Image analysis – Applications – Target tracking or detecting
Reexamination Certificate
1999-02-01
2001-06-19
Mancuso, Joseph (Department: 2621)
Image analysis
Applications
Target tracking or detecting
C348S169000
Reexamination Certificate
active
06249590
ABSTRACT:
FIELD OF THE INVENTION
This invention relates in general to a method for automatically locating instances of a target image pattern in a digital image, and more particularly relates to a method for automatically locating intervertebral disks in lumbar spine for computed radiography produced digital radiographic images.
BACKGROUND OF THE INVENTION
Computed radiography is a technique in which a latent radiographic image formed in a reusable storage phosphor is read out to produce a digital radiographic image.
Images acquired by computed radiography (CR) require a tone scale mapping for a diagnostically satisfactory visual examination. The tone scale is the mapping between the input code values and the output code values for display media, such as film density of a photographic film or luminance of a cathode-ray-tube monitor. Prior to the determination of the tone scale curve, the desired region of interest in a CR image (that is, the body parts) should be separated from the irrelevant objects (for example, the background and foreground) via a segmentation algorithm. For a low signal to noise ratio image, the segmentation task is particularly difficult. Hence, for some body parts and projections, the tone scale algorithm does not give the desired contrast look of the images. Lateral lumbar spine CR images are one of these types of image.
Various methods for locating/identifying the bone region of interest have been proposed. For example, in order to use a posterior/anterior (PA) chest radiograph to analyze lung texture, the inter-rib bones shown in the PA chest image need to be removed. U.S. Pat. No. 4,851,984, issued Jul. 25, 1989, to inventors K. Doi, et al. teaches a method to locate inter-rib spaces in digital chest radiograph images. First, a lung field is defined by determining the rib cage edge boundary. A horizontal signal profile is obtained at a predetermined vertical location. The pixel location at which the second derivative of this horizontal profile is minimum is defined as the rib cage edge boundary. Then two vertical profiles in the periphery of both lungs are fitted with a shift-invariant sinusoidal function. This technique assumes that the horizontal line is always perpendicular to the spinal column. Moreover, it assumes that the relative vertical locations of the objects are known a priori. Therefore, U.S. Pat. No. 4,851,984 does not teach a fully automatic method for locating instances of an image pattern. Furthermore, the lumbar spine bone structure is different from the inter-rib bone structure. The profiles with sinusoidal function do not fit the lumbar spine bone. U.S. Pat. No. 5,268,967, issued December, 1993, to inventors B. K. Jang, et al. teaches a method of using edge density for segmenting image into foreground, background, and body parts regions. Yet, the lumbar spine image needs further segmentation of lumbar spine bone region from soft tissue inside the body part. Because the lumbar spine region which is even separated from the background and the foreground still spans a large dynamic range, that is, a wide code value range. This causes low contrast look in an output display.
Histogram methods have been used to locate the bone region of interest. U.S. Pat. No. 5,228,068, issued Jul. 13, 1993, to inventor R. B. Mazess, teaches a method to determine and analyze vertebral morphology by evaluating the approximate center location of each vertebra from a digital lateral vertebral scan. The centers are located by evaluating the horizontal and vertical histograms. The horizontal histogram is constructed along a line cross each anterior-posterior border of the vertebra. The vertical histogram is obtained along a line cross superior-inferior border, which directs the spine column. Because the patient is supported in the supine position on a table so that the vertebrae of the spine are generally aligned with the scan direction. However, because of the curvature of the spine, the angle of the vertebrae, that is, the angle of an anterior border, a posterior border, a superior border, and an inferior border with respect to the scan direction will vary among vertebrae. This method requires that this variation be accommodated by the trained eye of a physician in estimating the initial positions of lines which horizontal and vertical histogram are generated from. The rigid assumption about the relative orientation of the image relative to the body makes the invention sensitive to orientation error and can not be used in automatic mode for general image orientations.
Another histogram-based method disclosed in U.S. Pat. No. 4,951,201 is used to determine the image body posture. In order to produce a better view of digital radiography, it is important to know if the object is imaged from its front side or lateral side. For example, for a chest image, the imaged thoracic vertebrae is of a relatively low density when it is imaged from its front side, and of relatively high density when it is imaged from its lateral side. This method determines imaged body posture by analyzing the histogram characteristics along a prescribed direction across the image. However, the method does not describe whether a prescribed direction is determined by a user interaction or by an automatic algorithm.
Other methods have used shape information to locate the bone region of interest. U.S. Pat. No. 4,922,915, issued May 8, 1990, to inventor B. A. Arnold, teaches a method to place an region of interest (ROI) of regular (e.g., elliptical) or irregular shape in a specific region of the image of the patient's anatomy, such as the trabecular bone region of the patient's spine. The cross-section image of an upper region of vertebra body contains the cortical bone image which appears as an outer ring, the trabecular bone image which occupies a portion of inside the ring, and the basivertebral vein image which occupies another small portion inside the ring. It is desired to exclude the basivertebrae vein from the trabecular bone image. This method requires that the operator positions the enlarged region of interest so that it encompasses the top region of the vertebral body including the trabecular bone image but excluding the basivertebral vein image. Then the template search and histogram analysis algorithm are used to place the trabecular bone region.
Some researchers have proposed methods to segment images of the hand bone for the purpose of age assessment (D. T. Morris et al., “Segmentation of the finger bones as a prerequisite for the determination of bone age,”
Image and Vision Computing,
Vol. 12, No. 4, pp. 239-246, May 1994; E. Pietka et al., “Feature extraction in carpal-bone analysis,”
IEEE Transactions on Medical Imaging,
Vol. 12, No. 1, pp. 44-49, 1994; E. Pietka et al., “Computer-assisted phalangeal analysis in skeletal age assessment,”
IEEE Transactions on Medical Imaging,
Vol. 10, No. 4, pp. 616-619, December 1991; C.-L. Chang et al., “Computer-aided diagnosis: detection and characterization of hyperparathyroidism in digital hand radiographs,”
Medical Physics,
20(4), pp. 983-992, July/August 1993.). In the first three publications, the hand bone region of interest was defined using a standard thresholding technique to separate the hand from the background. Unfortunately, thresholding is not a robust and reliable image segmentation method for the texture-rich CR medical images. In the last publication, the extraction of the hand region of interest was performed manually. Then, an edge filter was applied to extract edges of finger bones. However, an edge filter usually produces numerous false alarms and misses; a complicated post processing is needed to remove these false alarms. This proposed method also relies on the user (manual) supervision which is not desirable.
The lumbar spine bone has a more complicated structure than the hand bone. Simple edge filtering cannot extract lumbar spine bone features (P. P. Smyth et al., “Automatic measurement of vertebral shape using active shape models,”
Proceedings of British Machine Vision Conference,
Edinburge, Sep. 9-12
Lee Hsien-Che
Young Susan S.
Eastman Kodak Company
Mancuso Joseph
Noval William F.
Tabatabai Abolfazl
LandOfFree
Method for automatically locating image pattern in digital... 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 for automatically locating image pattern in digital..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for automatically locating image pattern in digital... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2517219