JP4797876B2 - X-ray image evaluation method and apparatus - Google Patents

Description

  The present invention relates to an X-ray image evaluation method and apparatus for evaluating an X-ray image based on an X-ray image obtained at each step, and more particularly to a technique for evaluating the graininess of an X-ray image.

  When an X-ray image is printed out or displayed on a screen, the X-ray image may be output with graininess. This is called “granularity”. Factors that deteriorate the granularity of X-ray images include quantum noise due to X-ray quanta, electrical noise due to electricity, and structural noise due to the structure of an X-ray detector, etc., but quantum noise is the most dominant. In addition, as electrical noise, it originates in an electrical circuit etc., for example. In addition, as the structural noise, for example, there are two-dimensional patterns (patterns such as vertical stripes and horizontal stripes) appearing on the image, which are specific patterns such as an X-ray grid pattern arranged on the incident surface of the X-ray detector. to cause. An index representing the granularity due to quantum noise includes X-ray capture efficiency (DQE), which is the efficiency of capturing photons converted from X-rays (hereinafter abbreviated as “X-ray photons”).

  However, in practice, it is difficult to directly measure the X-ray capture efficiency DQE at each medical facility. Further, the normal X-ray capture efficiency DQE is measured by the imaging dose, and when the X-ray image is a fluoroscopic image obtained by fluoroscopic imaging, the measurement of the X-ray capture efficiency DQE under the fluoroscopic condition is a quantum Noise other than noise will be included. Therefore, in the case of a fluoroscopic image, it is impossible to measure only the original X-ray capture efficiency DQE.

  The granularity of the fluoroscopic image greatly depends on the incident dose (panel incident dose) to the detection surface (panel surface) of the X-ray detector and the radiation quality. There is a conflict between the reduction in exposure and the granularity of the fluoroscopic image. Therefore, in order to reduce the exposure, it is necessary to know what the noise configuration of the fluoroscopic image at that time is. In order to improve the granularity of a fluoroscopic image, there is a need for a noise measurement method that can determine whether X-ray capture efficiency DQE should be improved or other noise such as electrical noise should be reduced. For this purpose, it is necessary to obtain each noise separately, in particular to obtain quantum noise, and to obtain the X-ray capture efficiency DQE that is an index thereof. Further, if the X-ray capture efficiency DQE is not known, there is a risk of shipping an imaging apparatus that captures X-ray images with poor graininess. Thus, it is required to know the X-ray capture efficiency DQE.

  The present invention has been made in view of such circumstances, and an object thereof is to provide an X-ray evaluation method and apparatus for easily evaluating an X-ray image by obtaining X-ray capture efficiency. To do.

In order to achieve such an object, the present invention has the following configuration.
That is, the invention described in claim 1 is a method for evaluating an X-ray image based on an X-ray image obtained at each step, and the quantum noise caused by the X-ray quantum in the i-th step is expressed as Nq (i ), And the X-ray image signal in the i-th step is S (i), the electrical noise due to electricity in the i-th step is Ne (i), and the total noise in the i-th step is Nt (i). Assuming that the physical quantity in the i-th step is Vq (i), the physical quantity Vq (i) is
Vq (i) = {Nq (i)} ² / S (i)
= [{Nt (i)} ² − {Ne (i)} ² ] / S (i)
The physical quantity Vq (i) is obtained by the following formula, and, based on the physical quantity Vq (i), by obtaining an X-ray capture efficiency that is an efficiency of capturing an X-ray photon that is a photon converted from an X-ray, If the value of the X-ray capture efficiency is small, the graininess is poor, and if the value of the X-ray capture efficiency is large, the graininess is good, and the X-ray image is evaluated with respect to the graininess .

[Operation and Effect] According to the first aspect of the present invention, Vq (i) = {Nq (i)} ² / S (i) = [{Nt (i)} ² − {Ne (i)} ² ] / S (i) is used to determine the physical quantity Vq (i) at the i-th step. Nq (i) is quantum noise in the i-th step, S (i) is an X-ray image signal in the i-th step, Ne (i) is electrical noise in the i-th step, and Nt (i) is i Total noise in the second step. As Vq (i) = {Nq (i)} ² / S (i), the physical quantity Vq (i) is an index of the quantum noise Ng (i), and {Nq (i)} ² / S (i) = [{Nt (i)} ² − {Ne (i)} ² ] / S (i) It is possible to separate each noise. Therefore, the physical quantity Vq (i) can be obtained by separating each noise. Since this physical quantity Vq (i) is also an index of the X-ray capture efficiency, the X-ray image can be easily evaluated by obtaining the X-ray capture efficiency based on the physical quantity Vq (i).

