JPS6382630A - Automatic extraction of body contour - Google Patents

Abstract

(57) [Abstract] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention performs threshold processing and differential processing on a digital image depicting a cross section of a patient and a bed obtained by an X-ray CT scanner, etc. The present invention relates to a body contour extraction method in an image processing device that generates and extracts the contours of various internal organs and the contour of a bed as binary images of a patient's body contour.

[Conventional technology]

Conventionally, in radiation treatment planning systems, X-rays, C
07 image data obtained by T scanner etc. to NoS,
In order to create the patient's body contour data necessary for treatment planning, coordinate data is manually entered using a digitizer, light pen, etc. based on the film containing the 07 image or the 07 image displayed on the display. was used.

[Problem that the invention seeks to solve]

The conventional body contour extraction method described above not only requires a large amount of manpower and attention, but also has a high possibility of inducing human errors due to parallax and carelessness during input, and the coordinate data obtained in this way The body contour data created based on this has the disadvantage that it causes errors in the radiation treatment plan and causes the wrong radiation treatment to be performed for 11m patients.

[Means for solving problems]

The automatic body contour extraction method of the present invention uses a point surrounded by the body contour on the generated binary image as the origin, and determines the position of each contour data constituting the two images as the radial length from the origin, The retained polar coordinate data is stored as polar coordinate data with a declination set at a predetermined pitch from a half-line with the origin as the starting point, and the declination is Search by incrementing at a predetermined pitch, and if there is one piece of polar coordinate data for a declination angle, use that data, if there are multiple pieces of polar coordinate data, use the maximum radial length as the body contour data for that declination angle, and if there are 0 pieces of polar coordinate data, If so, the declination is further incremented as a missing part, and after detecting the two body contour data, extrapolation is performed from the body contour data at the two most recent and smaller declinations among the polar coordinate data at each declination. The polar coordinate data corresponding to the data determined by the method is used as body contour data, and if there is no corresponding polar coordinate data, the argument is further incremented to set the next body contour data, and for missing parts, the argument is Polar coordinate data is set by interpolation from recent polar coordinate data that sandwich the polar coordinates, and the body contour is detected from the body contour data corresponding to each of these declination angles.

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 shows a case where the automatic body contour extraction method of the present invention is applied.
Two images subjected to threshold processing and differential processing,
FIG. 3 is a diagram showing a relationship with coordinates on a binary image.

Two threshold values T, ,
Set T2. These two thresholds T, , T
2 is used to perform the threshold processing shown in equation (1) to create three images composed of data values DATAn (which take any value of a, b, or c).

if DATA>Tl then DATAn
= aif T2≦DATA≦Tl than
DATAn = b − (1) if DATA<T
2 then DATAn = c However, formula (1
), DATA is the data value of each picture element constituting the base 07 image, and DATAn is the data value of each picture element of the three images created by threshold processing.

FIG. 2 is a diagram showing a picture element E to be processed and eight picture elements A, B, . . . , K in the vicinity of picture element E. For each picture element E of the three images obtained as described above, the neighboring picture elements B, F, H,
D is used to perform the differential processing shown in equation (2) to obtain a differential image En
Create.

In addition, another threshold value T3 is added to this differential image En.
Perform threshold processing using , and create two images.

As illustrated in FIG. 1, the two images obtained in this way include the patient's body contour as well as the threshold values T, , T2 . The outlines of various organs within the patient's body according to T3, the outline of the bed on which the patient was sleeping when the CT image was taken by the X-ray GT scanner, etc. are also depicted.

Processing is performed to automatically detect only the body contours of the two images. The method will be explained below.

As shown in FIG. 1, coordinate axes X and Y are set whose origin is one point surrounded only by the body contours of the two images of matrix MXN. For each of the picture elements that make up the two images, it is checked whether the picture element exists on some outline using the data value of that picture element, and if the picture element is on the outline, From the values of column number I and line number J in the matrix MXN of that picture element, the value of the coordinates (X, Y) of that picture element is determined using equation (3).

