JP2007068714A - Medical image signal processing method, medical image signal processing apparatus, and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a medical image signal processing method, a medical image signal processing apparatus, and a computer program for a stable interpolation of processing signals of medical images. <P>SOLUTION: Since the number of the maximum terms of an interpolation polynomial for interpolation of the processing signals of the medical images is dynamically acquired, the number of the maximum terms of the most suitable polynomial for interpolation of the processing signals of the medical image is determined. Thus, a stabler interpolation of the processing signals of the medical images is performed. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a medical image signal processing method, a medical image signal processing apparatus, and a computer program, and is particularly suitable for use in interpolating values of medical image signals.

  In the time-dependent CAD (Computer Aided Diagnosis) technique of an X-ray image, first, a first image obtained by imaging a part (for example, lung region) of a common subject (subject). For the second image, a plurality of sets of corresponding points are obtained. Then, a shift vector representing a positional deviation amount between corresponding points is obtained, and further interpolated to correct deformation of the lung region caused by various factors such as a change in posture of the subject and breathing, Difference processing is performed on the first image and the second image. In this way, a portion in which both the first image and the second image are changed is depicted in the difference image.

  At this time, by using the temporal difference technique, common normal structures such as bones and blood vessels can be offset in two images (the first image and the second image), and only a change in lesion can be extracted. Therefore, by using the time-difference CAD technology, it is possible to expect clinically early detection of lesions, particularly lesions hidden in normal structures such as ribs and blood vessels, prevention of missed lesions, and rapid interpretation.

  FIG. 15 shows an outline of the algorithm of the time difference CAD.

  In FIG. 15, the image processing system includes an X-ray imaging apparatus 10, a time difference processing apparatus 20, and a display apparatus 30. The X-ray imaging apparatus 10 is for obtaining an X-ray image of a subject by irradiating the subject with X-rays.

  The temporal difference processing device 20 has an image input unit 201, a matching unit 202, a shift vector interpolation unit 203, an image warping unit 204, and a difference processing unit 205 as functions for realizing processing in the temporal difference processing device 20. And have.

  The image input unit 201 inputs an X-ray image obtained by the X-ray imaging apparatus 10. For example, the image input unit 201 inputs two X-ray images of the same part of the subject taken at different times. The matching unit 202 aligns one X-ray image input by the image input unit 201 with the other image. Specifically, the matching unit 202 first roughly aligns both images to obtain an overall shift amount of the examination region. Based on this shift amount, a plurality of template ROIs are set for one X-ray image, and a plurality of search ROIs are set for corresponding points in the other image. Next, an area that most closely matches the template ROI is searched in the search ROI, and the horizontal / vertical movement amount between the center of the template ROI and the center of the area is set as a shift vector. By performing this process for each of the plurality of template ROIs, a plurality of shift vectors representing the positional deviation between the two X-ray images at each corresponding point are obtained as shown in FIG.

  The shift vector interpolation unit 203 performs two-dimensional polynomial interpolation on the horizontal and vertical components of the plurality of shift vectors detected by the matching unit 202.

Here, in the shift vector interpolation technique performed using a two-dimensional polynomial, each input point is used as a sample point, and each sample point is obtained by using the two-dimensional data at each sample point by the least square error method or the like. The coefficients of each polynomial term are obtained so as to minimize the input data and the interpolation data at, and the input two-dimensional data is approximated. In the interpolation of the shift vector using the two-dimensional polynomial, the coefficient is obtained over the entire region where the corresponding points exist, so that the input two-dimensional data can be smoothly interpolated and is hidden in the two-dimensional data. Noise can be suppressed to some extent. (See Patent Document 1)
Next, the warping unit 204 uses the polynomial obtained by the shift vector interpolation unit 203 to calculate the corresponding position of each pixel of one image in the other image, performs a warping process, and the difference processing unit 205 performs the other image. One image warped to the other is subtracted from the other image, and the difference result is displayed on the display device 30.
JP-A-7-37074

  In the time difference technique, in order to obtain a shift vector at the corresponding point (sample point), for example, a matching technique such as a cross-correlation operation is often used for the template ROI, data, and search ROI. In order to improve the alignment accuracy, it is necessary to obtain shift vectors at many corresponding points. However, when interpreting medical images, it is practical to be able to create differential images at high speed on the spot to improve throughput. This is desirable from the standpoint of optimization, and increasing the number of corresponding points has a problem that it takes time to execute due to an increase in the amount of computation.

  Here, the shift vector distribution tends to change relatively smoothly as a whole image. Therefore, by reducing the number of corresponding points, that is, the number of sample points, and lowering the maximum degree of the interpolation polynomial, the accuracy of alignment is reduced. The calculation time can be suppressed while ensuring the above. However, changes in the posture of the subject depicted in images taken at different times may be accompanied by complex deformations. For example, in Fig. 16, the upper right lung field shift vector points in the horizontal direction, which is different from the overall average direction. The alignment accuracy cannot be ensured.

  If the maximum degree of the interpolation polynomial is increased, there is a problem that if the number of sample points is small, the interpolation value will oscillate between the sample points. It was difficult to perform an interpolation process that can ensure alignment accuracy.

  The present invention has been made in view of the above circumstances, and an object thereof is to enable stable interpolation of a medical image processing signal.

In order to solve the above-described problem, the medical image signal processing method of the present invention includes a step of arranging a predetermined number of evaluation points at each sample point of a medical image processing signal;
A first interpolation value calculating step for obtaining an interpolation value at each of the sample points and the evaluation points;
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; It is characterized by having.

