Surgery – Diagnostic testing – Detecting nuclear – electromagnetic – or ultrasonic radiation
Reexamination Certificate
2001-10-25
2004-06-08
Mercader, Eleni Mantis (Department: 3737)
Surgery
Diagnostic testing
Detecting nuclear, electromagnetic, or ultrasonic radiation
C128S920000, C128S922000, C128S925000, C382S131000, C382S132000, C382S199000, C382S266000
Reexamination Certificate
active
06748257
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an image processing technique for automatically and accurately detecting an anatomical configuration from a chest image necessary for computer-aided diagnosis of the chest, and in particular, to an image processing technique that makes it possible to detect a boundary of a ribcage from a chest image with higher precision, even when its image is poor in quality.
Digitized chest radiographs have been used widely in the field of computer-aided diagnosis. There have been known a wide variety of types of computer-aided diagnosis capable of automatically detecting ribcage boundary information and landmark information both specifying anatomical configurations of the chest. One conventional technique is provided by “Xin-Wei Xu and Kunio Doi, Image feature analysis for computer-aided diagnosis: Accurate determination of ribcage boundary in chest radiographs, Med. Phys. 22(5), May 1995.” This technique is also provided by Japanese Patent Laid-open Publication NO.7-37074.
This diagnostic technique uses lesion-enhanced images in order to detect temporal changes of diseases such as lung diseases among digital chest images acquired at different times for the same patient's region. To raise diagnostic accuracy, this technique comprises the steps of obtaining previous and current digital chest images, positioning both previous and current digital images by performing non-linear warping processing based on a non-linear warping technique on either the first or second digital image, and making a subtraction between previous and current images one of which (has undergone the non-linear warping. The non-linear warping technique uses information detected from a chest image in relation to its anatomical structure and is based on a local matching technique to be applied to a number of tiny regions of interest (ROIs) selected on the basis of the information. The non-linear warping technique is mapping of amounts of matching shift obtained between corresponding locations in two frames of images. The mapping is realized using amounts of local matching resulting from a local matching technique applied to the locations and a weighted fitting technique that uses weighting coefficients resulting from image data analysis applied to the ROIs. In addition, the mapping of shift amounts is based on two-dimensional polynomial functions fitted to shift values.
The above conventional automatic detection technique provides a ribcage boundary in the procedures detailed in FIG.
1
.
First, as pre-processing for detecting a ribcage boundary from chest image data
40
, information indicative of landmarks
41
, which becomes landmark information in displaying a configuration of the chest (Step S
50
).
Then, based on the landmark information
41
, a series of upper lung ribcage boundary candidate points is detected (Step S
51
). Based on both landmark information
41
and upper lung ribcage boundary candidate point series
42
, a series
43
of right ribcage boundary candidate points and a series
44
of left ribcage boundary candidate points are both detected (Step S
52
).
Based on the landmark information
41
, the upper lung ribcage boundary candidate point series
42
is approximated with polynomials, so that an upper lung ribcage boundary point series
45
consisting of a series of points continuously aligned in the X-coordinate direction (horizontal direction) (Step S
53
). Like this, based on the landmark information
41
, both right and left ribcage boundary candidate point series
43
and
44
are approximated with polynomials, so that right and left ribcage boundary point series
46
and
47
each consisting of a series of points continuously aligned in the Y-coordinate direction (vertical direction) (Step S
54
).
Finally, on the basis of the landmark information
41
, both upper lung ribcage boundary point series
45
and right ribcage boundary point series
46
are combined to obtain a right ribcage boundary, and both upper ribcage boundary point series
45
and left ribcage boundary point series
47
are combined to obtain a left ribcage boundary
48
(Step S
55
).
This conventional technique adopts the lung length of ⅕ as the landmark information
41
for obtaining the upper lung ribcage boundary point series
45
.
However, as shown in
FIG. 2
, a region along the lung length of ⅕ within the lung field is relatively low in contrast and the original chest image includes artifacts, such as a blank
49
occurring in converting a chest image into a digital image by the use of a film digitizer and noises
50
occurring in image processing. Thus, in some occasions, the foregoing conventional technique fails in detecting search limit points to search a series of ribcage boundary candidate points.
FIGS. 3A and 3B
pictorially show examples in which the detection fails. In the case of
FIG. 3A
, an artifact that consists of a noise
50
existing at the right side of a chest image has influence on setting an outer search limit point in such a way that a blade bone portion is set as the outer search limit point by mistake. As a result, as shown in
FIG. 3B
, the processing fails in detecting a series
51
of upper lung right ribcage boundary candidate points.
Specifically, in the case of the conventional detection technique, a fail in detection of ribcage boundary candidate points due to influence of setting a reference for detecting the ribcage boundary candidate points and artifacts on an original chest image results in that a ribcage boundary obtained by approximating polynomials to a series of the ribcage boundary candidate points deviates from the true ribcage boundary, thereby lowering accuracy in computer-aided diagnosis as a whole.
In addition, both right and left ribcage boundary candidate point series are detected from the series of upper lung ribcage boundary candidate points. Hence, failing in detecting the series of upper lung ribcage boundary candidate points will give rise to failure in successively performed detection of both series of right and left ribcage boundary candidate points. A ribcage boundary computed by approximating polynomials to such erroneous ribcage boundary candidate point series is no longer the true ribcage boundary. This will lead to lowered accuracy in computer-aided diagnosis.
For using in computer-aided diagnosis features of a chest image and positional information indicative of an anatomical structure, it is significant to acquire more accurate information about a ribcage boundary and landmarks. It has therefore been strongly desired that the more accurate information be available for computer-aided diagnosis.
SUMMARY OF THE INVENTION
An object of the present invention is to prevent erroneous detection of a ribcage boundary, which was seen in the conventional technique, thereby acquiring accurate ribcage boundary information that will not be influenced by quality of chest images.
In order to realize the above object, one aspect of the present invention is provided by a method for detecting a ribcage boundary for computer-aided diagnosis requiring anatomical structure information to be detected from a digital chest image. The method comprises the steps of obtaining a profile of smoothed pixel-value integrated averages in each of right and left lung field of the image; deciding a threshold for each of the right and left lung fields with the profile taken as a reference; and searching each of the right and left lung fields from a central part of each lung field outwardly on the chest image so as to determine a position exceeding the threshold, the position being set as an outer search limit point for a series of an upper lung ribcage boundary candidate points for the ribcage boundary.
Preferably, the searching step further includes the steps of calculating a first derivative of the profile for examining changes in pixel value within each of the right and left lung fields in cases the position exceeding the threshold is not found; searching not only a position from the central part of each of the right
Mantis Mercader Eleni
Mitsubishi Space Software Co., Ltd.
Wenderoth , Lind & Ponack, L.L.P.
LandOfFree
Detection of ribcage boundary from digital chest 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 Detection of ribcage boundary from digital chest image, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Detection of ribcage boundary from digital chest image will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3350213