PROGRAM, IMAGE PROCESSING APPARATUS AND IMAGE PROCESSING METHOD

Abstract

To easily normalize a three-dimensional medical image.SOLUTION: An acquisition section 1 acquires a medical image. An analysis section 4 calculates an average value and a mode of a pixel value in each of a plurality of analysis areas included in a subject area in which a subject is photographed as a pixel average value and a pixel mode on the basis of the medical image. A calculation section 5 calculates a relational expression indicating a correlation between the pixel average value and the pixel mode in the analysis area on the basis of the pixel average value and the pixel mode in each of analysis areas. A correction section 6 generates a normalized image in which the pixel value at each pixel position of the medical image is corrected by calculating a correction coefficient on the basis of the relational expression and the medical image and multiplying the pixel value at each pixel position of the medical image by the correction coefficient.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to imaging techniques for nuclear medicine. [Background technology]

[0002] As a technology for administering a radiopharmaceutical to a subject and detecting the radiation emitted from the radiopharmaceutical to obtain medical images, technologies called SPECT (Single Photon Emission Computed Tomography) and PET (Positron Emission Tomography), which can obtain three-dimensional images, are attracting attention.

[0003] The pixel values ​​of three-dimensional medical images acquired by SPECT and PET are determined according to the count value of radiation emitted from radioactive materials accumulated in each tissue of the subject. Even if the condition of the tissue of the subject is the same, the count value of radiation varies depending on the characteristics of the imaging device that counts radiation, the examination time, the subject, etc., so it is not easy to compare three-dimensional medical images acquired by each of multiple examinations.

[0004] In response to this, Patent Document 1 discloses a nuclear medicine bone image analysis technique that normalizes three-dimensional image data obtained based on SPECT or PET to facilitate comparison of the three-dimensional image data. In this nuclear medicine bone image analysis technique, the three-dimensional image data is normalized by correcting each pixel value of the three-dimensional image data based on the amount of radioactivity administered to the subject and the subject's body weight or bone weight. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2014-174654 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the nuclear medicine bone image analysis technology described in Patent Document 1, in order to normalize the three-dimensional image data, the amount of radioactivity administered to the subject and the subject's body weight or bone weight must be obtained for each examination, which is time-consuming.

[0007] The present disclosure has been made in consideration of the above problems, and aims to provide a program, an image processing device, and an image processing method that are capable of easily normalizing three-dimensional medical images. [Means for solving the problem]

[0008] A program according to one aspect of the present disclosure causes a computer to implement an acquisition unit that acquires a medical image generated based on radiation from a subject administered a radioactive pharmaceutical; an analysis unit that calculates an average value and a most frequent value of pixel values ​​in each of a plurality of predetermined analysis regions included in a subject region in which the subject is shown in the medical image as a pixel average value and a pixel mode value; a calculation unit that calculates a relational equation indicating a correlation between the pixel average value and the pixel mode value in the analysis region based on the pixel average value and pixel mode value of each analysis region; and a correction unit that calculates a correction coefficient based on the relational equation and the medical image, and multiplies the pixel value of each pixel position of the medical image by the correction coefficient to generate a normalized image in which the pixel value of each pixel position of the medical image is corrected. Effect of the Invention

[0009] According to the present invention, it becomes possible to easily normalize three-dimensional medical images. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of an image processing device according to an embodiment of the present disclosure. [Diagram 2] 1 is a flowchart for explaining an operation of an image processing device according to an embodiment of the present disclosure. [Diagram 3]FIG. 3 is a diagram for explaining an example of steps S205 and S206 in FIG. [Figure 4] 11A and 11B are diagrams illustrating the relationship between analysis information, a relational expression, a reference value, and a target value in an embodiment of the present disclosure. [Diagram 5] FIG. 13 is a diagram illustrating an example of an evaluation result of a normalized image in the embodiment of the present disclosure. [Figure 6] FIG. 13 is a diagram showing another example of an evaluation result of a normalized image in the embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] FIG. 1 is a block diagram showing a configuration of an image processing device according to an embodiment of the present disclosure. The image processing device 100 shown in FIG. 1 is a device for generating a normalized image by normalizing a medical image generated based on radiation from a subject. Here, normalization refers to aligning pixel values ​​that should be at the same level in multiple medical images to the same level, thereby facilitating comparison of multiple medical images. More specifically, normalization refers to aligning pixel values ​​that differ depending on the imaging device, examination time, and subject that generates the medical images, to the same level depending on the condition of the subject's tissue, even if the condition of the subject's tissue is similar, thereby facilitating comparison of multiple medical images.