  Specifically, the signal S (i) and the total noise Nt (i) are obtained based on the X-ray image. Then, assuming that the electric noise Ne (i) is constant regardless of each step, the electric noise Ne (i) is set so that the variation with respect to the step of the physical quantity Vq (i) obtained by the above-described equation becomes a minimum value. Obtained (the invention according to claim 2). In order to obtain the electric noise Ne (i) so that the variation with respect to the step of the physical quantity Vq (i) becomes the minimum value, the electric noise Ne (i) is set to various values, and the set electric noise Ne ( For each setting, the variation of the physical quantity Vq (i) obtained by substituting i) into the above-described equation is obtained. The electric noise Ne (i) set so that the variation becomes the minimum value is the electric noise value to be obtained. As a representative method, there is a least square method. In the case of the least square method, the variation corresponds to a mean square error.

Further, the physical quantity Vq (i) is obtained based on the obtained signal S (i), the total noise Nt (i), and the electric noise Ne (i), and based on the average value of the physical quantity Vq (i) in each step. The X-ray capture efficiency described above is obtained (the invention according to claim 3). The average value of the physical quantity Vq (i) in each step is the Vq value finally obtained. Further, the quantum noise Nq (i) is obtained based on the obtained total noise Nt (i) and electrical noise Ne (i) (the invention according to claim 4). If the total noise Nt (i) and the electrical noise Ne (i) are obtained, an expression {Nq (i)} ² = {Nt (i)} ² − {Ne (i) obtained by dividing S (i) by the above-described expression. )} ² , the quantum noise Nq (i) can be obtained.

The invention described in claim 5 is an apparatus that evaluates an X-ray image based on an X-ray image obtained at each step, and obtains an X-ray image by performing signal processing based on the X-ray. Signal processing means, the signal processing means,
The quantum noise due to X-ray quanta in the i-th step is defined as Nq (i), the X-ray image signal in the i-th step is defined as S (i), and the electrical noise due to electricity in the i-th step is defined as Ne (i ), When the total noise in the i-th step is Nt (i) and the physical quantity in the i-th step is Vq (i), the physical quantity Vq (i) is
Vq (i) = {Nq (i)} ² / S (i)
= [{Nt (i)} ² − {Ne (i)} ² ] / S (i)
The physical quantity Vq (i) is obtained by the following formula, and signal processing for obtaining the X-ray capture efficiency, which is the efficiency of capturing the X-ray photons that are photons converted from the X-rays, based on the physical quantity Vq (i). If the X-ray capture efficiency value is small, the graininess is poor, and if the X-ray capture efficiency value is large, the graininess is good, and the X-ray image is evaluated with respect to the graininess .

  [Operation and Effect] According to the invention described in claim 5, the invention described in claim 1 can be suitably implemented by providing the signal processing means.

According to the X-ray image evaluation method and apparatus according to the present invention, Vq (i) = {Nq (i)} ² / S (i) = [{Nt (i)} ² − {Ne (i)} ² ] / S (i), the physical quantity Vq (i) at the i-th step can be obtained to separate the noises to obtain the physical quantity Vq (i). Since this physical quantity Vq (i) is also an index of the X-ray capture efficiency, the X-ray image can be easily evaluated by obtaining the X-ray capture efficiency based on the physical quantity Vq (i).

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram of an X-ray image evaluation apparatus according to an embodiment, and FIG. 2 is a schematic diagram illustrating a fluoroscopic image and a measurement location of a measurement phantom used in the embodiment. In the present embodiment, the configuration other than the signal processing unit described later in the X-ray image evaluation apparatus (for example, an X-ray tube or a flat panel X-ray detector) is used for the X-ray fluoroscopic apparatus after shipment. The following will be described.

  As shown in FIG. 1, an X-ray evaluation apparatus according to the present embodiment includes an X-ray tube 1 that irradiates X-rays to a phantom PH, and a flat panel X-ray detector 2 that detects X-rays transmitted through the phantom PH. (Hereinafter referred to as “FPD”) and a signal processing unit 3 that performs signal processing based on the detected X-rays and evaluates an X-ray image (here, a fluoroscopic image). The signal processing unit 3 corresponds to the signal processing means in this invention.