However, 1≦I≦M 1≦J≦N Furthermore, the radius of polar coordinates, which is the distance from the origin of the XY coordinate system to the picture element, is determined by the formula (0.

y= drTuxS = (4) where S is the length of one side of one picture element. The radius vector data γ for all picture elements on the contour obtained in this way are based on the angle α=k・θ (k=0.1, 2.~.n; θ is a predetermined pitch angle). , organize by angle α and store in memory. For any angle α (0 degrees ≦ α < 360 degrees), calculate the number of picture elements on various contours that exist in that direction.
In general, Q takes a value of O or 1 or more, and therefore the number of the radius vector data γ is also O or 1 or more.0 A predetermined angle α0 with only the body contour and organs is the body contour data. The search start angle is sequentially incremented by a predetermined pitch θ and the radius vector data γ at each angle α is checked. The search start angle is set as the search start angle, and the search is performed for 360° from the search start angle.

At the start of the search, if there is one radius vector data, it is used as body contour data, if there are multiple vectors, the one with the largest vector radius is used as body contour data, and if it is zero, the angle α is incremented as a missing part. continue. After two pieces of body contour data are detected, if radius vector data exists, it is extracted by extrapolation from the second radius vector data γ that has already been detected and stored as body contour data at two smaller angles. The calculated radius vector data γ corresponding to the predicted value of the radius vector data γ at the angle α of the current processing target is stored as body contour data at the angle α.

The above processing is performed for all angles α between 3130 degrees from the search start angle, and radius vector data corresponding to one body contour is obtained for each angle α. For the angle α of the defective part for which there is no radial data corresponding to the body contour, the radial vector data γ has already been set.

The radius vector data γ of the angles α on both sides thus obtained are determined by interpolation.

Finally, smoothing processing is performed to complete automatically extracted body contour data.

〔Effect of the invention〕

As explained above, the present invention provides X-ray C? The data is converted into polar coordinates from two images obtained from digital images such as a scanner, and since the body contour is the outermost part of the body except for the bed, the data is converted from the direction where the bed is not shown, and the polar coordinate declination is incremented. The body contour can be determined automatically by extracting only the data for one body contour and generating data for declination angles for which no data is found. 11
Le pay front table 1 + yot group person ke u # l + snake? ? f
It is possible to prevent errors and input errors, and it is possible to prevent errors in radiation treatment plans that require patient body contour data and radiation treatments performed based on the plans, and to improve accuracy. There is.

[Brief explanation of the drawing]

Figure 1 is a diagram showing the relationship between two images subjected to threshold processing and differential processing, to which the automatic body contour extraction method of the present invention is applied, and the coordinates on the binary image. It is a diagram showing a picture element E to be processed and eight picture elements A, B, -, K in the vicinity of picture element E. ■... Column number, J... Line number, (x, y)... Coordinates, γ... Radius, α... Angle (declination), α0... Search start angle, E ... Picture elements to be processed, A, B, C, D, F, G, H, K... Picture elements. Patent applicant: NEC Corporation - Nitno=

Claims (1)

[Claims] Threshold processing and differential processing are performed on a digital image depicting a patient's cross section and bed obtained by an X-ray CT scanner, etc., to obtain the patient's body contour, the contours of various organs within the body contour, and the bed. In an image processing device that generates and depicts a contour as a binary image, a point surrounded by the body contour on the generated binary image is set as the origin, and the position of each contour data making up the binary image is calculated from the origin. It is retained as polar coordinate data that has a radial length and a declination angle set at a predetermined pitch from a half-line starting from the origin, and the retained polar coordinate data is searched in one direction where only the body outline and organs are located. Search as a point by sequentially incrementing the declination at a predetermined pitch, and if there is one piece of polar coordinate data for the declination, use that data, or if there are multiple pieces of polar coordinate data, use the maximum radial length as the body contour of that declination. data, and if there is zero, the declination is further incremented as a missing part, and after detecting two body contour data, the two most recent and smaller declinations of the polar coordinate data at each declination are The polar coordinate data corresponding to the data obtained by extrapolation from the body contour data in is set as body contour data, and if there is no corresponding polar coordinate data, the declination is further incremented and the next body contour data is set, and the missing part is On the other hand, there is an automatic body contour extraction method that sets polar coordinate data by interpolation from recent polar coordinate data that sandwich the declination angle, and detects the body contour from the body contour data corresponding to each of these declination angles.