[0013] The image processing device 100 is configured, for example, by a computer system including a processor (computer) and a memory (both not shown). In this case, each component and each function of the image processing device 100 described below is realized, for example, by the processor reading a program and executing the read program. The program can be recorded in a computer-readable recording medium such as a memory.

[0014] The image processing device 100 is connected to an input / output device 101 and an auxiliary storage device 102. The input / output device 101 has an input device, such as a keyboard, a touch panel, and a pointing device, that receives various information from a user who uses the image processing device 100, and an output device, such as a display device and a printer, that outputs various information to the user. The input / output device 101 may also include a network interface device that transmits and receives various information via a communication network such as the Internet. The auxiliary storage device 102 is a storage device that stores various information, such as a large-capacity storage device. The auxiliary storage device 102 may store medical images.

[0015] The image processing device 100 includes an acquisition unit 1, a standardization unit 2, an extraction unit 3, an analysis unit 4, a calculation unit 5, and a correction unit 6.

[0016] The acquisition unit 1 acquires a medical image generated based on radiation from a subject administered with a radiopharmaceutical. The type of radiopharmaceutical administered to the subject is not particularly limited, and may be, for example, 99mTc-MDP, 99mTc-HMDP, or Na18F. The medical image is, for example, a three-dimensional image such as a SPECT image generated by SPECT or a PET image generated by PET. The three-dimensional image shows a plurality of pixel positions arranged in three dimensions and a pixel value of each pixel position. The pixel value is a value determined according to radiation emitted from the subject's tissue (for example, bone tissue and soft tissue), and is determined according to, for example, a count value obtained by counting radiation.

[0017] In this embodiment, the medical image is a bone SPECT image, which is a SPECT image generated for the purpose of examining bone tissue, particularly examining bone metastasis of cancer. The medical image also includes a bone region that captures the entire bone tissue of the subject as a subject region (subject image) that captures the subject.

[0018] The standardization unit 2 generates a standardized image by anatomically standardizing the medical image acquired by the acquisition unit 1 to a template image including a predetermined subject region.

[0019] Anatomical standardization is a process of correcting images so that corresponding points in two images have the same coordinates, and the standardization unit 2 can align (standardize) subject regions that differ according to the subject's physique, posture, etc., to a predetermined subject region by performing anatomical standardization. Note that anatomical standardization can be performed using, for example, SPM (Statistical Parametric Mapping) technology commonly used in the field of nuclear medicine.

[0020] In this embodiment, the predetermined subject region is a standard bone region that is an image of the entire bone tissue having a standard form (shape, size, etc.). In this case, the template image may be, for example, an image that includes the bone region of a specific real subject (preferably a subject with good posture and a standard body shape) as the standard bone region, or an image that includes the bone region of a fictitious subject created by a digital phantom or the like as the standard bone region.

[0021] The extraction unit 3 generates an extracted image by extracting, as a target region, a region consisting of pixel positions whose pixel values ​​fall within a predetermined range from the standard bone region, from the standardized image generated by the standardization unit 2. Note that the target region may include multiple regions separated from each other.

[0022] The predetermined range is a range including the average value of pixel values ​​in the entire standard bone region included in the standardized image, and is, for example, a range from mean-aSD to mean+bSD, where mean is the average value of pixel values ​​in the entire standard bone region and SD is the standard deviation of pixel values ​​in the entire standard bone region. a and b are values ​​greater than 0. Furthermore, a and b may be the same value or different values. However, when a and b are different from each other, either a or b may be 0.