  The phantom PH is for measurement, and may have a stepped shape (stepped shape) shown in FIG. 1 or a shape that is isotropic as shown in FIG. In the X-ray image evaluation processing shown in FIGS. 4 to 6 described later, the isotropic phantom PH shown in FIG. 2 is used.

The perspective image of the isotropic phantom PH shown in FIG. 2 has nine regions of interest ROI. As shown in FIG. 2, it is assumed that ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ . Then, ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ are divided into respective steps in this order, and an X-ray image signal is detected by FPD 2. It is assumed that the thickness of the phantom PH increases in the order of ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ . As the phantom PH for the fluoroscopic image shown in FIG. 2, for example, a copper plate is used. In ROI ₁ , the thickness of the phantom PH is about 0.5 mm, and in ROI ₉ , the thickness is about 4 mm.

The signal processing unit 3 includes a first image memory unit 31, an integration unit 32, a subtraction unit 33, an ROI setting unit 34, a second image memory unit 35, a third image memory unit 36, and a calculation unit 37. The first image memory unit 31 writes and stores X-ray image signals for a plurality of frames (for example, about 30 frames) detected by the FPD 2. That is, the first image memory unit 31 is a memory for continuous images. The integration unit 32 reads out X-ray image signals for a plurality of frames stored in the first image memory unit 31 and adds and averages the X-ray images of each frame to obtain an integrated image. That is, the integrating unit 32 obtains an integrated image composed of the structural noise for removing the structural noise.

The subtracting unit 33 reads out a representative X-ray image signal for one frame among the X-ray images for a plurality of frames stored in the first image memory unit 31, and outputs the read X-ray image. An integrated image obtained by the integrating unit 32 is subtracted from the signal to obtain a noise image composed of quantum noise and electrical noise. That is, the subtraction unit 33 subtracts the integral image made up of the structural noise, thereby removing the structural noise and extracting only the quantum noise and the electrical noise. The ROI setting unit 34 sets ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ shown in FIG. 2 described above.

The second image memory unit 35 is obtained by the subtraction unit 33, ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ set by the ROI setting unit 34. Based on the noise image, the standard deviation of the noise image for each ROI in each step is written and stored. Since the standard deviation of the noise image is also total noise, the second image memory unit 35 is a memory for total noise. The third image memory unit 36 is ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ set by the ROI setting unit 34, and the representative 1 described above. Based on the X-ray image signal for the frame, the average value of the X-ray image signal for each ROI in each step is written and stored.

  The calculation unit 37 reads the average value of the signals of the X-ray image stored in the third image memory unit 36 and also reads the total noise stored in the second image memory unit 35 and calculates the average value of these signals (the average value thereof) ) And the physical quantity based on the total noise. At this time, quantum noise is also obtained.

Steps correspond to ROIs, ROI ₁ is the first step, ROI ₂ is the second step, ROI ₃ is the third step, ROI ₄ is the fourth step, and ROI ₅ is the fifth step It is assumed that ROI ₆ is the sixth step, ROI ₇ is the seventh step, ROI ₈ is the eighth step, and ROI ₉ is the ninth step. Here, the region of interest in the i-th step is ROI _i , and i satisfies (1 ≦ i ≦ 9).

  The quantum noise in the i-th step is Nq (i), the X-ray image signal (average value) in the i-th step is S (i), the electrical noise in the i-th step is Ne (i), The total noise in the i-th step is Nt (i). Further, the physical quantity in the i-th step is assumed to be Vq (i).

[Definition of Vq]
Next, a general definition of the physical quantity Vq (i) will be described with reference to FIG. FIG. 3A shows an image acquisition model in FPD, and FIG. 3B shows a signal generation process for each image. Note that “(i)” representing the i-th step is omitted, the signal is S, and the physical quantity is Vq. The flat panel X-ray detector (FPD) 2 (see FIG. 1) includes an FPD sensor unit 21 and an FPD sensor control unit 22 as shown in FIG.

  The FPD sensor unit 21 captures X-ray photons that are photons converted from X-rays, and converts the captured X-ray photons into electrons (electrical signals). The efficiency for capturing X-ray photons at this time is called “X-ray capture efficiency”, and the efficiency for converting X-ray photons into electrons is called “photon-electron conversion efficiency”.

  The FPD sensor control unit 22 performs sensitivity correction so that a constant image signal is obtained with respect to a constant X-ray dose, and performs digital gain according to the application. Assuming that the above-mentioned X-ray capture efficiency is DQE, the photon-electron conversion efficiency is f, and the digital gain in the FPD sensor control unit 22 is g, signals and noise from X-ray incidence to image signal output are as follows: (See FIG. 3B).

