Image analysis – Applications – Biomedical applications
Reexamination Certificate
1998-12-01
2002-07-23
Boudreau, Leo (Department: 2721)
Image analysis
Applications
Biomedical applications
C382S128000, C250S455110
Reexamination Certificate
active
06424732
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to segregating selected object images from an image. In particular this invention relates to segregating selected object images from images containing a collection of object images to remove background artifacts and image leaks that result from closely spaced object images-and to view selected objects image independently.
BACKGROUND
Image segregation herein means defining the boundaries of an object image from an image in order to separate the object image from noise and other object images. Segregating an object image from 2-dimensional images involves defining the boundaries of the image that correspond to the object. Defining the boundaries of an object image can be difficult, especially if the object image is blurred or is positioned in an area of the image where there is a collection of other object images that leak or bleed into each other. Separating an object image form a more complex image or defining the boundaries of an object image within an image is useful for the identification of the object that has been imaged. For example, the authenticity of a blurred photograph could be ascertained if the true shape of the i objects in the photograph can be discerned. Furthermore, image segregation can be used to support automated indexing and retrieval of photographs from large photograph collections based on the photographs' contents, or could facilitate identification of bank robbers from poor quality video images. Of all the possible applications of image segregation, one of more useful applications has been in the area of medical imaging.
Computed Tomography (CT) is an important non-evasive method to image tissue volumes for medical diagnoses and therapy planning. The general concept of CT imaging involves collecting multiple stepwise images of a circumference of a tissue slice. A 2-dimensional image is reconstructed by back projection of the detected signals densities on to the imaged slice. Any number of 2-dimensional slices are acquired and can be stacked to generate a 3-dimensional image volume representing the imaged tissue.
Virtually all volume imaging methods use digitizing image data. Digitizing the image data allows a computer system to manipulate the data and stack the 2-dimensional images into a 3-dimensional image volume. Once the image volume is constructed the volume can be edited by any number of methods. There are several image acquisition methods, such as ultrasound and Magnetic Resonance (MR), for generating image volumes that use the same basic principle of generating 2-dimensional digitized images and stacking them to generate a 3-dimensional image volume which then can be manipulated either sectional or as an image volume by a computer.
It is often beneficial to selectively view a particular feature or object image in the imaged volume by a process of image segregation. For example, vascular tissue is difficult to view in the complex surrounding of an image containing large organs and bone tissue. There are many exemplar methods for image segregation; for a review see Pal, N. R., and Pal, S. K. A review on image segmentation techniques, Pattern Recognition, Volume 26(9), pp. 1277-1294. Image segregation in a three dimensional volume involves two basic steps. Firstly, at least one two dimensional cross section needs to be edited to select the image that is to be segmented. For example, a region of pixels containing Voxels of the object image of interest can be traced and the pixels outside of the traced region can have the image intensity set to zero. Prior art segregation methods involve manual tracing of the object image to be segmented. The procedure of manual tracing is very time consuming even for experienced radiologists and the results are not consistent from operator to operator. Image thresholding is another common technique for segmenting object images from an image. The basic principle is to assign image thresholds that correspond to the imaged object to be segmented. However, object images that appear in medical images often have overlapping intensity distribution and thresholding can result in the misclassification of Voxels. Users usually distinguish between good and inadequate thresholds by the shape of the regions whose Voxels intensities are contained within the thresholds and a priori knowledge of the shape of the object imaged. Deciding what is a good intensity threshold can be time consuming and can give unreasonable results because of overlapping intensities.
The second step to image segregation in three dimensions involves identifying and grouping the object images from each consecutive 2-dimensional cross section that belongs to the object to be segmented based on grouping criteria. For example the grouping criteria could be that object images that have a similar image intensity and position. Generally, all the method for image segregation in 2-dimensions can be extended to three dimensions but still have the same limitations and difficulties.
What is needed is a method for object image segregation that can be automated, does not require hand tracing and does not require a priori knowledge of shapes of the objects imaged. Further there is a need for a method that can segregate complex structures such as vascular tissue images. The method should reduce the time involve in the segregation, reduce artifacts in the image caused by leaks from surrounding object and be readily applied to image volumes.
OBJECTS AND ADVANTAGES
It is a primary object of this invention to provide a method for segregating object images from images, whereby the segregation method requires no a priori knowledge about the shapes of the object imaged or the object images. The method is general and can be used for segregating object images from continues images such as photographs and film
It is a further object of this invention to provide a method for segregating object images from an image using information obtained from intensity thresholding and the inherent shapes of the object images. The method generates segregated images of object that accurately represent the features of the object imaged.
It is a specific object of this invention to provide a method of segregating object images from images, wherein the method does not require manual tracing. The method eliminates the inconstancies that occur from operator tracing, reduces the time of segregating the object images, is not specific or sensitive to the image acquisition mode and can be fully automated.
It is a more specific object of the invention to provide a method of segregating complex object images from images. The method has applications in medical imaging and is particularly useful for segregating vessel tissue from an image volume, wherein the image volume is constructed from digitized data acquisitions obtained ultrasound, CT, and MR imaging techniques.
SUMMARY
The objects of the invention are obtained providing a method for segmenting complex object images, such as arterial structures, from an image volume. The image volume is preferably obtained by acquiring 2-dimensional images that represent imaged slices of an image volume. The method is most useful in the field of medical imaging where the 2-dimensional images are digitized images obtained from CT, MR or ultrasound acquisition data and are comprised of Voxels. While it is not required, in the preferred embodiment of this invention the 2-dimensional images are segmented to define the salient regions of the images containing the object images to be segmented prior to constructing a 3-dimensional image volume.
There are any number of methods to segment object images from 2-dimensional images using intensity thresholding, but the method of this invention uses the combinations of intensity thresholding and intrinsic imaged shapes in order to segment object images. The method can be completely automated and does not require prior knowledge of the shapes of the objects. imaged.
An intensity threshold or range of intensities is predetermined to give a sufficient number of divisions of
Napel Sandy
Shiffman Smadar
Boudreau Leo
Choobin Barry
Lumen Intellectual Property Services
The Board of Trustees of the Leland Stanford Junior University
LandOfFree
Object segregation in images does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Object segregation in images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Object segregation in images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2821163