JP6442310B2 - Technology for extracting tumor regions from nuclear medicine images - Google Patents

Description

  The invention disclosed in the present application generally relates to a technique for analyzing a nuclear medicine image, and more specifically to a technique for extracting a tumor region in a nuclear medicine image.

SPECT preparation ^99m Tc-HMDP and PET preparation ¹⁸ F-NaF are thought to be mainly chemically adsorbed to hydroxyapatite crystals, which are the basic composition of bone minerals. Because it adsorbs a lot, it is possible to visualize highly inflamed sites such as bone metastasis and fractures of malignant tumors.

The SPECT preparation ^99m Tc-HMDP is quickly distributed in bone tissue throughout the body after intravenous administration, and a high-quality bone scintigram can be obtained in about 1-2 hours after administration.

It has been reported that ¹⁸ F-NaF, a PET preparation, can provide a stable bone scintigram 30 minutes after administration. Moreover, since PET can accurately correct attenuation and scattering in the body, it is possible to obtain an image having high quantitativeness.

  Bone scintigrams can qualitatively (visually) detect abnormal lesions, but they also accumulate in the excretory pathways of the kidney and bladder, which may hinder analysis.

In order to solve this problem, the inventor of the present application discloses the following invention relating to a technique for analyzing nuclear medicine image data. This invention
Extracting a bone region from CT image data that has been aligned with the nuclear medicine image data;
-Displaying the data of the nuclear medicine image data that overlaps the extracted bone region;
-Of the nuclear medicine image data, do not display data of a region that does not overlap with the extracted bone region;
It is characterized by.

  For example, in the analysis of bone scintigrams, when using the conventional method of extracting a region above a predetermined threshold as a lesion, the organs such as kidney and bladder that are physiologically accumulated are displayed and extracted as highly accumulated sites. End up. On the other hand, according to the above-mentioned invention, since the data of the region that does not overlap the bone region is not displayed in the nuclear medicine image data, the influence of accumulation due to normal physiological reasons is excluded, and the nuclear medicine image data Can be observed and analyzed.

  In the present specification and claims, the term “nuclear medicine image data” is a concept including both two-dimensional nuclear medicine image data and three-dimensional nuclear medicine image data. The above invention can be applied to both two-dimensional nuclear medicine image data and three-dimensional nuclear medicine image data.

  When a tumor is observed in a bone scintigram, it may be necessary to extract a region corresponding to the tumor from nuclear medicine image data. In the conventional technique for doing this, nuclear medicine image data is displayed on a display and a tumor is extracted by tracing the outline of the tumor with a mouse or the like, and an appropriate pixel threshold is set, and a pixel value equal to or greater than the threshold is set. There is a method for determining a region having a tumor as a tumor.

  However, in the former case, the size of the extracted tumor varies depending on the skill and preference of the operator. In the latter case, the optimum threshold must be different depending on patient conditions such as dose and physiological accumulation, imaging conditions, and imaging apparatus. As a result, these two approaches can cause variations in the size of the extracted tumor.

In order to solve this problem, the inventor of the present application discloses the following invention relating to a technique for automatically extracting a tumor from nuclear medicine image data. This invention
Obtaining a pixel group of a region corresponding to a bone from a pixel group included in nuclear medicine image data;
Obtaining a standard deviation based on at least some pixel values of the acquired pixel group;
Extracting a bone tumor pixel group based on the standard deviation;
It is characterized by.

  According to the above invention, since the bone tumor pixel group is statistically extracted, variation in the extraction result by the operator is suppressed, and the tumor extraction can be performed objectively.

  The above-described invention can also be applied to two-dimensional nuclear medicine image data and three-dimensional nuclear medicine image data.

  When a tumor is observed in a three-dimensional bone scintigram, it may be necessary to extract the contour of the tumor three-dimensionally. Conventional techniques for doing this include a method of extracting contours by displaying nuclear medicine image data on a display and tracing the contour of a tumor with a mouse or the like, and an appropriate pixel threshold value after pre-designating a range including the tumor. And a boundary where the pixel value exceeds the threshold value is defined as a tumor contour.