  From the theory of photon statistics, it is known that when S ′ (X-ray) photons are incident as shown in FIG. 3B, the fluctuation (noise) becomes √ (S ′). . By multiplying by the X-ray capture efficiency DQE, the signal becomes DQE × S ′, and the fluctuation (noise) becomes √ (DQE × S ′).

  The captured X-ray photons are converted into electrons by the photon electron conversion efficiency f. At that time, the signal is DQE × S ′ × f, and the fluctuation (noise) is {√ (DQE × S ′)} × f. And output from the panel surface (detection surface) of the FPD 2.

  The signal and fluctuation (noise) output in this way are corrected for sensitivity by the amount of deterioration due to the X-ray capture efficiency DQE. In this correction, each is normalized by 1 / (DQE × f) by dividing by (DQE × f), so that the signal returns to S ′. On the other hand, the fluctuation (noise) is {√ (DQE × S ′)} × f / (DQE × f) = √ (S ′ / DQE).

  The FPD sensor control unit 22 adjusts the gain so as to output a constant (image signal) output value for a constant (incident) X-ray dose. When the final gain adjustment is g, the final signal S is S ′ × g (S = S ′ × g (1)), fluctuation (noise), that is, the quantum noise Nq is {√ (S ′ / DQE) } × g (Nq = {√ (S ′ / DQE)} × g (2)).

The physical quantity Vq is defined as Nq ² / S (Vq = Nq ² / S (3)). By substituting the above equations (1) and (2) into equation (3), the physical quantity Vq becomes g / DQE (Vq = g / DQE (4)), the physical quantity Vq is proportional to the digital gain g, It becomes inversely proportional to the X-ray capture efficiency DQE. As described above, the physical quantity Vq defined in this way is an index representing the X-ray capture efficiency DQE.

[Measurement method of Vq]
Next, X-ray image evaluation processing including measurement of the physical quantity Vq (i) will be described with reference to FIGS. FIG. 4 is a flowchart showing a series of X-ray image evaluation processes, FIG. 5 is a table listing each signal, noise, and root mean square error at each step, and FIG. 6 is an electric diagram by the least square method. It is a graph when calculating | requiring noise Ne (i). As described above, the isotropic phantom PH shown in FIG. 2 is used.

(Step S1) Perspective image of phantom X-ray tube 1 (see FIG. 1) emits X-rays to phantom PH, and X-rays transmitted through phantom PH are converted into flat panel X-ray detector 2 (see FIG. 1). ) Is detected. The fluoroscopic image at this time is as shown in FIG. This fluoroscopic image is an X-ray image signal for one frame, and is written and stored in the first image memory unit 31 (see FIG. 1).

(Step S2) Reached 30 frames?
It is determined whether or not 30 frames have been reached. If it has not reached 30 frames, the process returns to step S1 and X-ray image signals for 30 frames are continuously collected. If it has reached 30 frames, it is determined that X-ray image signals for 30 frames have been acquired, and the process proceeds to the next step S3. The signals of the X-ray image for one frame stored in the first image memory unit 31 are set as S ₁ , S ₂ ,..., S _F ,..., S ₂₉ , S _{30 in} order from 1 frame (1 ≦ F ≦ 30).

(Step S3) X-ray image addition average integration unit 32 (see FIG. 1) of each frame X-ray image signals S ₁ , S ₂ ,... For 30 frames stored in the first image memory unit 31. , S _F ,..., S ₂₉ , S ₃₀ are read out, and the X-ray image of each frame is added and averaged to obtain an integrated image. When this integrated image is _SADD , the integrated image _SADD is expressed by the following equation (5).

S _ADD = (1/30) ΣS _F (5)
However, ΣS _F is the sum of S ₁ , S ₂ ,..., S _F ,..., S ₂₉ , S ₃₀ . Thus, if averaged an X-ray image of each frame is a continuous image, the integral image S _ADD obtained by averaging random noise reduction Yowashi becomes the main structure and structure noise-only image.

