JPH0444545B2 - - Google Patents

Description

Detailed Description of the Invention (Industrial Application Field) The present invention sets a γ function curve that approximates a curve showing contrast agent dynamics in a tomographic image of an X-ray tomography apparatus, and The present invention relates to an image processing device that configures various characteristics of images as calculated images.

(Prior Art) An X-ray tomography apparatus is a device that obtains a tomographic image by irradiating X-rays from around a subject. In this tomography technique, a water-soluble contrast agent is injected intravenously, and as the contrast agent is absorbed sequentially, images of blood vessels, organs, etc. change over time. There are ways to make a diagnosis. 1 with tomographic image
Time-series measurement data of a point or a certain region of interest is said to form a γ function curve as shown in Figure 3 when the horizontal axis is the time axis and the vertical axis is the change in CT number. It is said that setting a γ function curve that most closely approximates the measured data (hereinafter referred to as γ fitting) will best approximate the measured data. Figure 3 shows data at a certain point in the same slice plane, and in the figure, the circles are measured data at each time (each slice) at that point, connected by the most approximate γ function curve 21. It is. This γ function is expressed by equation (1).

f=K(t- _t0 )〓exp{-β(t- _t0 )}...(1) (Problem to be solved by the invention) As a method of gamma fitting, minimize the error by using grid search method etc. There are methods such as finding the parameters asymptotically, or converting the γ function with another function, such as a logarithmic function, and bringing it into the normal equation. The former asymptotic method requires an extremely large amount of data and does not always yield a solution. Furthermore, in the case of using the latter logarithmic function, there are problems in terms of the enormous amount of calculation and accuracy, such as large errors in areas where data values are small, and it is not suitable for practical use. Therefore, it is difficult to perform accurate γ-fitting, and it is even more difficult to obtain a high-quality and easily viewable calculated image created using the γ-function curve.

The present invention has been made in view of the above points, and its purpose is to reduce the amount of calculation during γ-fitting, perform highly accurate γ-fitting in a short time, and create a high-quality calculated image. The objective is to realize an image processing device.

(Means for Solving the Problems) A first invention for solving the above problems sets a γ function curve that approximates a curve showing contrast agent dynamics in a tomographic image of an X-ray tomography apparatus, and In an image processing device that configures various characteristics of the γ function curve as a calculation image, the first data of the temporally changing measurement data of the image accompanied by contrast agent injection (the data immediately before the contrast agent arrives) is used as the origin of the coordinates. a data clipping means that performs coordinate transformation and converts negative data to positive values; logarithmic conversion means for performing the replacement, weighting means for weighting the output data of the logarithm conversion means, disturbance removal means for removing disturbance data to correct the disturbance of the curve due to the disturbance, and disturbance removal means for removing the disturbance data from the disturbance removal means. The second invention is characterized by comprising a γ function determining means for calculating unknown parameters of the γ function based on the removed data and setting a γ function curve to be approximated.
In an image processing device that sets a γ function curve that approximates a curve showing contrast agent dynamics in a tomographic image of a line tomography device, and configures various characteristics of the set γ function curve as a calculation image, contrast agent injection is performed. data clipping means for performing coordinate transformation using the leading data (data immediately before the arrival of the contrast agent) of the measurement data that changes within the time of the accompanying video as the origin of the coordinates, and converting negative data to positive values by the coordinate transformation; logarithmic conversion means for storing a three-dimensional linear equation obtained by logarithmically converting the γ function and replacing the coefficients with data from the data clipping means; weighting means for weighting the output data of the logarithmic conversion means; A disturbance removal means for removing disturbance data in order to correct the disturbance of the curve due to the disturbance removal means, and setting of a γ function curve to approximate by calculating unknown parameters of the γ function based on the data after the disturbance removal from the disturbance removal means. a γ-fitting means comprising a γ-function determining means to perform the operation; and a data control for controlling the operation of the γ-fitting means and creating a calculation image based on a γ-function curve by calculating with a γ-function determining parameter output from the γ-fitting means. means and
The present invention is characterized by comprising an interpolation means for interpolating pixels in a portion that cannot be calculated during the calculation process by the data control means to complete a calculated image.