  However, in the former case, the extracted tumor shape varies depending on the skill and preference of the operator. Further, manual contour extraction in a three-dimensional manner is a cumbersome operation and is troublesome for the operator, and the shape varies depending on the operator. In the latter case, the detected tumor contour may vary depending on the set threshold value.

In order to solve this problem, the inventor of the present application discloses the following invention relating to a technique for automatically extracting the outline of a tumor from three-dimensional nuclear medicine image data. This invention
In the nuclear medicine image data, setting a reference point in a region including a tumor;
Scanning the pixels from the reference point in a three-dimensional radial pattern and creating a pixel value profile in each scanning direction;
Creating a gradient intensity profile for the pixel value profile or a smoothing process applied to the pixel value profile;
Setting a contour determination range for the pixel value profile or the pixel value profile obtained by applying a smoothing process;
The tumor contour point in the pixel value profile is determined based on the vertex having the maximum kurtosis in the contour determination range in which the corresponding gradient strength profile or the gradient strength profile is subjected to smoothing processing. To do;
It is characterized by.

  According to the above invention, since the outline of the tumor is automatically extracted, variation in the extraction result by the operator can be suppressed, and the tumor outline can be objectively extracted. Further, since the contour extraction is automatically performed, the labor of the operator is remarkably reduced.

  The present invention generally relates to a system having means for performing at least one of the above characteristics, a program for causing the system to execute at least one of the above characteristics when executed by the processing means, and at least one of the above characteristics. It can be embodied as a method for performing any one of them.

An example of a preferred embodiment of the present invention is a computer program for analyzing nuclear medicine image data, the computer program being executed by a processing means of the system,
Obtaining a pixel group of a region corresponding to a bone from a pixel group included in the nuclear medicine image data;
Obtaining a standard deviation based on at least some pixel values of the acquired pixel group;
Extracting a bone tumor pixel group based on the standard deviation;
A computer program comprising a group of program instructions configured to perform
Another example of a preferred embodiment of the present invention is a system having processing means and storage means for storing the computer program.
Another example of a preferred embodiment of the present invention is a method of analyzing nuclear medicine image data performed by the processing means of the system performed by the system processing means:
Obtaining a pixel group of a region corresponding to a bone from a pixel group included in the nuclear medicine image data;
Obtaining a standard deviation based on at least some pixel values of the acquired pixel group;
Extracting a bone tumor pixel group based on the standard deviation.

  Some of the preferred embodiments of the present invention are specified in the claims included in the claims. However, the configurations specified in these claims do not necessarily include all of the novel technical ideas disclosed in this specification and the drawings. The applicant shall allege that he has the right to obtain a patent for all new technical ideas disclosed in the present specification and drawings, regardless of whether they are described in the present claims. Is noted.

FIG. 2 is a diagram for describing a main configuration of a system 100, which is an example of hardware that can execute various processes disclosed in the present specification. It is a flowchart for demonstrating the novel process 200 which extracts a bone area | region from the nuclear medicine image data 130. FIG. In step 210, it is a figure which introduces the example of the image of the bone created with the pixel which remained only by extraction by a threshold value. It is a figure which introduces the example of the image obtained by adding a closing process to the image of Drawing 3A. 11 is a diagram for explaining an example of display in step 214. FIG. 11 is a diagram for explaining an example of display in step 214. FIG. It is a flowchart for demonstrating the novel process 500 which extracts a tumor automatically from nuclear medicine image data. It is an example of a display of the bone tumor pixel group extracted by step 518. It is a flowchart for demonstrating the novel process 700 which extracts the outline of a tumor automatically from nuclear medicine image data. An example of processing that can be performed in step 720 is shown. The example of the display performed at step 708 is shown. An example of the reference point set in step 710 is shown. FIG. 7 is a diagram for explaining a state of scanning performed in step 712. 7 is an example of a pixel profile created in step 712. It is a figure which shows an example of the smoothing performed by step 714. 7 is an example of a gradient intensity profile created in step 716. It is a figure for demonstrating an example of the process which determines the range which determines the tumor outline point performed at step 720. FIG. It is a figure for demonstrating an example of the determination process of the tumor outline point performed at step 722. FIG. It is a figure for demonstrating an example of the determination process of the tumor outline point performed at step 722. FIG. It is a figure for demonstrating an example of the display of the extracted outline point performed at step 724. FIG. It is a flowchart for demonstrating the application example 1200 of the process 200 introduced using the flowchart of FIG. FIG. 11 is a diagram for describing an example of the effect of a process 1200.

