally distinct from surrounding healthy tissue. Some examples include multiple sclerosis and Alzheimer's disease, which are characterized by the development of brain lesions or plaques, and the sub-set of cancers that include the development of solid tumors in the brain, bones, or organs. In order to assess the progress or response to treatment of these diseases, it is necessary to obtain some measure of the patient's total lesion burden. In some cases, it is also helpful to know specific things about the structure of individual lesions.
Clearly, any accurate measure of a complex three-dimensional structure requires a three-dimensional image set. For this reason, magnetic resonance imaging (MRI) and computed tomography (CT), which provide complete volume imagery, are preferred over plain films for these assessments. Once the imagery has been obtained, however, any assessment requires the location, identification, and delineation of all lesions within the volume. Current standard practice requires an expert, typically a radiologist, to read each of the 30-100 images in the volume set, identify any lesions present, and trace out the boundaries of each lesion using specialized computer software. The traced boundaries are then used to calculate lesion volumes and other biomarkers.
This procedure has a number of obvious drawbacks. First, it is both tedious and time consuming. Manual tracing of a single volume data set can take anywhere from 15 minutes to two hours or more, depending on the number of images in the set and the number of lesions per image. Second, because manual outlining is heavily dependent on the opinion of the observer, it produces results that are subject to both error and bias. Recent studies have shown coefficients of variation of 5% or more for repeated tracings of the same structures by a single observer, and of up to 50% in some cases for tracings of structures by multiple observers. Such wide variability renders the results of such an analysis nearly useless, and points out a clear need for an improved, preferably automated, method of measurement. Some examples of prior work in this field include:
[1] E. Ashton et al., “Automated Measurement of Structures in CT and MR Imagery: A Validation Study” Proc. of IEEE-
Computer Based Medical Systems
, pp. 300-305 (2001).
[2] R. Chung, C. Ho, “3-D Reconstruction from tomographic data using 2-D active contours”
Computers and Biomedical Research
(33), pp. 186-210 (2000).
[3] K. Juottonen et al. “Volumes of the entorhinal and perirhinal cortices in Alzheimer's disease”
Neurobiology of Aging
(19), pp.15-22 (1998).
[4] E. Ashton, K. Parker, M. Berg, C. Chen, “A Novel Volumetric Feature Extraction Technique with Applications to MR Images” IEEE Trans.
Medical Imaging
(16), pp.365-371 (1997).
[5] E. Ashton et al., “Segmentation and Feature Extraction Techniques, with Applications to MRI Head Studies”
Magnetic Resonance in Medicine
(33), pp. 670-677 (1995).
[6] D. Taylor, W. Barrett, “Image segmentation using globally optimum growth in three dimensions with an adaptive feature set”
Visualization in Biomedical Computing
, pp. 98-107 (1994).
[7] I. Carlbom, D. Terzopoulos, K. Harris, “Computer assisted registration, segmentation, and 3-D reconstruction from images of neuronal tissue sections” IEEE
Trans. Medical Imaging
(13), pp. 351-362 (1994).
[8] F. Cendes et al., “MRI volumetric measurement of amygdala and hippocampus in temporal lobe epilepsy”
Neurology
(43), pp. 719-725 (1993).
All of the above referenced work describes schemes that require a human observer to identify the location of each lesion in the volume manually. Most also require some operator input regarding lesion shape and size. These limitations reduce the precision advantage provided over pure manual tracing by introducing subjective opinion into the identification process, and reduce the speed advantage by requiring extensive operator input. Prior work in the area of automated detection of abnormal regions in imagery using grayscale or spectral information includes:
[9] Z. Ge, V. Venkatesan, S. Mitra, “A Statistical 3-D Segmentation Algorithm for Classifying Brain Tissues in Multiple Sclerosis” Proc. of IEEE-
Computer Based Medical Systems
, pp. 455-460 (2001).
[10] K. Van Leemput et al., “Automated Segmentation of Multiple Sclerosis Lesions by Model Outlier Detection” IEEE
Trans. Medical Imaging
(20), pp. 677-688 (2001).
[11] E. Ashton, “Multialgorithm solution for automated multispectral target detection”
Optical Engineering
(38), pp. 717-724 (1999).
[12] E. Ashton, “Detection of sub-pixel anomalies in multspectral infrared imagery using an adaptive Bayesian classifier” IEEE
Trans. Geoscience and Remote Sensing
(36), pp. 506-517 (1998).
[13] E. Ashton, A. Schaum, “Algorithms for the Detection of Sub-Pixel Targets in Multispectral Imagery”
Photogrammetric Engineering and Remote Sensing
(64), pp. 723-731 (1998).
[14] R. Muise, “Coastal mine detection using the COBRA multispectral sensor” SPIE
Detection Remediation Tech. Mines Minelike Targets
(2765), pp. 15-24 (1996).
[15] T. Watanabe et al., “Improved contextual classifiers of multispectral image data” IEICE
Trans. Fundamentals Elect. Commun., Comput. Sci
,(E77-A), pp. 1445-1450 (1994).
[16] X. Yu, I. Reed, A. Stocker, “Comparative performance analysis of adaptive multi-spectral detectors” IEEE
Trans. Signal Processing
(41), pp. 2639-2656 (1993).
[17] I. Reed, X. Yu, “Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution” IEEE
Trans. Acoustics, Speech, Signal Processing
(38), pp. 1760-1770 (1990).
The systems described in these references have very similar theoretical bases and suffer from two common limitations. First, they make primary use only of either spatial/grayscale information (9,10) or spectral signature (11-17). Second, all of these systems operate by forming a statistical model of common background tissues and then searching for outliers. The resulting lack of a priori target information causes these systems to be non-specific and to have impractically high false alarm rates.
SUMMARY OF THE INVENTION
It will be readily apparent from the above that a need exists in the art to overcome the above-noted problems caused by existing techniques for identification and delineation of lesion boundaries. It is therefore an object of the invention to detect and delineate abnormal structures with higher accuracy.
It is another object of the invention to allow increased speed in the detection and delineation of abnormal structures.
It is yet another object of the invention to remove human error from the detection and delineation of abnormal structures.
It is still another object of the invention to reduce the rate of false alarms.
To achieve the above and other objects, the present invention makes use of spectral and spatial information to provide automated detection and delineation of abnormal structures. The present invention goes beyond and improves the work described in these references in two ways. First, it uses statistical techniques which permit the use of significant spatial and spectral information. Second, it allows for a directed search for a particular grayscale or spectral anomaly, presented, e.g., as a user-defined exemplar, whereas the prior work focuses on background characterization and generalized anomaly detection. This allows the system described in this work to be both sensitive and sp
Blank Rome LLP
Johns Andrew W.
VirtualScopics LLC
LandOfFree
Method and system for automatic identification and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for automatic identification and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for automatic identification and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3232630