[0023] The analysis unit 4 calculates the average value and the most frequent value of the pixel values ​​in each of a plurality of analysis regions preset in the standard bone region based on the pixel values ​​of each pixel position of the extracted image generated by the extraction unit 3, as the pixel average value and the pixel mode value, and generates analysis information indicating the pixel average value and the pixel mode value of each analysis region.

[0024] The analysis region is preferably, for example, a bone region that has few cases of bone metastasis empirically, has strong radiopharmaceutical accumulation, and has a certain volume of bone tissue. In this case, as described below, a correction coefficient for normalizing the medical image is calculated based on the pixel average value and pixel mode value of each analysis region, so that even if the subject is a patient with bone metastasis, normalization can be performed based on a region that is unlikely to have abnormally high radiopharmaceutical accumulation. Examples of such bones include the left femur, right femur, left sacrum, right sacrum, left ilium, right ilium, and spine. Since the bone density differs from bone to bone, the amount of radiopharmaceutical accumulation also differs from bone to bone. Therefore, the pixel average value and pixel mode value differ from analysis region to analysis region, and it is considered that, for example, the sacrum region is larger than the femur region.

[0025] The more the number of analysis regions, the more the effect of the characteristics of one analysis region on the normalized image can be reduced, and the more the effect of each analysis region can be homogenized. However, if the number of analysis regions exceeds a certain level, the processing load increases and such an effect is weakened. For this reason, it is desirable that the analysis regions are about 4 to 8 regions including four regions each showing the left femur, right femur, left sacrum, and right sacrum. In addition, the entire bone region may be set as one of the analysis regions. In addition, two or more analysis regions may partially overlap each other.

[0026] The calculation unit 5 calculates a relational equation showing the correlation between the pixel average value and the pixel mode value in the entire analysis region based on the analysis information generated by the analysis unit 4. The relational equation can be calculated, for example, by applying a predetermined approximation method such as the least squares method to the pixel average value and the pixel mode value of each analysis region. In this embodiment, the relational equation is a linear function (linear equation), but is not limited to this example. Note that by setting the analysis region to a region with few cases of bone metastasis as described above, the relational equation reflects the correlation between the pixel average value and the pixel mode value in normal areas without cancer cells.

[0027] The correction unit 6 calculates a correction coefficient based on the relational expression calculated by the calculation unit 5 and the extraction image generated by the extraction unit 3.

[0028] Specifically, the correction unit 6 first substitutes a predetermined reference value as the pixel average value of the analysis region into the relational expression calculated by the calculation unit 5, and specifies the pixel mode of the analysis region when the pixel average value of the analysis region is the reference value as the target value. The correction unit calculates the ratio between the specified target value and the pixel mode of the entire bone region shown in the extracted image extracted by the extraction unit 3 as a correction coefficient. The reference value is a pixel value that is a standard for normalization, and may be, for example, a fixed value or a value that can be set by the user.

[0029] At this time, the correction unit 6 may substitute a reference value as the pixel mode of the analysis region into the relational expression, and specify the pixel average value of the analysis region when the pixel mode of the analysis region is the reference value as the target value. In this case, the correction unit 6 calculates the ratio between the target value and the average pixel value of the entire bone region shown in the extracted image extracted by the extraction unit 3 as the correction coefficient.

[0030] After calculating the correction coefficient, the correction unit 6 multiplies the pixel value at each pixel position of the standardized image generated by the standardization unit 2 by the correction coefficient to generate a normalized image in which the pixel value at each pixel position of the standardized image is corrected.

[0031] FIG. 2 is a flowchart for explaining the operation of the image processing device 100. As shown in FIG.

[0032] First, the acquisition unit 1 acquires a bone SPECT image as a medical image (step S201). For example, the acquisition unit 1 may acquire the medical image from the outside via the input / output device 101, or may acquire the medical image stored in the auxiliary storage device 102.