DESCRIPTION OF PREFERRED EMBODIMENTS

  FIG. 1 is a diagram for explaining a main configuration of a system 100, which is an example of hardware capable of executing various processes disclosed in the present specification. As illustrated in FIG. 1, the system 100 is similar to a general computer in hardware, and includes a CPU 102, a main storage device 104, a mass storage device 106, a display interface 107, a peripheral device interface 108, a network. An interface 109 or the like can be provided. As with a general computer, a high-speed RAM (Random Access Memory) can be used as the main storage device 104, and an inexpensive and large-capacity hard disk or SSD can be used as the large-capacity storage device 106. it can. A display for displaying information can be connected to the system 100, which is connected via a display interface 107. In addition, a user interface such as a keyboard, a mouse, and a touch panel can be connected to the system 100, and this is connected via the peripheral device interface 108. The network interface 109 can be used to connect to another computer or the Internet via a network.

  The mass storage device 106 may store an operating system (OS) 110, a nuclear medicine image analysis program 120 for providing characteristic processing disclosed in the present specification, and the like. The most basic functions of the system 100 are provided by the OS 110 being executed by the CPU 102. The characteristic processing disclosed in the present specification is provided when at least a part of a program instruction group included in the nuclear medicine image analysis program 120 is executed by the CPU 102. The characteristic processes disclosed in the present specification can be roughly divided into three types, one of which is a process of removing a physiological accumulation site, the other of which is an automatic extraction process of a tumor region, and the last one. Is an automatic extraction process of the tumor contour. The nuclear medicine image analysis program 120 may include a program instruction group or program modules 122, 124, and 126 in order to perform each of these three types of processing. Depending on the embodiment, each of these program instructions or program modules may be stored, copied, downloaded, or sold independently of other program instructions or programs modules. Depending on the embodiment, these program instruction groups or program modules may be inseparably integrated with each other to form a single nuclear medicine image analysis program 120. As is well known, there are various implementation forms of the program, and all these forms are included in the scope of the invention disclosed in the present application.

  The mass storage device 106 further includes nuclear medicine image data 130 to be analyzed by the nuclear medicine image analysis program 120, CT image data 132 corresponding to the nuclear medicine image data 130, other data 134, 136, 138, and the like. Can also be stored.

  In addition to the elements depicted in FIG. 1, the system 100 can have a configuration similar to that of an apparatus included in a normal computer system, such as a power supply or a cooling device. The implementation form of the computer system includes storage device distribution / redundancy and virtualization, use of multiple CPUs, CPU virtualization, use of a processor specialized for specific processing such as DSP, and implementation of specific processing in hardware. Various forms using various techniques, such as combining, are known. The invention disclosed in the present application may be mounted on any form of computer system, and the scope thereof is not limited by the form of the computer system. The technical idea disclosed in this specification is generally (1) executed by a processing unit to cause an apparatus or a system including the processing unit to perform various processes described in this specification. (2) an operation method of an apparatus or a system realized by the processing means executing the program, and (3) the program and the program. The present invention can be embodied as an apparatus or system provided with processing means. As described above, some software processing may be implemented as hardware.

  It should be noted that the data 130 to 138 are often not stored in the mass storage device 106 when the system 100 is manufactured and sold or activated. These data may be data transferred from an external device to the system 100 via the peripheral device interface 108 or the network interface 109, for example. Depending on the embodiment, the data 134 to 138 may be formed by the CPU 120 executing the program 120 or another program. Further, depending on the implementation form of the program 120 and the OS 110, the data 134 to 138 may not be stored in the mass storage device 106 but may be stored only in the main storage device 104. It should be noted that the scope of the invention disclosed in the present application is not limited by the presence or absence of the data 130 to 138.