Further, the medical image signal processing method of the present invention includes a step of arranging a predetermined number of evaluation points at each sample point of the medical image processing signal;
A first interpolation value calculation step for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative step to obtain each overall error;
And an interpolation step for interpolating the processing signal of the medical image using the highest-order polynomial that minimizes the overall error.

The medical image signal processing apparatus of the present invention comprises means for arranging a predetermined number of evaluation points at each sample point of a medical image processing signal,
First interpolation value calculation means for obtaining an interpolation value at each of the sample points and evaluation points;
First error evaluation means for obtaining an error of an interpolation value at each sample point;
Second error evaluation means for obtaining an error of an interpolation value at the arranged evaluation points;
Total error evaluation means for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; It is characterized by having.

Further, the medical image signal processing apparatus of the present invention comprises means for arranging a predetermined number of evaluation points at each sample point of the medical image processing signal,
First interpolation value calculation means for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
First error evaluation means for obtaining an error of an interpolation value at each sample point;
Second error evaluation means for obtaining an error of an interpolation value at the arranged evaluation points;
Total error evaluation means for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative means of obtaining each of the total errors,
A medical image signal processing apparatus comprising: an interpolation step that performs interpolation processing of a medical image processing signal using a polynomial of the highest order that minimizes the overall error.

The computer program of the present invention includes a step of arranging a predetermined number of evaluation points at each sample point of a processing signal of a medical image,
A first interpolation value calculating step for obtaining an interpolation value at each of the sample points and the evaluation points;
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; A computer program for causing a computer to execute.

Further, the computer program of the present invention comprises a step of arranging a predetermined number of evaluation points at each sample point of the medical image processing signal,
A first interpolation value calculation step for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative step to obtain each overall error;
A computer program that causes a computer to execute an interpolation step of performing interpolation processing of a processing signal of a medical image using a polynomial of the highest order that minimizes the overall error.

  According to the present invention, in addition to the evaluation of the interpolation values at the sample points of the processing signal of the medical image, the interpolation values at the evaluation points arranged between the sample points are evaluated, and these are comprehensively evaluated. Therefore, regardless of only the interpolation accuracy at the sample point, the vibration of the interpolation data generated by the higher-order interpolation equation can also be measured, and the overall accuracy of the interpolation equation can be evaluated.

  Further, according to the present invention, since the maximum number of terms of the interpolation polynomial for interpolating the processing signal of the medical image is obtained dynamically, regardless of the magnitude of the data change at the corresponding point (sample point) Therefore, it is possible to determine the maximum number of polynomial terms that is optimal for interpolating the medical image processing signal, and to more stably interpolate the medical image processing signal. In addition, when obtaining a temporal difference signal representing a temporal change in a plurality of medical image signals, the number of sample points can be reduced, so that the processing speed can be improved.

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

(First embodiment)
FIG. 1 is a block diagram showing an example of a temporal difference processing system and a shift vector interpolation unit in the system according to the first embodiment of the present invention.

  In FIG. 1, the temporal difference processing system includes an X-ray imaging device 10, a temporal difference processing device 20, and a display device 30.

  In the temporal difference processing apparatus 20, the shift vector interpolation unit 203 includes an overall error evaluation unit 203a, an optimum term number measurement unit 203b, an interpolation formula determination unit 203c, and an interpolation value calculation unit 203d.

  FIG. 2 shows a block diagram of the total error evaluation unit 203a.

  The data input unit 1001 reads the shift vector from the matching unit 202, and uses the horizontal and vertical shift amounts of the shift vector at each corresponding point and the coordinate position of the corresponding point as sample point data, respectively. 1002 and the evaluation point insertion unit 1005.

  The first shift vector interpolation unit 1002 obtains a polynomial coefficient by the least square error method using the provisional maximum term number K set by the provisional term number setting unit 1004 over the entire area occupying the sample points.

Here, the interpolation polynomial f (x, y) is shown in Equation (1). Here, m is the highest degree of the abscissa x of each interpolating polynomial term x ⁱ y ^j , and n is the highest degree of the ordinate y. K is the maximum number of terms in the interpolation polynomial. Polynomial interpolation uses the coordinate position as a variable and calculates the value of the position.

f (x, y) = c _K x ^m y ⁿ + c _K−1 x ^m−1 y ⁿ + Λc ₃ y + c ₂ x + c ₁ (1)
However, K = (m + 1) (n + 1)
In order to obtain each coefficient of the polynomial, the first shift vector interpolation unit 1002 solves the equation shown in Equation (2). In equation (2), L is the number of sample points. The sample points are arranged in two-dimensional data such as an image, but are rearranged in the scanning order and handled as one-dimensional data. As described above, the number of sample points L is larger than the maximum number K of terms in order to compensate for errors in the shift vector obtained in the matching unit 202 in this embodiment, and the polynomial coefficient c _k (k = 1,..., K ) Cannot be uniquely determined, each coefficient of the interpolation polynomial is obtained by Equation (4) using the least square error method that minimizes the weighted square error S at each sample point shown in Equation (3). Here, w _i is the weight of the i-th sample, and W is a diagonal matrix in which w _i are arranged diagonally. When calculating the ROI matching by the matching unit 202, w _i is determined on the basis of the amount of good matching, for example, the size of the cross-correlation coefficient when cross-correlation is used for matching.