[0033] The standardization unit 2 generates a standardized image by anatomically standardizing the medical image acquired by the acquisition unit 1 to a template image (step S202). The template image may be stored in advance in the image processing device 100 or the auxiliary storage device 102, or may be acquired from the outside via the input / output device 101 when performing anatomical standardization.

[0034] The extraction unit 3 generates an extracted image by extracting, as a target region, a region consisting of pixel positions whose pixel values ​​fall within a predetermined range from the standard bone region, from the standardized image generated by the standardization unit 2 (step S203).

[0035] The analysis unit 4 calculates the pixel average value and the pixel mode value in each of the multiple analysis regions based on the pixel values ​​of each pixel position of the extracted image generated by the extraction unit 3, and generates analysis information indicating the pixel average value and the pixel mode value of each analysis region (step S204).

[0036] The calculation unit 5 calculates a relational expression indicating the correlation between the pixel average value and the pixel mode value in the entire analysis region based on the analysis information generated by the analysis unit 4 (step S205).

[0037] The correction unit 6 calculates a correction coefficient based on the relational expression calculated by the calculation unit 5 and the extracted image generated by the extraction unit 3 (step S206). The correction unit 6 multiplies the pixel value of each pixel position of the standardized image generated by the standardization unit 2 by the correction coefficient to generate a normalized image in which the pixel value of each pixel position of the standardized image is corrected (step S207), and the process ends. The generated normalized image may be output via the input / output device 101 or may be stored in the auxiliary storage device 102.

[0038] Fig. 3 is a diagram for explaining an example of the processing in steps S205 and S206. Fig. 3 shows a diagram in which the horizontal axis represents the pixel mode value and the vertical axis represents the pixel average value. In this diagram, the analysis information (pixel average value and pixel mode value of each analysis region) A calculated in step S204 is shown as a point (black circle).

[0039] The calculation unit 5 applies a predetermined approximation method to the pixel average value and the pixel mode value shown at each point on the diagram to obtain a relational expression F that indicates the correlation between the pixel average value and the pixel mode value in the entire analysis region. Here, the calculation unit 5 uses the least squares method as the predetermined approximation method to obtain the relational expression F as a linear expression.

[0040] The correction unit 6 uses the relational expression F to calculate the pixel mode in the analysis region where the pixel average value is the reference value Y "300" as the target value X. Then, the correction unit 6 calculates the ratio of the target value X to the pixel mode Z of the entire bone region in the extracted image (specifically, the target value X / (pixel mode Z of the entire bone region)) as a correction coefficient. The correction unit 6 generates a normalized image by multiplying the pixel value of each pixel position of the standardized image generated by the standardization unit 2 by the correction coefficient. EXAMPLES

[0041] Next, an embodiment will be described. In this example, 21 subjects suspected of having bone metastasis from prostate cancer were administered Techne (registered trademark) MDP injection as a radiopharmaceutical, and bone SPECT images were obtained and used as medical images.

[0042] Of the 21 subjects, 9 were normal patients without bone metastasis and 12 were metastatic patients with bone metastasis. The imaging device used to image the subjects and generate bone SPECT images was the SIEMENS Symbia Intevo.

[0043] The acquisition conditions for bone SPECT images were as follows: the dose of radiopharmaceutical administered to the subject (administered radioactivity) was 1,041 ± 60 (mean ± SD) [MBq], the collimator was LEHR, the acquisition matrix size was 256 × 256 (pixel size: 2.4 mm), the number of projections was 120 [views], the acquisition method was 360° acquisition, the acquisition time was 15 [sec / view], the tube voltage was 130 [kV], and the tube current was 70 [mAs].

[0044] The bone SPECT images obtained by imaging each subject were used to generate standardized images by anatomically standardizing the subject area of ​​each subject using SPM technology. From the standardized images, pixel regions consisting of pixel positions whose pixel values ​​were within the average value of the entire bone area + 3SD were extracted as target regions.