Next, the nuclear medicine image data 130 to be processed in the embodiment disclosed in this specification will be described. The nuclear medicine image data 130 can be, for example, two-dimensional image data or three-dimensional image data obtained by a SPECT examination performed for the purpose of a bone metastasis examination of a tumor or the like. More specifically, for example, ^99m Tc-HMDP as a radiopharmaceutical is intravenously administered to a subject, radiation emitted from the body is detected by a SPECT device, and two-dimensional image data formed based on the radiation count value Or three-dimensional image data. Usually, each pixel value of the pixels constituting these images has a value corresponding to the radioactivity count value, that is, each pixel value represents the intensity of radioactivity.

It should be noted that the data that can be analyzed by the embodiment disclosed in the present specification is not limited to SPECT image data. There are various types of data that can be analyzed by the examples. For example, PET image data using ¹⁸ F-NaF as a radiopharmaceutical can also be analyzed.

  The CT image data 132 can be three-dimensional image data obtained by using the X-ray CT for the same person as the subject who created the nuclear medicine image data 130. Currently, an apparatus in which SPECT and CT are integrated has been developed, and it is possible to take a CT image in conjunction with the execution of SPECT inspection. Image data 130 and 132 may be image data taken together using such a device.

  Next, some examples of characteristic processing disclosed in this specification will be described with reference to FIG.

  FIG. 2 is a flowchart for explaining the bone region extraction processing 200 disclosed in this specification. The process introduced in this flowchart can be, for example, a process performed by the system 100 when the bone region extraction program module 122 of the nuclear medicine image analysis program 120 is executed by the CPU 102.

  Step 202 indicates the start of processing. In step 204, the three-dimensional nuclear medicine image data 130 to be processed in this embodiment is read. For example, the CPU 102 copies at least a part of the image data 130 from the mass storage device 106 to the main storage device 104 in accordance with an instruction of a program instruction group included in the nuclear medicine image analysis program 120.

  In some embodiments, the nuclear medicine image data 130 can be stored in a storage medium, read by a dedicated reading device, and taken into the system 100 via the peripheral device interface 108. In another aspect, the nuclear medicine image data 130 may be captured via the network interface 109 as a data signal superimposed on a carrier wave. Depending on the embodiment, the captured nuclear medicine image data 130 may be temporarily stored in the mass storage device 106 and copied to the main storage device 104, or may be directly stored in the main storage device 104 and used for processing. May be.

  In step 206, CT image data 132 is read.

  In step 208, the nuclear medicine image data 130 and the CT image data 132 are aligned (co-registration). That is, the orientation, size, and position of the body are three-dimensionally matched between the nuclear medicine image 130 and the CT image 132. Through this processing, the nuclear medicine image 130 and the CT image 132 can be compared with each other. This process is sometimes simply referred to as registration. Since the co-registration function is included in many nuclear medicine image data analysis programs available on the market, in this embodiment, the existing methods implemented in such a program and other known methods are also used. May be used for co-registration. Co-registration is a manual method, that is, both nuclear medicine image data and CT image data are displayed as images, and one image is translated and rotated by the operation of a mouse etc. You may carry out by the method of matching.

  In some embodiments, the co-registration between the image data 130 and 132 may have already been completed. For example, when the image data 130 and 132 are taken together by an apparatus in which SPECT or PET and CT are integrated, the co-registration of the nuclear medicine image data 130 and the CT image data 132 is performed by the apparatus. May have already been completed at the output stage. In that case, of course, the process of step 208 is unnecessary.

In step 210, a bone region is extracted three-dimensionally using the CT image data 132. Depending on the embodiment, this processing may be simply extracting a region having a pixel value equal to or greater than a threshold value on the CT image as a bone region. However, there are cases where the inside of the bone is lost or noise is generated only by extraction based on the threshold. Therefore, depending on the embodiment, the remaining pixels obtained by applying the closing process by a known method to the image data constructed by the pixels having the pixel value equal to or larger than the threshold value may be determined as the bone region. Further, as a noise removal method, a method of removing a volume whose number of Voxels is equal to or less than a certain threshold value or a method of manually specifying and removing noise may be used.
As the threshold used for extracting the bone region, a value generally known to those skilled in the art as a Hans field value indicating the bone region can be used. In this embodiment, 152 is used.