Ｃ＝（Ｘ^TＷＸ）⁻¹Ｘ^TＷＵ
次に、第1シフトベクトル補間部１００２は求めた多項式係数を、サンプル点誤差計算部１００３と評価点誤差計算部１００７に出力し、それぞれ、サンプル点と評価点の多項式補間値を計算する。

C = (X ^T WX) ⁻¹ X ^T WU
Next, the first shift vector interpolation unit 1002 outputs the obtained polynomial coefficients to the sample point error calculation unit 1003 and the evaluation point error calculation unit 1007, and calculates the polynomial interpolation values of the sample points and the evaluation points, respectively.

The sample point error calculation unit 1003 receives the polynomial coefficient from the first shift vector interpolation unit 1002, calculates the interpolation value v _i at all sample points from the equation (5), and calculates the interpolation value and data from the equation (6). Weighted sum of squares with true value from input unit 101

をサンプル点の総誤差とする。なお、本計算部で、重みづけ二乗和ではなく、単に真値と補間値の二乗和を総誤差としてもよい。

Is the total error of the sampling points. In this calculation unit, the total error may be simply the sum of squares of the true value and the interpolated value instead of the weighted sum of squares.

v _i = c _K x _i ^m y _i ^m + c _K−1 x _i ^m−1 y _i ⁿ + Λc ₃ y _i + c ₂ x _i + c ₁ i = 1, Λ, L (5)

一方、評価点挿入部１００５は、データ入力部１００１からのサンプル点位置データを受け、評価点を挿入する。図３に示すように、w×hの検査領域にN個サンプル点がある場合、評価点挿入部１００５は、検査領域をｌ×ｌの正方形区間に分割する。検査領域の周辺に領域のサイズに応じて、分割区間の大きさを変更することがある。

On the other hand, the evaluation point insertion unit 1005 receives sample point position data from the data input unit 1001 and inserts evaluation points. As shown in FIG. 3, when there are N sample points in the w × h inspection area, the evaluation point insertion unit 1005 divides the inspection area into l × l square sections. The size of the divided section may be changed in accordance with the size of the area around the inspection area.

分割された区間にサンプル点の有無をチェックして、indexマップを作成する。サンプル点がある場合Indexを1にし、ない場合、Indexを0とする。

Check the existence of sample points in the divided sections and create an index map. If there is a sample point, set Index to 1, otherwise set Index to 0.

  The second shift vector interpolation unit 1006 receives the sample point data and coordinates from the data input unit 1001 and the evaluation point coordinates from the evaluation point insertion unit 1005, and uses the local interpolation method to insert the first evaluation point at the inserted evaluation point. 2 is obtained. As the local interpolation method, the nearest neighbor pixel interpolation method may be used, but bilinear interpolation is used in order to predict the data value at the evaluation point with higher accuracy. Hereinafter, the local interpolation method will be described by taking bilinear interpolation as an example.

  FIG. 4 shows a block diagram of the second shift vector interpolation unit 1006.

  The representative sample point determination unit 2001 selects one sample point in the section with Index == 1, and records the position, value, and weight information using the sample point as the representative sample point.

Then evaluation point coordinate determination unit 2002 arranges the evaluation points P _e in the upper left of the Index == 0 interval. The sample point search unit 2003 in the vicinity of the evaluation point is centered on the arranged evaluation point _Pe , and as shown in FIG. 5, the sample point P ₀ P ₁ P ₂ P ₃ nearest to each of the four quadrants in zigzag order. Explore.

As shown in FIG. 6, the position coordinate normalization unit 2004 normalizes the coordinates of the four sample points P ₀ and P ₁ P ₂ P ₃ to a unit square Q ₀ Q ₁ Q ₂ Q ₃ . A normalized position Q _e corresponding to the evaluation point P _e is also obtained. Hereinafter, as shown in FIG. 7, the coordinates of the sample points P ₀ P ₁ P ₂ P ₃ are (u ₀ , v ₀ ), (u ₁ , v ₁ ), (u ₂ , v ₂ ), (u ₃ , v ₃ ), and a method of obtaining the normalized coordinates Q _e (t, s) of the evaluation point P _e (u _e , v _e ) will be described.

The coordinate value (t, s) in the normalized coordinates of the evaluation point _Pe is obtained by the following equation.

各直線上の各点は直線関係があるので、式(８)は以下式(９)になる。

Since each point on each straight line has a linear relationship, equation (8) becomes equation (9) below.

tを求めるために、以下の方程式を用いる。

To find t, we use the following equation:

ここで、式(１１)を用いて、式(１０)を変換すると、
ｘ＝ｕ−ｕ₀
ｙ＝ｖ−ｖ₀
ｘ’＝ｕ−ｕ₁
ｙ’＝ｖ−ｖ₁ （１１）
（x₃y₂'−x₂'y₃）・t²＋(x_e'y₃＋x₂'y_e−x_ey₂'−x₃y_e')・t＋(x_ey_e'−x_e'y_e)＝0 （１２）
式(１２)を解け、正の解をtの値とする。

Here, when Formula (10) is converted using Formula (11),
x = u−u ₀
y = v−v ₀
x ′ = u−u ₁
y ′ = v−v ₁ (11)
(X ₃ y ₂ '−x ₂ ' y ₃ ) ・ t ² + (x _e 'y ₃ + x ₂ ' y _e −x _e y ₂ '−x ₃ y _e ') ・ t + (x _e y _e '− x _e 'y _e ) = 0 (12)
Equation (12) is solved, and the positive solution is taken as the value of t.

