Image analysis – Applications – Biomedical applications
Reexamination Certificate
1999-06-24
2004-03-09
Patel, Jayanti K. (Department: 2625)
Image analysis
Applications
Biomedical applications
C382S226000, C382S240000, C128S922000, C250S582000
Reexamination Certificate
active
06704440
ABSTRACT:
CROSS-REFERENCE TO RELATED APPLICATIONS
Not applicable.
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
Not applicable.
BACKGROUND OF THE INVENTION
The present invention generally relates to image processing carried out in connection with medical imaging, such as digital radiography. More particularly, the present invention relates to a method and apparatus for identifying non-clinical regions in a medical image and correcting therefor. In addition, the present invention relates to a method and apparatus for enhancing imaging processing of edges of medical diagnostic images.
In the past, radiation imaging systems have been proposed which capture images representative of an X-ray scan of a region of interest obtained from a patient. The scan of the region of interest is captured upon a film, screen, digital detector or the like and later converted by a radiation image processing system to a radiation image comprised of gray scale pixels representative of the region of interest of the patient displayable to a physician. Before display, the radiation image is image processed, such as through noise filtering, contrast enhancement, data compression, and the like.
According to one image processing technique, a digital signal representation of a radiation image is first decomposed into a multi-resolution (MR) representation containing localized image detail at multiple scales or frequencies. During MR processing, the image is decomposed into a series of processed images at multiple resolution levels. Multi-resolution imaging decomposes an original or basic image into different resolution or frequency images/bands. The decomposed image also contains a residual image at an even lower resolution level. The images at each level may be referred to as the levels of a pyramid. The differences in frequency for the images of each pyramid level illustrate various image features at different resolutions. Once the decomposition function is completed, the images formed at the various levels of the pyramid are modified in a desired manner, such as to perform edge or contrast enhancement and the like. The images at the various levels of the pyramid may also be combined through weighted functions to afford each pyramid level image a different amount of impact upon a resulting image. The final processed image is computed through a reconstruction algorithm. For instance, edge enhancement may be achieved by heavily weighting the first few decomposed images (i.e., pyramid level zero).
Many different algorithms have been proposed for performing multi-resolution imaging. A common characteristic of multi-resolution imaging algorithms is that decomposition is achieved by applying a filter to the pixel values of the image. For instance, a low pass filter may be used in a convolution process to create each pyramid level. During each iteration through the multi-resolution processing technique, the low pass filter operates upon each pixel within the image, including all of the image border pixels. The filter is applied to the image border by overlaying a filter kernel upon each pixel along the image border and modifying the image pixel based upon the surrounding pixel values overlaid by the filter kernel. For instance, an image pixel overlaid by the center element of a filter kernel may be replaced with an average of the sum of the products of the filter kernel elements and overlaid image pixels. When a filter operates upon the border pixels of the image, a portion of the filter kernel extends beyond the edge of the image. The filter kernel elements that extend beyond the image edge still contribute to the filtered value that replaces an image pixel. Hence, an error is introduced into the resulting filtered image pixel since a portion of the filter kernel extends beyond the image edge. Errors formed by the filter kernel when processing an image edge appear as an artifact in the image to be displayed ultimately.
When using multi-resolution algorithms to process images, such as during digital radiography, incorrect treatment of edges and edge regions may create image artifacts which extend far into the interior of the image. The edges may be caused by collimators located about the patient region of interest during the X-ray process. Edges may also be caused by defective pixels along the border of the X-ray detector. Also, the edge of the radiation field also creates artifacts when processed.
During multi-resolution (MR) processing, the images are decomposed into successive lower resolution images (or pyramid levels) via the convolution of the radiation image with a low pass filter of finite size (e.g., a 3×3, 5×5, 7×7 element array, etc.). Each subsequent pyramid level is typically one-half the size of its predecessor level. Hence, it is preferable that the original image size be integer divisible by 2
N
, where N is the number of levels to be computed during decomposition. Each pyramid level contains specific frequency content. After the transformation of each pyramid level, an output image is constructed by reversing the decomposition process.
One problem that occurs during decomposition is, when the filter kernel is centered over pixels at the edge of the radiation image, no image data exists for convolution with the outer elements of the filter kernel extending beyond the image. For instance, when the filter kernel is centered over a corner pixel of a radiation image and the filter kernel is a 3×3 array, five elements of the filter kernel have no underlying image data, upon which to operate. In the past, it has been proposed to “pad” the radiation image with a border of zeros, where the width of the border was dependent upon the size of the filter kernel. Decomposition is a recursive algorithm and thus the padding must be iteratively placed around the radiation image at each pyramid level. For instance, if the MR algorithm uses eight pyramid levels, the image at each of the eight pyramid levels must be padded with zeros. The padding does not accurately reflect the image data values along the image edge and thus creates an artifact during decomposition in each pyramid level image. The artifacts are carried to each lower level pyramid image and magnified at each pyramid level when a new border of zero padding is added until reaching the bottom of the pyramid where the error becomes quite large.
X-ray systems have recently been proposed which utilize digital X-ray detectors that offer much improved resolution. Digital detectors have in turn enabled X-ray systems to greatly enhance illustrations of small image features. Heretofore, conventional digital detectors inherently exhibited enough noise to mask the artifacts or errors caused by MR processing of image borders. The inherent detector noise covered up artifacts created during the image processing of the borders. Modern digital detectors now offer higher signal-to-noise ratios and thus artifacts created during MR processing of an image border have become more noticeable.
X-ray systems utilize image detectors having a generally fixed size. However, it is not always necessary to view a region of a patient as large as the detector. The size of the patient region that is exposed to X-rays is reduced by blocking a portion of the X-ray source from the patient with a collimator. By way of example, an X-ray detector may capture a radiation image formed of a 2K×2K array of pixel values. The radiation image may include a region associated with (e.g. located behind) the collimator. The border between the collimator and the patient's area of interest is defined by an edge since the collimator blocks radiation, while the patient's body passed a majority of the radiation. The edge has a broad frequency signal component.
An operator may collimate the field of interest to expose a smaller region of interest than the full field of view. When a collimator is used, a region exists within the image which is termed a non-clinical region. Non-clinical regions may be removed or “cropped” such that the resulting image no longe
Choobin Barry
Dellapenna Michael J.
McAndrews Held & Malloy Ltd.
Patel Jayanti K.
Vogel Peter J.
LandOfFree
Method and apparatus for processing a medical image... 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 apparatus for processing a medical image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for processing a medical image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3255590