  FIG. 3A introduces an example of a bone image created by pixels remaining only by extraction using a threshold. FIG. 3B introduces an example of an image obtained by applying a closing process to the image of FIG. 3A. As shown in FIG. 3A, it can be seen that even within the bone, it is displayed black depending on the location, and there is a portion that is not extracted as a bone region. On the other hand, in the image of FIG. 3B, the inside of the bone and the space between the vertebrae are displayed in white, and it can be seen that the bone region is extracted better than the image of FIG. 3A. Therefore, in this embodiment, data obtained by applying a closing process to image data extracted by a threshold, that is, data that can be imaged as illustrated in FIG. 3B is used as an extracted bone region for subsequent processing. To do. In the present embodiment, the number of closing processes is three. The extracted bone region data has, for example, a pixel value of 1 or more for pixels extracted as a bone region, and 0 or a NULL code is assigned to a pixel value for pixels not extracted as a bone region. It may be image data. The extracted bone region data may be stored, for example, in the main storage device 104 and / or may be stored, for example, as data 134 in the mass storage device 106 (step 212).

  In step 210, it is preferable that the threshold value used for extracting the bone region and the number of closing processes can be arbitrarily adjusted by the user. The nuclear medicine image analysis program 120 (or bone region extraction program module 122) is preferably configured to allow such arbitrary settings.

  In step 214, the nuclear medicine image data 130 is displayed using the bone region data obtained in step 210. In the display process in this step, among the pixels included in the nuclear medicine image data 130, pixels that overlap the bone region extracted in step 210 are displayed, but pixels that do not overlap the bone region are not displayed. And For example, for each pixel of the nuclear medicine image data 130, it is checked whether or not there is a valid pixel value other than 0 or NULL code in the corresponding pixel of the bone region data 134 obtained in step 210. Only when a value exists, the pixel may be displayed (that is, a pixel not included in the bone region data 134 is not displayed). As another aspect, for each pixel of the nuclear medicine image data 130, a process such as adding 0 or a NULL code to a pixel other than the corresponding pixel of the bone region data 134 obtained in step 210 is performed. good.

  An example of what the display in step 214 looks like will be described with reference to FIGS. 4A and 4B. First, FIG. 4A is an image in which an appropriate cross section is cut out and displayed from the CT image data 132 and the same cross section is cut out from the nuclear medicine image data 130 and displayed. Radioactivity accumulation can be observed at the locations indicated by reference numerals 402 and 404. However, the place indicated by reference numeral 402 is the bladder, and the reason why the radioactivity is accumulated is the excretion effect, not the presence of the lesion. In contrast, the accumulation observed at the location indicated by 404 is probably due to the presence of a lesion.

  FIG. 4B is also an image displayed by superimposing the corresponding cross section of the nuclear medicine image data 130 on the same CT image used in FIG. 4A. However, the displayed pixels of the nuclear medicine image data are only the pixels that overlap the bone region extracted in step 210. As can be readily seen, no radioactivity accumulation is displayed in the kidney or bladder region. On the other hand, the accumulation observed at the location indicated by reference numeral 404 remains as it is. That is, in FIG. 4B, the accumulation due to physiological reasons hardly appears in the displayed nuclear medicine image, while the accumulation due to the lesion appears properly in the displayed nuclear medicine image. Therefore, according to the above processing, it becomes possible to perform observation and analysis of nuclear medicine image data without the influence of accumulation due to normal physiological reasons.

Regarding the above processing, Non-Patent Document 1 listed below discloses a technique for detecting an abnormal part from a CT image based on a normal model constructed using a plurality of CT image data. However, a technique for extracting a bone region from nuclear medicine image data and accurately detecting a tumor site is not disclosed. In addition, there is no example disclosed so far for a method for effectively removing a physiological accumulation site.
Ei Zhou Mukai, Hiroshi Fujita, “Recognition of Normal Anatomical Structure in Simple CT Images of the Trunk Region”, MEDICAL IMAGING TECHNOLOGY, Vol.24 No.3, May 2006, p.167-172

  The difference between FIG. 4A and FIG. 4B becomes clearer when displayed in color, and it becomes possible to observe very clearly that nuclear medicine information is displayed only in the bone region. (Actually, FIGS. 4A and 4B are color images, but are represented in gray scale due to limitations of the drawings that can be used in patent applications.) Also, FIGS. 4A, 4B, and 3A Note that FIG. 3B and FIG. 3B are images created using one section of the three-dimensional image data, and the nuclear medicine image data 130 and the CT image data 132 are three-dimensional image data.