  Similarly, the value of s is determined using the following equation:

同じく、式(9)を用いて式(8)を変換すると、
ｘ＝ｕ−ｕ₀
ｙ＝ｖ−ｖ₀
ｘ”＝ｕ−ｕ₃
ｙ”＝ｖ−ｖ₃ （１４）
（x₂”y₁−x₁y₂”）・s²＋(x_ey₂”＋x₁y_e”−x₂”y_e−x_e”y₁)・s＋(x_e”y_e−x_ey_e”)＝0 （１５）
同じく、正の解をsの値とする。

Similarly, if equation (8) is transformed using equation (9),
x = u−u ₀
y = v−v ₀
x ″ = u−u ₃
y ″ = v−v ₃ (14)
(X ₂ ”y ₁ −x ₁ y ₂ ”) · s ² + (x _e y ₂ ”+ x ₁ y _e ” −x ₂ ”y _e −x _e ” y ₁ ) · s + (x _e ”y _e − x _e y _e ”) = 0 (15)
Similarly, let the positive solution be the value of s.

Finally, the interpolation unit 2005 uses the data (d ₀ , d ₁ , d ₂ , d ₃ ) of the sample points P ₀ P ₁ P ₂ P ₃ to calculate the second evaluation point by the equation (16). Calculate the interpolation value. When the number of sample points is three or less in the vicinity of the evaluation point, the value of the evaluation point is calculated using Affine transformation or nearest neighbor pixel interpolation. Similarly, evaluation point weights are calculated from neighboring sample point weights.

d _e = d ₀ + (d ₃ −d ₀ ) · t + (d ₁ −d ₀ ) · s + (d ₂ −d ₁ −d ₃ + d ₀ ) · t · s (16)
The evaluation point error calculation unit 1007 calculates the polynomial interpolation value v _i at the evaluation point using the polynomial interpolation coefficient from the first shift vector interpolation unit 102, and simultaneously calculates the local interpolation value v ′ _i at the evaluation point. Received from the second shift vector interpolating unit 1006, the total error at the evaluation point is calculated by Expression (17). However, L 'is a score of evaluation points.

図８に、一列のシフトベクトル水平方向移動量を用いて、多項式補間と局所的な補間の違いを示す。黒い点はサンプル点の値を示し、白い点は評価位置における局所的な補間結果の仮真値である。円滑な曲線は多項式補間の結果である。サンプル点の値と多項式補間結果の差の重み付け二乗和をサンプル点の総誤差とし、評価点における仮真値と多項式補間結果の差の重み付け二乗和を評価点の総誤差とする。

FIG. 8 shows the difference between polynomial interpolation and local interpolation using a single shift vector horizontal movement amount. A black point indicates the value of the sample point, and a white point is a provisional true value of a local interpolation result at the evaluation position. The smooth curve is the result of polynomial interpolation. The weighted square sum of the difference between the sample point value and the polynomial interpolation result is defined as the total error of the sample points, and the weighted square sum of the difference between the provisional true value and the polynomial interpolation result at the evaluation point is defined as the total error of the evaluation point.

The total error calculation unit 1008 calculates an error E _K obtained by combining the sample points and the evaluation points in the maximum number of terms _K by Expression (18). The calculated total error E _K is stored in the storage unit 1009. Here, w is the weight of the total error at the sample points, and may be determined so that the total error at the evaluation points can be sufficiently considered empirically.

続いて、最適項数測定部２０３ｂは、記憶部１００９に記憶された総合誤差E_Kを読み出して、設定された全ての仮最大項数Kに渡って、E_Kの最小値を探し、それに対応する項数ｋを最適項数とする。

Subsequently, the optimum term number measurement unit 203b reads the total error E _K stored in the storage unit 1009, searches for the minimum value of E _K over all the set provisional maximum terms K, and responds to it. The number of terms k to be used is the optimum number of terms.

  The interpolation formula determination unit 203c sets the optimum term number k obtained by the optimum term number measurement unit 203b as the maximum number of terms, and recalculates the optimum interpolation polynomial coefficient.

  Note that, as a modification of the present embodiment shown in FIG. 9, when the storage unit (2) 1010 stores the polynomial coefficient in each provisional maximum term number in the memory and outputs it to the processing unit, the interpolation formula determination unit 203c May simply read out the polynomial coefficient at the optimum number of terms k from the memory.

  The interpolation value calculation unit 203d calculates the value of the point to be interpolated over the entire area using the polynomial coefficient determined by the interpolation formula determination unit 203c.

  Finally, the image warping unit 204 warps the other image to one image as a reference using the shift vector obtained by the shift vector interpolation unit 203. The difference image processing unit 205 performs a difference between one image and the other warped image to create a difference image. Displayed on the display device 30.

  In this way, not only the interpolation results at the sample points of a plurality of images are evaluated, but also evaluation points are inserted between the sample points, and the interpolation results at the evaluation points are also evaluated, and these are comprehensively evaluated. As a result, the interpolation of the shift vector can be performed more stably than before, and the number of sample points can be reduced.

  A flowchart of this embodiment is shown in FIG.

  S101: The data input unit 1001 reads shift vector data from the matching unit 202, and outputs the horizontal and vertical components, weights, and coordinates of the shift vector to subsequent processing.