(Step S4) Subtraction of integral image from X-ray image The subtraction unit 33 (see FIG. 1) stores signals S ₁ , S ₂ ,..., 30 frames of X-ray images stored in the first image memory unit 31. Of S _F ,..., S ₂₉ , S ₃₀ , a representative X-ray image signal for one frame is read out. Here, you select the S ₁ as a signal for one frame of the X-ray image as a representative. Note that the signal of one frame of the X-ray image as a representative is not limited to _{S 1, S 1, S 2} , ..., S F, ..., it may be arbitrarily selected from S _{29, S 30.} The read X-ray integral image S _ADD obtained by the integration unit 32 from signals S ₁ of the image by subtracting, seek noise image. _Assuming that this noise image is S _NOISE , the noise image S _NOISE is expressed by the following equation (6).

S _NOISE = S ₁ -S _ADD (6)
By subtracting the integral image S _ADD composed of structural noise, the structural noise is removed and only the quantum noise and electrical noise are extracted, and the noise image S _NOISE obtained by the subtraction is composed of quantum noise and electrical noise. Become an image.

(Step S5) Acquisition of Average Value of X-Ray Image Signal The third image memory unit 36 (see FIG. 1) is set to ROI ₁ , ROI ₂ , ROI ₃ set by the ROI setting unit 34 (see FIG. 1). , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ , ROI ₈ , ROI ₉ and the X-ray image signal S _{1 of} one representative frame as described above, the X-ray image for each ROI in each step S (i), which is the signal (average value), is written and stored. That is, an average value in each pixel included in each region of interest ROI is obtained for each ROI, and the average value is used as an X-ray image signal S (i) for each ROI.

Therefore, the X-ray image signal at ROI ₁ becomes the X-ray image signal S (1) at the first step, and the X-ray image signal at ROI ₂ becomes the X-ray image signal at the second step. S (2), the X-ray image signal in ROI ₃ becomes the X-ray image signal S (3) in the third step, and the X-ray image signal in ROI ₄ becomes the X-ray signal in the fourth step. The X-ray image signal at ROI ₅ becomes the X-ray image signal S (5) at the fifth step, and the X-ray image signal at ROI ₆ becomes the _sixth image signal S (4). The X-ray image signal S (6) at the ROI _{7 is obtained} , and the X-ray image signal at the ROI ₇ becomes the X-ray image signal S (7) at the seventh step, and the X-ray image signal at the ROI ₈ is obtained. Is the X-ray image in the 8th step Signal S (8) next to the signal of the X-ray image in the ROI ₉ is a signal S (9) of the X-ray images in the ninth step. That is, an X-ray image signal at ROI _i is an X-ray image signal S (i) in the i-th step.

(Step S6) Acquisition of Total Noise The second image memory unit 35 (see FIG. 1) is set to ROI ₁ , ROI ₂ , ROI ₃ , ROI ₄ , ROI ₅ , ROI ₆ , ROI ₇ set by the ROI setting unit 34. , ROI ₈ , ROI ₉ and the noise image S _NOISE obtained by the subtractor 33, the standard deviation of the noise image for each ROI in each step is written and stored. The standard deviation of this noise image is also the total noise. The total noise Nt (i) in ROI _i is expressed by the following equation (7).

Nt (i) =
√ [Σ [S _NOISE −Σ {S _NOISE / n}] ² / (n−1)] (7)
The equation (7) is also an equation representing the standard deviation (square root of variance) as described above. Note that Σ {S _NOISE / n} is an arithmetic average of S _NOISE in each pixel included in ROI _i .

Therefore, the total noise at ROI ₁ is the total noise Nt (1) at the first step, and the total noise at ROI ₂ is the total noise Nt (2) at the second step, and the total noise at ROI ₃ Is the total noise Nt (3) at the third step, the total noise at ROI ₄ is the total noise Nt (4) at the fourth step, and the total noise at ROI ₅ is the total noise at the fifth step. The total noise at ROI ₆ becomes the total noise Nt (6) at the sixth step, and the total noise at ROI ₇ becomes the total noise Nt (7) at the seventh step. Overall noise at _8, total noise Nt (8) next to the eighth step, integrated noise in ROI ₉ is The total noise Nt (9) in the ninth step. That is, the total noise at ROI _i is the total noise Nt (i) at the i-th step.

(Step S7) Electric Noise Separation / Physical Quantity Calculation The calculation unit 37 (see FIG. 1) reads the X-ray image signal S (i) stored in the third image memory unit 36, and also uses the second image memory. The total noise Nt (i) stored in the unit 35 is read out, and the physical quantity Vq (i) is obtained based on these signals S (i) and the total noise Nt (i). In order to obtain the physical quantity Vq (i), the electric noise Ne (i) is obtained by separating each noise from the total noise Nt (i) based on the noise image S _NOISE extracted only to the quantum noise and the electric noise. At this time, the separated quantum noise Nq (i) is also obtained.