(Effect) The γ function, which approximates the measured data showing the contrast agent dynamics, is converted into a linear equation by logarithmic transformation to reduce the amount of calculation, improve accuracy, and take measures to remove disturbances.
A function curve is set, and calculation image data is calculated from the γ function curve set for each pixel, and data is interpolated for pixels that cannot be calculated to create a smooth calculation image.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a structural block diagram of an embodiment of the γ-fitting device of the first invention. In the figure,
1 was obtained by slicing with multiple scans,
Convert time-series measurement data of a single point in multiple tomographic images or a certain region of interest into a coordinate system based on the time and CT number of the first data, and at that time, convert data whose CT number is negative to a positive value close to 0. The data clipping device processes the input data as follows. An example of input data is shown in FIG. 4, and processing for this data will be explained. Each point in the figure is a plot of slice data at each time. The number assigned to each data is the data number, and data 1 is the data of CT number CT ₁ of the first slice at time T ₁ , data i
is the CT number of the i-th slice at time Ti
Data for CH ₁ is shown. In order to γ-fit the data shown in FIG. 4, the data clipping device 1 converts the data shown in FIG. 4 into a coordinate system based on the time _T1 and CT number _CT1 of data 1, which is the leading data, that is, the data immediately before the arrival of the contrast agent. The coordinate transformed data is
As shown in the figure. In the graph of Figure 5, data 2 is
Since the increment of the CT number is negative, the data clipping device 1 replaces the negative data with a preset value A close to 0 (A>0). Therefore, the coordinate transformation is performed as shown in equation (2) (T _i , CT _i )→(t _i , M _i ′) ...(2) However, t _i =T _i −T ₁ M _i ′= CT _i − CT ₁ (when CT _i − CT ₁ ≧A) = A (when CT _i − CT ₁ <A) A logarithmic conversion device that sequentially substitutes the output data of the data clipping device 1 into the coefficients of the equation is used to logarithmically transform the equation of the γ function in equation (1) to obtain the three-dimensional linear equation shown in equation (3). It is held in the form of

M=ax+by+cz...(3) However, M=log _e f (hereinafter log _e is simply written as log) x=log K, y=α, z=β a=1 b=log(t-t ₀ ) c =-(t- _t0 ) In the data clipping device 1, each coefficient M _{i ,} b _{i ,}
c _i is given. Therefore, equation (3) becomes equation (4). Apply an _appropriate value to t ₀ until it _is determined later ₍ see Figure 3) However, since it is in logarithmic space, it is a fitting that is greatly affected by a small area of measured data, and the accuracy is poor for approximation in real space. This is a weighting device to make it equivalent to fitting in real space. To determine the γ function, as shown in equation (5), set the residual error when the left side of equation (4) at each measurement point i is not 0 as e _i ', and minimize the sum of its squares. When x, y, and z are found, K, α, and β of the γ function in equation (1) are found, and the γ function can be determined.

_o 〓 ⁱ⁼¹ e _i ′ ² = _o 〓 ⁱ⁼¹ (x+b _i y+c _i z−Mi) ² ...(5) As mentioned above, the evaluation of equation (5) is in logarithmic space, so the precision is In order to improve the accuracy, real space accuracy is improved compared to equation (5) by evaluating equation (5) using a weighting function w _i that is a function of M _i .

_o 〓 ⁱ⁼¹ e _i ² = _o 〓 ⁱ⁼¹ (w _i e _i ′) ² = _o 〓 ⁱ⁼¹ {(x+b _i y+c _i z−M _i )w _i } ² ……(6) Above Although various weighting functions w _i can be considered, one example will be explained with reference to FIG. 6. Now, if we logarithmically transform this data set X in real space,
Consider logX. FIG. 6 is a diagram showing a curve of Y=logX for data x in real space. In the figure, if the coordinates of P are (X ₀ , logX ₀ ), the sensitivity of the change in the ordinate Y to the change in the abscissa X of data P is expressed by the differential coefficient 1/X ₀ at the point P of the curve logX be able to. Therefore, by multiplying the residual in logarithmic space by the reciprocal of the differential coefficient, it becomes possible to perform γ fitting on arbitrary data X without being affected by nonlinearity caused by logarithmic transformation. From the above, in the case of conversion to logarithmic space, w(X) = 1/(1/X) = X
And it is sufficient. In the weighting device 3, w _i of equation (6)
By setting w _i =M _i ′, maintaining this formula, and successively substituting input data t _i and M _i ′, it is possible to obtain γ fitting equivalent to that in real space. Therefore, the equation corrected by the weight function of logarithmic space correction is as shown in equation (7).