[0045] The five bone regions corresponding to the left femur, right femur, left sacrum, right sacrum, and whole bone were analyzed, and the pixel average value and pixel mode value were calculated for each analysis region. The least squares method was applied to these pixel average values ​​and pixel mode values ​​to calculate the relational expression F as a linear equation.

[0046] In relational expression F, the reference value Y of "300" was substituted as the pixel average value for the entire analysis region, and the pixel mode value for the entire analysis region when the pixel average value was 300 was calculated as the target value X, and the correction coefficient was calculated as target value X / (pixel mode value Z for the entire bone region).

[0047] Then, a normalized image was generated by multiplying the pixel value at each pixel position of the standardized image by the correction coefficient.

[0048] Fig. 4 is a diagram showing the relationship between the analysis information A, the relational expression F, the reference value Y, and the target value X in this embodiment. Fig. 3 shows a diagram with the pixel mode on the horizontal axis and the pixel average on the vertical axis. In the diagram, the pixel average value and the pixel mode value of each analysis region, which are the analysis information A, are shown as dots (black circles). When the pixel mode value is x and the pixel average value is y, the relational expression F is calculated as y = 1.0647x - 49.612. Since the reference value Y is 300, the target value X is approximately 328.

[0049] Figures 5 and 6 are diagrams showing the evaluation results of normalized images. Specifically, Figure 5(a) shows the pixel average value and pixel mode value of the bone SPECT images before normalization for each of nine normal patients (No. 1 to No. 9), and Figure 5(b) shows the pixel average value and pixel mode value of the normalized images for each of the nine normal patients. Also, Figure 6(a) shows the pixel average value and pixel mode value of the bone SPECT images before normalization for each of twelve metastatic patients (No. 1 to No. 12), and Figure 6(b) shows the pixel average value and pixel mode value of the normalized images for each of the twelve metastatic patients.

[0050] As shown in Fig. 5(a) and Fig. 6(a), the pixel mean value and pixel mode value of the bone SPECT images vary widely between subjects, both in normal and metastatic patients.

[0051] In contrast, in the normalized image of a normal patient, as shown in FIG. 5(b), both the pixel average value and the pixel mode value are approximately 300, indicating that normalization has been performed with sufficient accuracy.

[0052] As shown in Fig. 6(b), for metastatic patients, the pixel mode and pixel average of the normalized images were approximately 300, except for subjects No. 6 and No. 12, and it can be seen that normalization was performed with sufficient accuracy. Note that for subjects No. 6 and No. 12, the pixel average was lower than 300 due to their symptoms. Furthermore, subjects No. 6 and No. 12 showed diffuse accumulation of the radiopharmaceutical due to bone metastasis, and the analysis region used as the standard for normalization was the diseased area rather than the normal area, so the pixel mode was larger than the pixel average.

[0053] As described above, according to the present disclosure, the analysis unit 4 calculates the average value and the most frequent value of pixel values ​​in each of a plurality of analysis regions included in the subject region in which the subject is depicted as a pixel average value and a pixel mode value based on the medical image. The calculation unit 5 calculates a relational expression indicating the correlation between the pixel average value and the pixel mode value in the analysis region based on the pixel average value and the pixel mode value of each analysis region. The correction unit 6 calculates a correction coefficient based on the relational expression and the medical image, and multiplies the pixel value of each pixel position of the medical image by the correction coefficient to generate a normalized image in which the pixel value of each pixel position of the medical image is corrected.

[0054] This makes it possible to generate normalized images from medical images without obtaining information such as the amount of radioactivity administered to the subject, or the subject's body weight and bone weight, making it possible to easily normalize three-dimensional medical images.

[0055] Furthermore, according to the present disclosure, the correction unit 6 calculates a correction coefficient, which is the ratio between a target value, which is the pixel mode of the analysis region when the pixel average value of the analysis region is a reference value, and the pixel mode of the entire subject region shown in the extracted image extracted by the extraction unit 3, and generates a normalized image by multiplying the pixel value of each pixel position of the three-dimensional image by the correction coefficient. Therefore, it becomes possible to generate a normalized image based on a desired pixel average value, thereby enabling more appropriate normalization.