Specifically, considering the relationship between the total noise Nt (i) and the quantum noise / electrical noise, the power (dispersion) of the random noise is equal between the total noise and the quantum noise / electrical noise. That is, the square of the total noise Nt (i), which is the standard deviation, is the variance of the noise image S _NOISE , and the variance of the noise image S _NOISE is the sum of the variance of the image relating to quantum noise and the variance of the image relating to electrical noise. Become. Therefore, the square of the total noise Nt (i) is the sum of the square of the quantum noise Nq (i) and the square of the electric noise Ne (i), and is expressed as the following equation (8).

{Nt (i)} ² = {Nq (i)} ² + {Ne (i)} ² (8)
On the other hand, since the physical quantity Vq is defined by Nq ² / S as in the above expression (3), the physical quantity Vq (i) is expressed as in the following expression (9) using the above expression (8).

Vq (i) = {Nq (i)} ² / S (i)
= [{Nt (i)} ² − {Ne (i)} ² ] / S (i) (9)
That is, the physical quantity Vq (i) is subject to fluctuations in X-ray photons, that is, quantum noise Nq (i). When the X-ray image is a fluoroscopic image as described above, electrical noise is included in the fluoroscopic image. Ne (i) and structural noise are also included. Therefore, it is necessary to separate these noises. Regarding the structural noise, the structural noise is removed in step S4 described above. Therefore, if the electrical noise Ne (i) is obtained and separated, the remaining quantum noise Nq (i) can be obtained, and the physical quantity Vq (i) can also be obtained. Although the fluoroscopic dose differs for each ROI, it is considered that Vq (i) in each step becomes equal if the electrical noise Ne (i) and the structural noise are separated and removed. That is, the physical quantity Vq depends only on the X-ray capture efficiency VQE of the FPD 2 and the digital gain g of the system regardless of the magnitude of the signal S (i) of the X-ray image.

Specifically, in the field of radiation, the measurement values obtained from radiation measurements are handled according to the Poisson distribution (“Radiation Technology for Medical Care (Vol. 1) Nanedo Tatsuhiro Iida, supervised by Inaya Kiyoya P386”). The average value m of Poisson distribution and the standard deviation σ have the following relationship. That is, σ = √m. Now, that the physical quantity Vq is, according to the above (9) Vq in formulas ^{(i) = {Nq (i} )} 2 / S (i), Vq (i) = {Nq (i)} 2 / S ( i) = σ ² / m = a constant value. From this, it can be said that the physical quantity Vq depends only on VQE and g irrespective of the magnitude of the signal S (i) of the X-ray image. This indicates that the physical quantity Vq does not change even if the dose is changed in the graph of FIG. 7 described later. Therefore, the physical quantity Vq can be measured without precisely defining the incident dose.

For this purpose, the electric noise Ne (i) to be separated is obtained by the least square method in this embodiment. Assuming that the electric noise Ne (i) is constant regardless of each step, the electric noise Ne (i) is set so that the variation with respect to the step of the physical quantity Vq (i) obtained by the above equation (9) becomes the minimum value. Ask. In the case of the least square method as in the present embodiment, the variation is a mean square error. Therefore, the signal S (i) and the total noise Nt (i) obtained at each step are arranged as shown in FIG. 5, and the signal S (i) and the total noise Nt (i) at each step correspond to each other. Let At this time, since the electrical noise Ne (i) is constant regardless of each step, for example, the electrical noise Ne (i) is set to each value of Ne ₁ , Ne ₂ ,..., Ne _J ,. The variation of the physical quantity Vq (i) obtained by substituting the set electrical noise Ne (i) into the above equation (9) is obtained for each setting.

FIG. 5 is a list when the electric noise Ne (i) is set to a value of Ne _J , for example. Since the set Ne _J is assumed to be constant regardless of each step, the electric noise Ne (i) is always set to the value of Ne _J at each step. When the set value of Ne _J is substituted into the above equation (9), the obtained physical quantity Vq (i) is as shown in FIG. In the present embodiment, since the mean square error of the physical quantity Vq (i) is a variation with respect to the step of the physical quantity Vq (i), Σ [Vq (i) −Σ {Vq (i) / 9}] which is the mean square error. ^The electrical noise Ne (i) may be set so that ² becomes the minimum value. However, ΣVq (i) is the sum of Vq (1), Vq (2),..., Vq (9), and Σ {Vq (i) / 9} is an arithmetic mean of Vq (i).