  S102: The evaluation point insertion unit 1005 determines that the inspection area size w × h and the number of sample points N are

の正方形の区間に分割する。画像の周辺に画像の境界により、分割区間のサイズをトリミングすることがある。分割された区間にサンプル点の有無をチェックして、サンプル点がある場合Indexを1にし、ない場合、Indexを0とし、indexマップを作成する。

Is divided into square sections. In some cases, the size of the divided section is trimmed around the image by the boundary of the image. Check whether there is a sample point in the divided section. If there is a sample point, set Index to 1, otherwise set Index to 0 and create an index map.

  S103: The representative sample point determination unit 2001 of the second shift vector interpolation unit 1006 determines whether or not all the divided sections of the Index map have been scanned. If all the divided sections have been scanned, the process proceeds to step S107. Proceed to S104.

  S104: The representative sample point determination unit 2001 of the second shift vector interpolation unit 1006 checks whether the Index of the main grid is 1, and if it is 1, the process proceeds to step S105. Otherwise, the process proceeds to step S106.

  S105: The representative sample point determining unit 2001 of the second shift vector interpolating unit 1006 selects one sample point for this grid and sets it as the representative sample point.

  S106: The representative sample point determination unit 2001 of the second shift vector interpolation unit 1006 moves to the next divided section and repeats the steps of selecting the representative sample points from S103 to S105.

  S107: The evaluation point coordinate determining unit 2002 of the second shift vector interpolating unit 1006 determines whether or not all the divided sections of the Index map have been scanned. If all of the divided sections have been scanned, the process proceeds to step S120. Proceed to S108.

  S108: The evaluation point coordinate determination unit 2002 of the second shift vector interpolation unit 1006 checks whether the Index of this grid is 0, and if it is 0, the process proceeds to step S109. Otherwise, the process proceeds to step S114.

  S109: The evaluation point coordinate determination unit 2002 of the second shift vector interpolation unit 1006 determines the upper left coordinate position of this grid as the evaluation point position.

S110: second evaluation points near the sample point searching unit 2003 of the shift vector interpolation unit 1006, the evaluation point centered on P _e, 4 quadrants closest to in zigzag order each sample point P ₀ P ₁ P ₂ P ₃ Explore.

S111: The position coordinate normalization unit 2004 of the second shift vector interpolation unit 1006 normalizes the coordinates of the four sample points P ₀ P ₁ P ₂ P ₃ to a unit square Q ₀ Q ₁ Q ₂ Q ₃ . A normalized position Q _e corresponding to the evaluation point P _e is also obtained.

  S112: The interpolation unit 2005 of the second shift vector interpolation unit 1006 calculates the evaluation point value using a local interpolation method.

  S113: The interpolation unit 2005 of the second shift vector interpolation unit 1006 calculates the weight of the evaluation point using a local interpolation method.

  S114: The evaluation point coordinate determination unit 2002 of the second shift vector interpolation unit 1006 moves to the next divided section, and repeats the steps from S107 to S113.

  S120: The provisional term number setting unit 1004 sets the provisional maximum term number k to 1.

  S121: The provisional term number setting unit 1004 determines whether the provisional maximum term number k is larger than a predetermined maximum value K. If it is larger, the process of the comprehensive error evaluation unit 203a is terminated, and the process proceeds to step S140. If so, go to step 122.

  S122: The first shift vector interpolation unit 1002 uses the sample point data to obtain a global polynomial coefficient.

  S123: The sample point error calculation unit 1003 calculates a polynomial interpolation value at each sample point.

  S124: The sample point error calculation unit 1003 determines the total error at the sample points.

を計算する。

Calculate

  S125: The evaluation point error calculation unit 1007 calculates a polynomial interpolation value at each evaluation point.

  S126: The evaluation point error calculation unit 1007 displays a total error at the evaluation point.

を計算する。

Calculate

S127: The total error calculation unit 1008 calculates a total error E _K at the sample points and the evaluation points.

S128: the storage unit 1009, stores the total error E _K, the output.

  S129: The provisional term number setting unit 1004 increases the provisional maximum number of items k and repeats the above processing up to a predetermined maximum value K.

S140: After obtaining the total error for all provisional maximum terms, the optimum term number measurement unit 203b searches for k that minimizes the total error E _K.

S141: interpolation equation determining unit 203c, using the k the stored E _K is minimized to step S128, to recalculate the polynomial coefficients.

  S142: The interpolation value calculation unit 203d interpolates the horizontal and vertical components or weights of the shift vector using the polynomial coefficient obtained in step S141.

In order to increase the speed, in this embodiment, in step S128, in addition to the total error E _K and the polynomial interpolation coefficient, the sample points obtained in steps S123 and S125 and the polynomial interpolation values at the evaluation points are also stored. In step S142, the data can be read and reused.

  The size of the divided section may be changed as long as the highest frequency component of the shift vector data can be reproduced.

  Further, in this embodiment, the representative sample point determination unit 2001 may not be provided, and the sample point search unit 2003 near the evaluation point may search for the nearest sample point in each of the four quadrants around the evaluation point. .

(Second Embodiment)
Next, a second embodiment of the present invention will be described. In the shift vector interpolation unit of the first embodiment, interpolation processing is performed while the shift vector data from the matching unit 202 remains two-dimensional. However, as shown in FIG. 11, the interpolation of two-dimensional data can be realized in the first stage of interpolating a row having sample points in the horizontal direction and the second stage of interpolating each row in the vertical direction. In the second embodiment of the present invention, the shift vector interpolation unit interpolates the shift vector by horizontal and vertical two-step interpolation.