[0056] Furthermore, according to the present disclosure, the correction unit 6 calculates a correction coefficient, which is the ratio between a target value, which is the average pixel value of the analysis region when the pixel mode value of the analysis region is a reference value, and the average pixel value of the entire subject region shown in the extracted image extracted by the extraction unit 3, and generates a normalized image by multiplying the pixel value of each pixel position of the three-dimensional image by the correction coefficient. Therefore, it becomes possible to generate a normalized image based on a desired pixel mode value, thereby enabling more appropriate normalization.

[0057] According to the present disclosure, the standardization unit 2 generates a standardized image by anatomically standardizing the medical image to a template image including a predetermined subject region. In this case, it is possible to normalize the medical image after aligning the subject region that differs according to the subject's physique and posture to the predetermined subject region, and therefore it is possible to generate a normalized image that can be compared more easily.

[0058] According to the present disclosure, the extraction unit 3 generates an extracted image by extracting a pixel region consisting of pixel positions whose pixel values ​​are within a predetermined range from the subject region from the medical image. In this case, it becomes possible to normalize the medical image by removing the influence of artifact parts such as the bladder where radiopharmaceuticals unintentionally accumulate, thereby enabling more appropriate normalization.

[0059] According to the present disclosure, the predetermined range is a range that includes the average value of the pixel values ​​in the subject region, which makes it possible to appropriately eliminate the influence of artifacts and enables more appropriate normalization.

[0060] According to the present disclosure, the analysis region includes four regions, each of which corresponds to the left femur, the right femur, the left sacrum, and the right sacrum. In this case, normalization can be performed based on the region that is less likely to have abnormally high accumulation of the radiopharmaceutical, thereby enabling more appropriate normalization.

[0061] The above-described embodiments and examples of the present disclosure are illustrative examples of the present disclosure, and are not intended to limit the scope of the present disclosure to only those embodiments. A person skilled in the art can implement the present disclosure in various other forms without departing from the scope of the present disclosure.

[0062] For example, the medical image may be a two-dimensional image obtained by a bone scintigraphy test or the like. At least one of the standardization unit 2 and the extraction unit 3 may be omitted. The extraction unit 3 may generate an extracted image from the medical image, and the standardization unit 2 may generate a standardized image from the extracted image. [Explanation of symbols]

[0063] 1: Acquisition unit, 2: Standardization unit, 3: Extraction unit, 4: Analysis unit, 5: Calculation unit, 6: Correction unit, 100: Image processing device, 101: Input / output device, 102: Auxiliary storage device

Claims

1. An acquisition unit for acquiring a bone SPECT image of a subject; a standardization unit that generates a standardized image by anatomically standardizing the bone SPECT image to a template image including a subject region, the subject region includes a plurality of bone regions; The bone region includes at least one of a left femur, a right femur, a left sacrum, a right sacrum, a left ilium, a right ilium, and a spine.

2. The computer-implemented program of claim 1 , wherein the bone regions include two or more of a left femur, a right femur, a left sacrum, a right sacrum, a left ilium, a right ilium, and a spine.

3. The program according to claim 1 , wherein the total number of bone regions is 4 to 8.

4. The method of claim 1 , wherein the bone region includes at least one of a left femur, a right femur, a left sacrum, and a right sacrum.

5. the total number of bone regions is seven; The method of claim 1 , wherein the bone regions include a left femur, a right femur, a left sacrum, a right sacrum, a left ilium, a right ilium, and a spine.

6. The total number of bone regions is four; The computer-implemented program of claim 4 , wherein the bone regions are a left femur, a right femur, a left sacrum, and a right sacrum.

7. The program described in claim 1, wherein the bone region is composed of two or more of the left femur, right femur, left sacrum, right sacrum, left ilium, right ilium, and spine.

8. The program described in claim 1, wherein the bone region is composed of two or more of the left femur, right femur, left sacrum, and right sacrum.

Images