_o 〓 ⁱ⁼¹ e _i ² = _o 〓 ⁱ⁼¹ {(x+b _i y+c _i z−M _i )M _i ′} ² ...(7) 4 indicates that the weighted data of each slice from the weighting device 3 is This is a disturbance removal device that eliminates data errors caused by disturbances such as having a secondary peak at the falling edge of the curve as shown in FIG. 7 or slowing down the falling edge of the curve due to the influence of blood recirculation. In the figure, (1, 2, 3, . . . m . . . p . . . n) are measured data, m+1 is data with a slow fall, and data p is data showing a secondary peak. The curve connecting the measured data is the curve of the γ function to be determined, and data m+1, data p, etc. are data found to not represent the correct value of the original data due to disturbance. The disturbance removal device 4 is preset with a 100-minute ratio RP with respect to the peak value PV appearing in the measurement data (see FIG. 7). The measurement data m of the first slice whose measurement data falls below (RP/100)×PV with respect to the peak value PV is found, and this measurement data m is used as the final data to be used for γ fitting. Therefore, the disturbance removal device 4 converts equation (7) into equation (8).

E= _n 〓 ⁱ⁼¹ {(x+biy+c _i z−M _i )M _i ′} ² ...(8) However, m: Last data used for γ fitting E: Sum of squares of residuals to be evaluated 5 is the above The value of the variable that minimizes the sum of squares E of the residual errors in equation (8) obtained by the disturbance removal device 4 is determined, and the following calculation is performed by the γ function determination device that determines the γ function. That is, from equation (8), a normal equation with the partial differential set to 0 is established as shown in the following equation.

∂／∂xE=0 ∂／∂yE=0 ∂／∂zE=0 This simultaneous equation is written as equation (9).

〓〓=〓 ...(9) Here, in order to solve equation (9), another parameter t ₀ (see equation (3)) other than x, y, and z must be determined. Therefore, in the device of this embodiment,
Several to several dozen values selected as described later are expressed in equation (8).
Substituting b _i and c _i into t ₀ and solving equation (9) independently,
As a result, the γ function is determined by K, α, β, and t ₀ that minimize the square sum E of all the obtained γ-fitting residuals. FIG. 8 shows how to select the several to several tens of ₀ values mentioned above. 100 minute ratio AP to peak value PV that determines the fitting start time point in advance
and the number of fitting repetitions N. Compare the measurement data for each slice (circle mark in the figure) with the value of (AP/100) x PV, and calculate (AP/100)
Find the first slice that exceeds ×PV, and let T be the time at which it occurs. The value chosen as in the following equation is t ₀
γ fitting is performed N times.

t ₀ =ΔT×j However, ΔT=T/N, j=0, 1, 2,...N-1 Determine as a function to obtain the γ-fitting when E becomes the minimum as a result of N-times γ-fitting .

Assume that t ₀ determined as described above is the position shown in FIG. If the slice immediately after t ₀ is 1, the slices before 1-1 will not be used. However, although it is not used for γ fitting, it is used for evaluating the sum of squares E of the residuals as shown in equation (10), so that a correct evaluation can be made for any _t0 .

