Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
2000-06-07
2003-12-23
Mariam, Daniel (Department: 2621)
Image analysis
Pattern recognition
Feature extraction
C382S128000, C382S287000, C382S311000, C382S201000
Reexamination Certificate
active
06668083
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to a method of deriving geometrical data of a structure from an image of the structure, in which
marker points are selected in the image,
a typical contour is associated with the marker points, and
geometrical data is calculated from the typical contour.
The invention also relates to a data processor which is arranged to carry out such a method.
BACKGROUND INFORMATION
A method of this kind is known from the article “
Digital radiography segmentation of scoliotic vertebral body using deformable models
”, published in SPIE Vol. 3034, 1997.
The known method is intended notably for deriving the degree of deformation, relative to a healthy spinal column, from an image of a spinal column of a patient to be examined. According to the known method, every vertebra is approximately by a deformed standard template. The templates are slightly rounded trapezia having dimensions which correspond to statistically mean dimensions of the respective vertebrae. Such a statistical mean value is determined from the dimensions of the vertebrae of a large number of patients having a spinal column deformed by scoliosis. The respective standard templates are deformed by scaling the height and the individual parallel sides in order to make them correspond as well as possible to the relevant vertebrae in the image. The standard templates are characterized by their height and a number of scale variables. The dimensions of the sides of the standard templates and the radius of curvature of concave sides of the standard templates are respective products of the height and one of the scale variables. Separate values of the height and the scale variables exist for different vertebrae. The known method thus takes into account the fact that the spinal column comprises vertebrae having different shapes and dimensions. The values of the scale variables have been estimated on the basis of data known from literature. The values for the heights have been derived from a large number of X-ray images of scoliotic patients. The typical contour corresponding to the edge of the spinal column in the image is derived from the standard templates. Furthermore, specific positions are indicated on the derived typical contour. A three-dimensional reconstruction of the imaged spinal column is carried out on the basis of said specific positions.
In order to carry out the known method it is necessary to determine characteristic dimensions of a large number of vertebrae in spinal columns of a large number of patients so as to form the standard templates. In order to make the standard templates correspond to the vertebrae in the image, the central axis of the spinal column is estimated by looking and a number of marker points is indicated on the estimated central axis. The central axis is subsequently approximated by a cubic curve through the indicated marker points on the central axis. Subsequently, the standard templates are deformed in such a manner that they are distributed as well as possible along the cubic curve. This makes the known method rather cumbersome. Because the known method utilizes standard templates relating to an average spinal column suffering from scoliosis, the known method is not very well suitable for accurately deriving spatial geometrical data relating to a spinal column exhibiting serious anomalies.
Notably when the image concerns an X-ray image of the spinal column it often occurs that a part of the marker points is not visible. The contrast of the vertebrae reproduced in an X-ray image often is not sufficient to see the marker points. Even if all marker points are visible in the X-ray image, usually not all of them are visible clearly enough to enable accurate indication of their position in the image.
OBJECTS AND SUMMARY OF THE INVENTION
It is an object of the invention to provide a method of deriving geometrical data of a structure from an image of the relevant structure which is less cumbersome than the known method and, moreover, enables more accurate results to be obtained.
This object is achieved by means of a method of deriving geometrical data of a structure from an image of the relevant structure according to the invention which is characterized in that
correctness values are associated with the marker points,
a cost function is associated with the typical contour,
which cost function is dependent on the shape of the typical contour and on the values of the correctness values, and
which typical contour is chosen in such a manner that the cost function has an optimum value.
For each of the selected marker points the correctness values represent the likelihood that the selected marker point accurately corresponds to a typical anatomical detail of the structure. The use of the method according to the invention notably concerns the determination of spatial geometrical data of the spinal column of a patient to be examined. In this application the marker points are selected to be situated at the positions in the image which correspond to the corner points of the vertebrae and to anatomically typical points at the edges of the relevant vertebral body. It appears that the marker points are not always suitably visibly reproduced in the image; this occurs notably when the image is an X-ray image of the spinal column. It is to be noted, however, that the marker points are not always clearly visible either in, for example magnetic resonance images of the spinal column. By associating the typical contour with the marker points the method according to the invention takes into account that the fact that not all marker points correspond equally accurately to the relevant positions in the spinal column. Upon association of the typical contour the value of the cost function is calculated and the typical contour is determined in such a manner that the cost function has an optimum value. The cost function is dependent on the correctness values of the marker points. It is thus achieved that the typical contour for which the cost function is optimized is determined predominantly by marker points having a high correctness value, i.e. by marker points for which it is rather certain that they accurately correspond to their relevant anatomically typical positions in the structure, such as the spinal column. The typical contour accurately corresponds with the boundary of the structure at issue in the image. Often, such a typical contour is a closed loop, e.g. when the structure is a vertebra. The typical contour may also comprise separate disconnected curves or straight line portions. For example, the correctness value of individual marker points is dependent on the differences between the brightness value in the marker point and brightness values in positions in the image in the vicinity of the relevant marker point. Generally speaking, suitable marker points correspond to anatomically typical positions; in that case differences between the brightness value in the relevant marker point and in individual directions adjacent the relevant marker point have values which are more or less predetermined. For example, such differences occur when the marker point is situated at the corner of a bone, such as a vertebra, which is surrounded by softer and less dense tissue, such as muscular tissue or lung tissue. The better the correspondence between such differences in the image and the predetermined differences, the higher the likelihood of accurate correspondence between the relevant marker point and the corresponding typical anatomical position. Therefore, the correctness values are preferably dependent on the predetermined differences or gradients in separate directions of brightness values in the vicinity of the relevant marker point. Marker points for which it is uncertain whether they correspond to their relevant anatomical position have a comparatively small effect only on the typical contour. Because the cost function is also dependent on the shape of the typical contour, it is achieved that the association of a typical contour having improbable shapes is counteracted and that,
Devillers Sylvain
Lacour Philippe
Makram-Ebeid Sherif
Verdonck Bert L. A.
Koninklijke Philips Electronics , N.V.
Mariam Daniel
Vodopia John
LandOfFree
Deriving geometrical data of a structure from an 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 Deriving geometrical data of a structure from an image, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Deriving geometrical data of a structure from an image will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3145443