Therefore, as shown in FIG. 6, the horizontal axis represents the electric noise Ne (i), and the vertical axis represents Σ [Vq (i) −Σ {Vq (1) / 9}] ² that is the mean square error. The electric noise Ne (i) is set to each of Ne ₁ , Ne ₂ ,..., Ne _J ,..., And the value (in FIG. 6, the square mean error becomes a minimum value (“Min” in FIG. 6)). Find "Ne _Min "). By finding such a value, that value becomes the value of electrical noise to be obtained.

  Substituting the electrical noise Ne (i) obtained in this way into the above equation (9), and substituting the previously obtained signal S (i) and the total noise Nt (i) into the above equation (9). Thus, the physical quantity Vq (i) can be obtained. At this time, the remaining quantum noise Nq (i) can also be obtained by substituting the total noise Nt (i) and the electric noise Ne (i) into the above equation (8).

  As described above, the physical quantity Vq (i) is obtained based on the obtained signal S (i), the total noise Nt (i), and the electric noise Ne (i), and the average value of the physical quantity Vq (i) in each step is obtained. Also ask. The average value of the physical quantity Vq (i) in each step is ΣVq (i) / 9 (where ΣVq (i) is the sum of Vq (1), Vq (2),..., Vq (9)). Based on the average value of the physical quantity Vq (i) in each step, the above-described X-ray capture efficiency DQE is obtained. The average value of the physical quantity Vq (i) in each step is the Vq value finally obtained. Further, the quantum noise Nq (i) is obtained based on the obtained total noise Nt (i) and electrical noise Ne (i). Once the total noise Nt (i) and the electrical noise Ne (i) are obtained, the quantum noise Nq (i) is obtained by the above equation (8), which is an equation obtained by dividing S (i) by the above equation (9). Can do.

According to the X-ray image evaluation method according to the present embodiment, Vq (i) = {Nq (i)} ² / S (i) = [{Nt (i)} ² − {Ne, which is the above equation (9). The physical quantity Vq (i) in the i-th step is obtained by the equation (i)} ² ] / S (i). The physical quantity Vq (i) is an index of the quantum noise Ng (i) as Vq (i) = {Nq (i)} ² / S (i) in the above expression (9), and the above expression (9) It is possible to separate each noise as {Nq (i)} ² / S (i) = [{Nt (i)} ² − {Ne (i)} ² ] / S (i) . Therefore, the physical quantity Vq (i) can be obtained by separating each noise. Since this physical quantity Vq (i) is also an index of the X-ray capture efficiency DQE, the X-ray image (the fluoroscopic image in this embodiment) is easily evaluated by obtaining the X-ray capture efficiency DQE based on the physical quantity Vq (i). be able to.

  Further, according to the X-ray image evaluation apparatus shown in FIG. 1, the X-ray image evaluation method according to the present embodiment can be suitably implemented by including the signal processing unit 3.

  Next, evaluation examples under various conditions are shown in FIGS. In each figure, “Mean” represents an average value of pixel values, is a value proportional to the incident dose, “Vq” represents a physical quantity, and “Ne2” represents electrical noise.

  FIG. 7 shows an example in which the dose is changed without changing the tube voltage. In FIG. 7, the average value (see “Mean” in FIG. 7) increases, but the physical quantity (see “Vq” in FIG. 7) and electrical noise (see “Ne2” in FIG. 7). Has not changed. Vq does not change because Vq is an index that indirectly represents the X-ray capture efficiency DQE.

  FIG. 8 shows an example in which the ground wire is intentionally weakened to increase electrical noise. Only the electrical noise increases and there is no change in Vq.

  FIG. 9 shows the case where the tube voltage is changed with the dose kept constant. In addition, “Cu” in FIG. 9 is a copper plate used as a phantom, and “0.0 mm”, “1.0 mm”, “2.0 mm”, and “3.0 mm” indicate the thicknesses, respectively. It can be seen that Vq increases as the radiation quality becomes harder and the X-ray capture efficiency DQE deteriorates.

  FIG. 10 shows an example in which the digital gain of the system is changed. All have similar changes.

  The present invention is not limited to the above-described embodiment, and can be modified as follows.

  (1) In the above-described embodiments, the X-ray image evaluation apparatus has been described as being used in an X-ray fluoroscopic apparatus after shipment. However, other X-ray imaging apparatuses (for example, a tomographic apparatus and an X-ray CT apparatus) ) May be used.