E= _l-1 〓 ⁱ⁼¹ M _i ² + _n 〓 ^i=l {(x+b _i y+c _i z−M _i ) M _i ′} ² ...(10) In equation (10), the first term on the right side is The second term, which is the measurement data of the rice not used for γ fitting, is the slice used for γ fitting, and the slices l to m are reflected in equation (9). The γ function determining device 5 solves the normal equation (9) for N times t ₀ as described above, and has K, α, β, and t ₀ when the sum of squares E of the residuals is minimum. It is determined as the γ function obtained by

The operation of the γ fitting device configured as described above will be explained. The data group obtained by slices from multiple scans detected by the X-ray detector is converted by the data clipping device 1 into a coordinate system based on the time and CT number of the first data, and the negative data is If there is data that becomes 0, it is replaced with a predetermined positive value A close to 0. The coordinates (t _i , M _i ') of this output data group are input to the logarithmic conversion device 2 . In the logarithmic transformation device 2, the coefficients of the equation of the γ function, which has been logarithmically transformed into a three-dimensional linear equation, are replaced by M _i , b _i , and c _i to create a group of equations equal to the number of slices.

The equation groups formed as described above are each multiplied by a weighting function w _i in the weighting device 3 in order to make the accuracy in logarithmic space equivalent to the accuracy in real space. The equation group weighted by the weighting device 3 is as shown in equation (7). The output signal of the weighting device 3 is input to a disturbance removing device 4, and the effects of slowing of the fall and secondary peaks are removed from the curve formed by each measurement data, and the signal is input to the γ function determining device 5. The γ function determining device 5 calculates K, α, and β represented by x, y, and z in equation (4), which is a three-dimensional linear equation.
The γ function to be obtained is determined by performing an operation to determine other parameters t ₀ and obtaining the unknowns K, α, β, and t ₀ in the formula of the γ function.

As explained above, the embodiment of the first invention has the following advantages.

(1) By performing least squares approximation by logarithmically transforming the measured quantity and solving a normal equation, the amount of calculation can be significantly reduced and the calculation time can be significantly reduced.

(2) By performing weighting in logarithmic space, it is possible to perform γ fitting equivalent to γ fitting in real space, and the accuracy of γ fitting is improved.

(3) Accurate γ-fitting using original data becomes possible by realizing disturbance removal.

(4) The optimal γ-fitting start time can be automatically determined by the method for determining t ₀ and the evaluation using equation (10).

Note that the present invention is not limited to the above embodiments. In the embodiment, data such as time and CT number increment have been described, but any time series data can be approximated by the γ function. Furthermore, although w(X)=X is used as an example of the weighting function, it is also possible to use other functions.

Next, FIG. 2 is a configuration block diagram of an embodiment of the computational image creation device of the second invention.

In the figure, 11 denotes parameters K, α,
the aforementioned γ fitting device outputting β, t ₀ ;
12 controls the γ-fitting device 11 to perform calculations to obtain a γ-function fitted to the measurement data of each pixel in order to construct a calculated image based on the input scan image data, and as a result, the γ-fitting device 11 The _γ function curve of the data of each pixel obtained from ,
This is a data control device that reproduces a γ function curve using the parameters K, α, β, and t ₀ described above and calculates various calculation image data such as a peak value and an area under the curve, which will be described later. In the calculation image calculation process described above, if a pixel occurs that causes an overflow due to the appearance of an excessive numerical value, the calculation is interrupted and a calculation interruption marking is stored in place of the data of that pixel. 13, when a calculation interruption marking is found in the input data from the data control device 12, the input data is searched again across the N×N matrix centered on that pixel, and the N×N
When there are M or more valid data in the matrix, the blank data is filled by replacing the marking of the calculation interrupted pixel with the average value of those valid data,
This is an interpolation device that replenishes and completes the calculated image formed by the data control device 12 when all pixel data is available. The calculated image output from the interpolator 13 is displayed on a display device (not shown). Here, the numerical values of N and M shall be determined as appropriate depending on the observation location, the degree of the lesion, etc.

The operation of the embodiment configured as described above will be explained. The data control device 12 controls γ for each pixel.
In order to obtain fitting data, the γ fitting device 11 is instructed to perform γ fitting of a desired pixel, and based on the obtained γ fitting, one of the various calculation images specified in advance, such as those shown below, is Calculate.