  FIG. 12 shows a shift vector interpolation unit according to the second embodiment of the present invention. The difference from the first embodiment is a one-dimensional data conversion unit 203e. As shown in FIG. 11, the one-dimensional data conversion unit 203e first extracts the shift vector data in a row with each sample point in the horizontal direction, and outputs it to the total error evaluation unit 203a, 203a, 203b, 203c, 203d Through this process, the row is interpolated. Next, in the vertical direction, the data interpolated in each column is taken out, processed 203a, 203b, 203c, and 203d, and interpolated. The interpolated data is stored in the interpolation value calculation unit 203d, and is output to the image warping unit 204 when interpolation of all the columns is completed.

  In the total error evaluation unit 203a, the overall block diagram is the same as that in the first embodiment, but the evaluation point insertion unit 1005 and the second shift vector interpolation unit 1006 are different.

The input data of the evaluation point insertion unit 1005 is one-dimensional, and the evaluation points are inserted as shown in FIG. The start point coordinate P ₀ and the end point coordinate P _{n of the} inspection area are assumed to be set. If there are n sample points, the length of the segment

になる。分割された区間にサンプル点の有無をチェックして、indexマップを作成する。サンプル点がある場合Indexを1にし、ない場合、Indexを0とする。

become. Check the existence of sample points in the divided sections and create an index map. If there is a sample point, set Index to 1, otherwise set Index to 0.

  The second shift vector interpolation unit 1006 receives the sample point data and coordinates from the data input unit 1001 and the evaluation point coordinates from the evaluation point insertion unit 1005, and uses the local interpolation method to insert the first evaluation point at the inserted evaluation point. 2 is obtained. As a local interpolation method, one-dimensional nearest neighbor pixel interpolation method, linear interpolation method, Cubic Convolution or Cubic Spline method can be used.

  Further, a flowchart of the shift vector interpolation unit 203 of this embodiment is shown in FIG.

  S201: The one-dimensional data conversion unit 203e reads the two-dimensional data of the shift vector from the matching unit 202 and decomposes it into the horizontal and vertical components and weight data of the shift vector, along with the coordinate information in the horizontal and vertical directions. Convert to one-dimensional data and output to subsequent processing.

  S210: The data input unit 1001 of the total error evaluation unit 203a determines whether or not the processing from step S211 to step S228 has performed all sample rows in the horizontal direction. If all the steps have been performed, the process proceeds to step S240. If not, the process proceeds to step S211.

S211: The evaluation point insertion unit 1005 of the total error evaluation unit 203a determines the length in the horizontal direction based on the start point P ₀ , end point P _n of the inspection region and the number of sample points.

の区間に分割する。画像の周辺に画像の境界により、分割区間のサイズをトリミングすることがある。分割された区間にサンプル点の有無をチェックして、サンプル点がある場合Indexを1にし、ない場合、Indexを0とし、indexマップを作成する。

Is divided into sections. In some cases, the size of the divided section is trimmed around the image by the boundary of the image. Check whether there is a sample point in the divided section. If there is a sample point, set Index to 1, otherwise set Index to 0 and create an index map.

  S212: The evaluation point insertion unit 1005 of the total error evaluation unit 203a scans all the divided sections and determines an evaluation point position in a section where Index is 0. As an example, the center position of the section is set as the evaluation point position.

  S213: The second shift vector interpolation unit 1006 of the total error evaluation unit 203a determines a representative sample point in the section with Index = 1, and interpolates the evaluation point by a local interpolation method from the representative sample point near the evaluation point. . Examples of the local interpolation method include nearest neighbor pixel interpolation, linear interpolation, Cubic Convolution interpolation, and Cubic Spline interpolation.

  S220: The temporary term number setting unit 1004 of the total error evaluation unit 203a determines whether or not the maximum number of terms of the interpolation polynomial has become maximum. If the maximum value is reached, the process proceeds to step S226; otherwise, the process proceeds to step S221.

  S221: The first shift vector interpolation unit 1002 of the total error evaluation unit 203a obtains a global polynomial coefficient using the sample point data.

  S222: The sample point error calculation unit 1003 of the total error evaluation unit 203a calculates a polynomial interpolation value at each sample point, and the total error at the sample point.

を求める。

Ask for.

  S223: The evaluation point error calculation unit 1007 of the total error evaluation unit 203a calculates a polynomial interpolation value at each evaluation point, and uses the local interpolation value of the evaluation point calculated in step S213 to determine the total error at the evaluation point.

を求める。

Ask for.

S224: The total error calculation unit 1008 of the total error evaluation unit 203a calculates the total error E _K at the sample points and the evaluation points, and stores it in the storage unit 1009.

  S225: The provisional number setting unit 1004 of the total error evaluation unit 203a increases the number of provisional maximum terms of the polynomial, and repeats the processing from steps S221 to S224.

S226: After obtaining the total error for all provisional maximum terms, the optimum term number measurement unit 203b searches for k that minimizes the total error E _K.

S227: interpolation equation determining unit 203c, using the k the stored E _K is minimized to step S128, recalculates the polynomial coefficients, the interpolated value calculating unit 203d is, horizontal and vertical components of the shift vector or weights, Is interpolated and stored.

