Image analysis – Applications – Biomedical applications
Reexamination Certificate
1998-01-29
2001-04-03
Mehta, Bhavesh (Department: 2621)
Image analysis
Applications
Biomedical applications
C382S173000
Reexamination Certificate
active
06212291
ABSTRACT:
MICROFICHE APPENDIX
This application has a microfiche appendix of one microfiche with twenty frames.
FIELD OF THE INVENTION
This invention relates in general to digital image processing, and more particularly relates to a method for the detection of collimation or, equivalently, detection of the radiation field in digital radiography to facilitate optimal tone scale enhancement, to minimize the viewing flare caused by the unexposed area, and to improve image segmentation, image compression, and body part identification.
BACKGROUND OF THE INVENTION
In digital radiography, computed radiography (CR) is a medical imaging technique that captures the information using storage phosphor screens. The information is stored as a latent image for later read-out by a laser scanning system (R. J. Pizzutiello and J. E. Cullinam,
Introduction to Medical Radiographic Imaging,
Eastman Kodak Company, 1993). The latent image signal remains stable for a certain period of time ranging from minutes to days. Using a red or near-infrared light to stimulate the phosphor which will then release its stored energy in the form of visible light, one can observe the phenomenon known as photostimulable luminescence. The intensity of the stimulated luminescence is proportional to the number of x-ray photons absorbed by the storage phosphor. In practice, the read-out process is accomplished using a laser scanner in order to provide high resolution data and also not to flood the screen with stimulating light. The stimulated light emitted from each point on the screen is collected and converted into digital signal. The screen can be refreshed by flooding the entire screen with high illuminance light. This allows the screen to be reused.
Recently, a new medical imaging technique called direct radiography (DR) has been introduced which uses an electronic detector to directly acquire digital signal (L. E. Antonuk et al., “Empirical investigation of the signal performance of a high-resolution, indirect detection, active matrix flat-panel imager (AMFPI) for fluoroscopic and radiographic operation”, Med. Phys. 24(1), January 1997, pp. 51-70.).
In their practice, radiologists may use x-ray opaque materials to limit the radiation scattering, and shield the subjects from unnecessary exposure to x-rays. Such x-ray opaque materials constitute the “collimation”. As illustrated in
FIG. 9
, the collimation blades
18
,
20
are placed in between the x-ray source
10
and the subject
14
, and are typically closer to the source. As a result, those body parts of the patient which are not important to diagnosis but may be vulnerable are protected. Furthermore, the use of collimation can also limit the radiation scattering from unintended regions from fogging the storage phosphor plate
16
. Moreover, a CR image may consist of one or more sub-images if the radiologists choose to make multiple exposures, typically of different projections of the same body part, on the same screen. Each sub-image corresponds to one x-ray irradiation field or one exposure.
In general, the resulting digital images need to be enhanced through code-value remapping in order to provide maximum visual contrast in the region of interest prior to display or printing. Such a process is referred to as tone scaling. The dynamic range of the CR systems (over 10,000:1) provides significant advantage over conventional film in terms of exposure latitude so that CR imaging is very tolerable to improper selection of exposure conditions. However, to optimally render such data on desired printing or display devices, it is necessary to develop tone scale remapping function. To this end, it is desirable to exclude the regions shadowed by the collimation from the calculation of the histogram statistics because these regions provide no useful information but distort the intensity histogram. Moreover, since the shadow regions are usually with the highest brightness levels (corresponding to minimum density), the flare can be reduced by setting the shadow regions to a comfortable brightness level or reducing their average brightness level. Such a process is referred to as masking. In the case of multiple-exposure images, it is desirable to separate individual radiation fields so that a best tone scale can be applied to each radiation field and masking can be performed to the best extent.
Radiation field recognition for one or more radiation fields, preferably done automatically, is the key to masking and is also important to tone scaling. However, it is a very challenging task. In CR, we have significant difficulties to overcome: (1) the boundaries between the region of interest (radiation field) and the shadow regions (collimation) are usually fuzzy due to the radiation scattering, (2) the region of interest often has significant modulation including prominent edges, (3) the shadow regions may have some comparable modulation and therefore are not uniform due to the radiation scattering from the region of interest, (4) boundaries may be invisible near very dense body parts due to the lack of x-ray penetration or the frequent underexposure in order to minimize the x-ray dosage.
Upon the extraction of the individual radiation fields, image processing can be performed in selective ways, including but not limited to: rendering the image using a globally optimal tone scale based on statistics gathered from the collection of extracted radiation fields, or rendering each radiation field using a locally optimal tone scale based on statistics gathered from a given radiation field, or rendering the collimation shadow region. etc.
Due to the practical use and the technical difficulties in collimation recognition, there have been considerable efforts on this subject in the past. Thus, U.S. Pat. No. 4,952,807 assumes that the collimation does not touch the object of the image and the search of collimation boundary is by scanning inwards from image borders to detect the object portion surrounded by the background region). The assumption is not always valid, in particular in shoulder or skull examinations. U.S. Pat. No. 4,804,842 excludes the shadow region by finding the first valley and thus removing the lower part of the histogram. The basic assumption of this invention is that the code values of the those pixels inside the radiation field should be higher than those of the pixels in the collimation shadow regions, which is not always valid. These two patents only address the tone scale problem which does not demand an explicit detection of the collimation boundaries. U.S. Pat. No. 4,970,393 thresholds 1st derivatives along predetermined scan lines. European Patent 0,342,379 (also Japan Patent 2,071,247 and U.S. Pat. No. 5,081,580) assumes strong signal-to-shadow boundaries in terms of the amplitude of the ID differentiation along the radial lines from the center of the image. European Patent 0,285,174, also referred to as U.S. Pat. No. 4,995,093, applies Hough Transform to edge pixels with significant gradient magnitude. European Patent 0,360,231, also referred to as U.S. Pat. No. 4,977,504, and U.S. Pat. No. 5,268,967 describe a distinctive approach which classifies non-overlapping small tiles. U.S. Pat. No. 4,952,805 tests a series of possible hypotheses with respect to the characteristics of the inside and the outside histograms to determine the presence or absence of a radiation field. Japanese Patent 7,181,609 is unique in that constructs a decision network consisting several parallel recognition processes, including a first means for detecting rectangular boundaries using boundary contrast, a second means for close-contour tracing to explicitly deal with arbitrarily curved shape irradiation field, a third means for irradiation field determination using imaging information (collimation shape and position, exposure, etc.) and exam type information, and a forth means using the projections. The final radiation is determined through a coordinating procedure.
There are a few patents particularly related to multiple-exposure images. U.S. Pat. No. 5,032,733, detects unexposed regions in
Luo Jiebo
Senn Robert A.
Wang Xiao-hui
Eastman Kodak Company
Mehta Bhavesh
Noval William F.
Patel Kanji
LandOfFree
Method for recognizing multiple irradiation fields in... 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 recognizing multiple irradiation fields in..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for recognizing multiple irradiation fields in... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2542652