(1) Peak value of the γ function curve (CT number) (2) Time from the origin of the coordinate axis to the peak value (3) Area occupied by the curve (4) Width of the inflection point (each change in the rise and fall of the curve) time width between curved points) (5) Appearance time (time from the origin until the CT number of the measurement data reaches 15% of the peak value on the curve) (6) Average passing time (passing from the origin to that pixel point) (7) Corrected average transit time (time obtained by subtracting appearance time from average transit time) (8) Relative flow rate (reciprocal of corrected average transit time) Pixels where overflow occurred during the calculation process of the above calculation image If so, the calculation is interrupted and a marking is made to that effect. When the interpolation device 13 finds the marking, it interpolates by calculating the average value of M or more valid data among the N×N data in the vicinity.
The unfinished calculation image calculation including the interpolated data is completed and displayed on the display device.

As explained above, according to this embodiment, by introducing an interpolation device when creating a calculation image, it is possible to compensate for pixel values that cannot be calculated due to overflow etc. by smoothly interpolating from the surrounding area, thereby improving image quality. A good, easy-to-see plan image can be obtained.

(Effects of the Invention) As explained in detail above, according to the present invention,
The amount of calculation during γ-fitting is significantly reduced, highly accurate γ-fitting can be obtained in a short time, and an image processing device that can obtain calculated images with high precision and quality has been realized, which has practical effects. is big.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the construction of an embodiment of the γ-fitting device of the first invention, FIG. 2 is a block diagram of the construction of an embodiment of the computational image creation device of the second invention,
Figure 3 is a diagram of the γ function curve that approximates the measured data,
Figure 4 is a diagram showing an example of input measurement data, Figure 5 is a diagram of coordinate transformation by the data clipping device, and Figure 6 is a diagram showing an example of input measurement data.
The figure is a relationship curve diagram between real space and logarithmic space, FIG. 7 is a diagram of data including disturbance, and FIG. 8 is an explanatory diagram of t ₀ determination. 1... Data clipping device, 2... Logarithmic conversion device, 3... Weighting device, 4... Disturbance removal device,
5... γ function determining device, 11... γ fitting device, 12... data control device, 13
...Interpolator, 21...γ function curve.

Claims (1)

[Scope of Claims] 1. An image in which a γ function curve that approximates a curve showing contrast agent dynamics in a tomographic image of an X-ray tomography device is set, and various characteristics of the set γ function curve are configured as a calculation image. In the processing device, a coordinate transformation is performed using the first data of the temporally changing measurement data of the image with contrast agent injection (data immediately before the contrast agent arrives) as the origin of the coordinates, and data that becomes negative due to the coordinate transformation is converted to a positive value. a data clipping means for migrating to
logarithmic conversion means for storing a three-dimensional linear equation obtained by logarithmically converting the γ function and replacing coefficients with data from the data clipping means; weighting means for weighting the output data of the logarithm conversion means;
Disturbance removal means for removing disturbance data in order to correct the disturbance of the curve due to disturbance, and setting of a γ function curve for calculating and approximating unknown parameters of the γ function based on the data after the disturbance removal from the disturbance removal means. 1. An image processing apparatus comprising: a γ fitting means comprising a γ function determining means for performing the following steps. 2. In an image processing device that sets a γ function curve that approximates a curve showing contrast agent dynamics in a tomographic image of an X-ray tomography device, and configures various characteristics of the set γ function curve as a calculation image, the contrast agent Data clipping means that performs coordinate transformation using the leading data of the temporally changing measurement data of the image accompanying the injection (data immediately before the arrival of the contrast agent) as the origin of the coordinates, and converts negative data to positive values by the coordinate transformation. and,
Logarithmic conversion means stores a three-dimensional linear equation obtained by logarithmically converting the γ function and replaces coefficients with data from the data clipping means; weighting means weights the output data of the logarithmic conversion means; A disturbance removal means for removing disturbance data in order to correct the disturbance of the curve, and a γ function curve to be approximated by calculating the unknown parameters of the γ function based on the data after the disturbance removal from the disturbance removal means. γ-fitting means comprising a γ-function determining means; and data for controlling the operation of the γ-fitting means and calculating using a γ-function determining parameter output from the γ-fitting means to create a calculation image based on a γ-function curve. An image processing apparatus comprising: a control means; and an interpolation means for completing a calculated image by interpolating pixels in a portion that cannot be calculated during the calculation process by the data control means.