  (2) In the above-described embodiment, the X-ray detection means is a flat panel X-ray detector. However, if the detection means is a commonly used detection means such as an image intensifier (II), There is no particular limitation.

  (3) In the above-described embodiments, the least square method has been described as an example, and the mean square error has been described as an example of variation. However, the present invention is not limited to this. As long as the physical quantity represents variation, the variation may be represented by a difference from a simple average value.

It is a block diagram of the X-ray image evaluation apparatus which concerns on an Example. It is the schematic which showed the fluoroscopic image and measurement location of the measurement phantom used for an Example. (A) is an image acquisition model in a flat panel X-ray detector (FPD), and (b) is a signal generation process of each image. It is the flowchart which showed a series of X-ray image evaluation processes. It is the table | surface which listed each signal in each step, noise, and a root mean square error. It is a graph when calculating | requiring an electrical noise by the least square method. This is an example of evaluation under various conditions, in which the dose was changed without changing the tube voltage. In the evaluation example under various conditions, the ground wire was intentionally weakened to increase the electric noise. In the evaluation example under various conditions, the tube voltage was changed with the dose constant. This is an example in which the digital gain of the system is changed in an evaluation example under various conditions.

Explanation of symbols

3 ... Signal processing unit Nq (i) ... Quantum noise (by X-ray quanta) S (i) ... Signal (of X-ray image) Ne (i) ... (Electric) electric noise Nt (i) ... Total noise Vq ( i) ... Physical quantity VQE ... X-ray capture efficiency

Claims (5)

A method of evaluating an X-ray image based on an X-ray image obtained at each step, wherein the quantum noise due to the X-ray quantum in the i-th step is set to Nq (i), and the X-ray image in the i-th step The line image signal is S (i), the electrical noise due to electricity in the i-th step is Ne (i), the total noise in the i-th step is Nt (i), and the physical quantity in the i-th step is Vq ( i), the physical quantity Vq (i) is
Vq (i) = {Nq (i)} ² / S (i)
= [{Nt (i)} ² − {Ne (i)} ² ] / S (i)
The physical quantity Vq (i) is obtained by the following formula, and, based on the physical quantity Vq (i), by obtaining an X-ray capture efficiency that is an efficiency of capturing an X-ray photon that is a photon converted from an X-ray, An X-ray image evaluation method characterized in that if the value of X-ray capture efficiency is small, the graininess is poor, and if the value of X-ray capture efficiency is large, the X-ray image is evaluated for graininess if the graininess is good .   The X-ray image evaluation method according to claim 1, wherein the signal S (i) and the total noise Nt (i) are obtained based on the X-ray image, and the electric noise Ne (i) does not depend on each step. An X-ray image evaluation method characterized in that the electric noise Ne (i) is obtained so that the variation with respect to the step of the physical quantity Vq (i) obtained by the above equation becomes a minimum value, assuming that the constant is constant.   The X-ray image evaluation method according to claim 2, wherein the physical quantity Vq (i) is obtained based on the obtained signal S (i), the total noise Nt (i), and the electric noise Ne (i). An X-ray image evaluation method, wherein the X-ray capture efficiency is obtained based on an average value of physical quantities Vq (i) in each step.   3. The X-ray image evaluation method according to claim 2, wherein the quantum noise Nq (i) is obtained based on the obtained total noise Nt (i) and electric noise Ne (i). Image evaluation method. An apparatus for evaluating an X-ray image based on an X-ray image obtained at each step, comprising signal processing means for performing signal processing based on the X-ray to obtain an X-ray image, the signal processing means comprising: ,
The quantum noise due to X-ray quanta in the i-th step is defined as Nq (i), the X-ray image signal in the i-th step is defined as S (i), and the electrical noise due to electricity in the i-th step is defined as Ne (i ), When the total noise in the i-th step is Nt (i) and the physical quantity in the i-th step is Vq (i), the physical quantity Vq (i) is
Vq (i) = {Nq (i)} ² / S (i)
= [{Nt (i)} ² − {Ne (i)} ² ] / S (i)
The physical quantity Vq (i) is obtained by the following formula, and signal processing for obtaining the X-ray capture efficiency, which is the efficiency of capturing the X-ray photons that are photons converted from the X-rays, based on the physical quantity Vq (i). X-ray image evaluation apparatus characterized in that if X-ray capture efficiency value is small, graininess is poor, and if X-ray capture efficiency value is large, graininess is good and X-ray image is evaluated for graininess .

Images