Also, as shown in FIG. 9, in S227, the polynomial coefficient of each provisional maximum term number is stored in the storage unit (2) 1010, and the interpolation formula determination unit 203c reads out from the storage unit (2) 1010 and shifts. Each component of the vector may be interpolated. Further, in S227, the interpolation value for each provisional maximum number of terms is stored in the storage unit (2) 1010, and when k that minimizes the total error E _K is determined, it may be read out at 203d.

  S228: The data input unit 1001 of the total error evaluation unit 203a increments the row number and moves to the next row.

  S240: The data input unit 1001 of the total error evaluation unit 203a determines whether all the columns have been processed in the vertical direction. If all the columns have been processed, the process proceeds to step S259. Otherwise, the process proceeds to step S241.

  Since the processing from S241 to S258 is the same as the processing from S211 to S228 except for the difference between the horizontal direction and the vertical direction, detailed description thereof is omitted here.

  S259: The interpolation value calculation unit 203d outputs the value of each interpolation point calculated in step S257.

  The length of the divided section may be changed as long as the highest frequency component of the shift vector data can be reproduced.

  In this embodiment, the representative sample point determination unit 2001 may not be provided, and the sample point search unit 2003 in the vicinity of the evaluation point may search for the nearest sample point with the evaluation point as the center.

  Further, the present invention may be applied to a system constituted by a plurality of devices (for example, a host computer, an interface device, an imaging device, a web application, etc.) or an apparatus constituted by a single device.

(Other embodiments of the present invention)
In order to operate various devices to realize the functions of the above-described embodiments, program codes of software for realizing the functions of the above-described embodiments are provided to an apparatus or a computer in the system connected to the various devices. What is implemented by operating the various devices according to a program supplied and stored in a computer (CPU or MPU) of the system or apparatus is also included in the scope of the present invention.

  In this case, the program code of the software itself realizes the functions of the above-described embodiments, and the program code itself and means for supplying the program code to the computer, for example, the program code are stored. The recorded medium constitutes the present invention. As a recording medium for storing the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used. When the present invention is applied to the recording medium, the recording medium stores program codes corresponding to the flowcharts described above.

  Further, by executing the program code supplied by the computer, not only the functions of the above-described embodiments are realized, but also the OS (operating system) or other application software in which the program code is running on the computer, etc. It goes without saying that the program code is also included in the embodiment of the present invention even when the functions of the above-described embodiment are realized in cooperation with the embodiment.

  Further, after the supplied program code is stored in the memory provided in the function expansion board of the computer or the function expansion unit connected to the computer, the CPU provided in the function expansion board or function expansion unit based on the instruction of the program code Needless to say, the present invention includes a case where the functions of the above-described embodiment are realized by performing part or all of the actual processing.

2 shows a temporal difference processing system and a shift vector interpolation unit of the system. It is a comprehensive error evaluation part block diagram. This is a division of the inspection area. FIG. 6 is a block diagram of a second shift vector interpolation unit. zigzag search order. It is a position coordinate normalization of bilinear interpolation. Normalization of evaluation point coordinates. It is the difference between polynomial interpolation and local interpolation at the evaluation point. It is a block diagram of a comprehensive error evaluation part deformation | transformation. 3 is a flowchart of the first embodiment. Each row and column for horizontal and vertical two-step interpolation. FIG. 10 is a block diagram of a shift vector interpolation unit of the second embodiment. Evaluation point insertion for one-dimensional data. 6 is a flowchart of the second embodiment. It is a block diagram of a time difference process. 3 is an X-ray image photograph showing a shift vector data sample.

Claims (23)

Placing a predetermined number of evaluation points at each sample point in a processing signal of a medical image;
A first interpolation value calculating step for obtaining an interpolation value at each of the sample points and the evaluation points;
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; A medical image signal processing method comprising:   The medical image processing signal is a shift vector signal that sets a plurality of regions of interest in a pair of image signals of the same part, performs matching processing, and indicates a horizontal and vertical movement amount in each region of interest. The medical image signal processing method according to claim 1. When separating the processing signal of the medical image into a one-dimensional signal, the step of arranging a predetermined number of evaluation points between the sample points,
Determining the length of the divided section according to the number of sample points, and dividing the section into sections of the same length as the processing signal of the one-dimensional medical image;
2. The medical image signal processing method according to claim 1, further comprising: searching for a section having no sample point in the divided section and arranging an evaluation point. 3. When the processing signal of the medical image is a two-dimensional signal, the step of arranging a predetermined number of evaluation points between the sample points is as follows:
Determining the area of the divided section according to the number of sample points, and dividing the section into areas of an area such as a processing signal of a medical image;
The medical image signal processing method according to claim 1, further comprising a step of searching for an interval having no sample point in the divided interval and arranging an evaluation point.   2. The medical interpolation according to claim 1, wherein the first interpolation value calculating step obtains a polynomial interpolation value at each sample point and evaluation point of the processing signal of the medical image using a polynomial as the interpolation formula. Image signal processing method.   2. The first error evaluation step obtains a difference between a first interpolation value of polynomial interpolation at each sample point and a value of a processing signal of a medical image at the sample point. Medical image signal processing method.   In the second error evaluation step, local interpolation is performed using a value and a weight of a processing signal of a medical image at a sample point arranged in the vicinity of the evaluation point, and a second interpolation value at the evaluation point is performed. 2. The medical image signal processing method according to claim 1, wherein a difference from the first interpolation value obtained using the polynomial is obtained.   8. The medical image signal according to claim 7, wherein when the processing signal of the medical image is separated into a one-dimensional signal, the local interpolation method is nearest neighbor pixel interpolation, linear interpolation, or cubic interpolation. Processing method. When the processing signal of the medical image is a two-dimensional signal, the local interpolation method is:
Creating an index map indicating the presence or absence of sample points in the divided area;
Selecting one sample point in an area with the sample point;
In the four quadrants centered on the evaluation point, respectively, a step of searching for a sample point closest to the evaluation point,
The medical image signal processing method according to claim 7, further comprising a step of interpolating from the searched sample points using nearest pixel interpolation or bilinear interpolation.   The medical image signal processing method according to claim 1, wherein the total error evaluation step for obtaining the total error calculates a weighted average of the error at the sample point and the error at the evaluation point. Placing a predetermined number of evaluation points at each sample point of the medical image processing signal;
A first interpolation value calculation step for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative step to obtain each overall error;
A medical image signal processing method comprising: an interpolation step for performing interpolation processing of a medical image processing signal using a polynomial of the highest order that minimizes the overall error. Means for arranging a predetermined number of evaluation points at each sample point of the processing signal of the medical image;
First interpolation value calculation means for obtaining an interpolation value at each of the sample points and evaluation points;
First error evaluation means for obtaining an error of an interpolation value at each sample point;
Second error evaluation means for obtaining an error of an interpolation value at the arranged evaluation points;
Total error evaluation means for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; A medical image signal processing apparatus comprising: When separating the processing signal of the medical image into a one-dimensional signal, means for arranging a predetermined number of evaluation points between the sample points,
Means for determining the length of the divided section by the number of sample points, and dividing the processing signal of the one-dimensional medical image into equal-length sections;
The medical image signal processing apparatus according to claim 12, further comprising means for searching for a section having no sample point in the divided section and arranging an evaluation point. When the processing signal of the medical image is a two-dimensional signal, means for arranging a predetermined number of evaluation points between the sample points
Means for determining the area of the division section based on the number of sample points, and dividing the section into areas of the medical image processing signal and the like;
13. The medical image signal processing apparatus according to claim 12, further comprising means for searching for a section having no sample points in the divided sections and arranging evaluation points.   13. The medical interpolation according to claim 12, wherein the first interpolation value calculation means obtains a polynomial interpolation value at each sample point and evaluation point of the processing signal of the medical image using a polynomial as the interpolation formula. Image signal processing device.   The first error evaluation unit obtains a difference between a first interpolation value of polynomial interpolation at each sample point and a value of a processing signal of a medical image at the sample point. Medical image signal processing apparatus.   The second error evaluation means performs local interpolation using a value and a weight of a processing signal of a medical image at a sample point arranged in the vicinity of the evaluation point, and a second interpolation value at the evaluation point The medical image signal processing apparatus according to claim 12, wherein a difference from the first interpolation value obtained using the polynomial is obtained.   18. The medical image signal according to claim 17, wherein when the processing signal of the medical image is separated into a one-dimensional signal, the local interpolation unit is nearest neighbor pixel interpolation, linear interpolation, or cubic interpolation. Processing equipment. When the processing signal of the medical image is a two-dimensional signal, the local interpolation means is
Means for creating an index map indicating the presence or absence of sample points in the divided area;
Means for selecting one sample point in a region with the sample point;
In the four quadrants centered on the evaluation point, respectively, the search means for the sample point closest to the evaluation point,
18. The medical image signal processing apparatus according to claim 17, further comprising means for interpolating from the searched sample points using nearest pixel interpolation or bilinear interpolation.   13. The medical image signal processing method according to claim 12, wherein the total error evaluation means for obtaining the total error calculates a weighted average of the error at the sample point and the error at the evaluation point. Means for arranging a predetermined number of evaluation points at each sample point of the processing signal of the medical image;
First interpolation value calculation means for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
First error evaluation means for obtaining an error of an interpolation value at each sample point;
Second error evaluation means for obtaining an error of an interpolation value at the arranged evaluation points;
Total error evaluation means for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative means of obtaining each of the total errors,
A medical image signal processing apparatus comprising: an interpolation step that performs interpolation processing of a medical image processing signal using a polynomial of the highest order that minimizes the overall error. Placing a predetermined number of evaluation points at each sample point of the medical image processing signal;
A first interpolation value calculating step for obtaining an interpolation value at each of the sample points and the evaluation points;
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of the interpolation value at the evaluation point arranged as described above;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; A computer program for causing a computer to execute. Placing a predetermined number of evaluation points at each sample point of the medical image processing signal;
A first interpolation value calculation step for obtaining an interpolation value at each of the sample points and the evaluation points using a polynomial having a highest order of 1 to n (n is a natural number of 2 or more);
A first error evaluation step for obtaining an error of an interpolation value at each sample point;
A second error evaluation step for obtaining an error of an interpolation value at the arranged evaluation points;
A total error evaluation step for obtaining a total error of an interpolation formula used for interpolating the processing signal of the medical image based on the error of the interpolation value at each sample point and the error of the interpolation value at the evaluation point; The first interpolation value calculation step, the first error evaluation step, the second interpolation value error evaluation step, and the total error evaluation step are repeated by changing the highest order of the polynomial to correspond to the polynomial in each highest order. An iterative step to obtain each overall error;
A computer program that causes a computer to execute an interpolation step of performing interpolation processing of a processing signal of a medical image using a polynomial of the highest order that minimizes the overall error.

Images