  By the way, the display of the image data may be performed by sending a digital or analog signal for displaying the image data through the display interface 107 to a display device connected to the display interface 107. Therefore, the phrase “display” used in the specification and claims of the present application is not only the actual display on the display device but also the digital or analog for displaying the image on the display device. Note that this also includes generating

  In FIG. 4B, the nuclear medicine image data superimposed on the CT image has substantially only a pixel value that overlaps the bone region (eg, other than 0 or Null code) and does not overlap the bone region. The image data is such that a pixel value such as 0 or a NULL code is assigned to the pixel. Depending on the embodiment, such image data may be stored in, for example, the mass storage device 106 as, for example, the bone region nuclear medicine image data 136 (step 216).

  Step 218 indicates the end of the process.

  Next, a new process for automatically extracting a tumor from nuclear medicine image data will be described with reference to FIG.

  FIG. 5 is a flowchart for explaining a process 500 for extracting a tumor from the nuclear medicine image data 130 disclosed in the present specification. The process introduced in this flowchart can be, for example, a process performed by the system 100 when the tumor extraction program module 124 of the nuclear medicine image analysis program 120 is executed by the CPU 102.

  Step 502 indicates the start of processing. In step 504, the three-dimensional nuclear medicine image data 130 to be processed 500 is read. In step 506, a pixel group in a region corresponding to a bone is acquired from the nuclear medicine image data 130. Depending on the embodiment, this process can be the process 200 itself described with reference to FIG. Thus, in some embodiments, step 506 may be to read the bone region nuclear medicine image data 136 stored in step 216 of process 200 from storage means (eg, mass storage device 106).

In step 508, a histogram is created based on the pixel values of the pixel group acquired in step 506. Depending on the embodiment, a histogram of the pixel value itself may be created. However, the pixel value of each pixel of the image data 130 normally represents a count value of radiation, and is a value that varies depending on the weight of the subject, the radioactive dose, and the like, and is a value that varies between measurements. Therefore, in a preferred embodiment of this step, it is preferable to first convert each pixel value into a normalized value such as SUV (Standardized Uptake Value) to create a histogram such as SUV. This is because SUV is a normalized value, and if it is a similar tissue, it is a value that can be expected to have a similar value to some extent regardless of measurement conditions. The SUV is a value defined as follows, as is well known in the technical field of the present application.

SUV = {radioactivity in the region of interest corrected for attenuation (kBq) ÷ region of interest volume (ml)} / {radioactivity dose (MBq) ÷ subject's body weight (kg)}

  That is, the SUV is a value obtained by normalizing the radioactivity concentration of the region of interest by the radioactivity dose per body weight, and is a value that can serve as an index reflecting the radioactivity or radiopharmaceutical intake.

  In step 510, the peak value of the created histogram is determined. That is, the largest frequency (frequency) is specified. In step 512, a class value whose frequency is a predetermined ratio of the peak value is determined in the created histogram. For example, the class value at which the frequency is 5% of the peak value is determined. Two class values whose frequency is 5% of the peak value are determined before and after the class value giving the peak value. In step 512, the section defined by these two class values is determined as the section for calculating the standard deviation calculated in the next step.

  In step 514, the standard deviation of the class value included in the section determined in step 512 is calculated.

In step 516, a threshold value used for extraction of the tumor pixel group is determined. The threshold value T calculated in step 516 can be expressed by the following equation when the large class value of the two class values calculated in step 512 is expressed as BaseValue and the standard deviation calculated in step 514 is expressed as SD.

T = BaseValue + α · SD

Here, α is a constant. When α is large, the number of pixels extracted as tumor pixels decreases, and when α is small, the number of pixels extracted increases. For this reason, it is preferable that the value of α can be adjusted depending on the difference in the nuclear medicine examination apparatus and the preference of the doctor. The nuclear medicine image analysis program 120 (or tumor extraction program module 124) is preferably configured to allow arbitrary setting of α. However, according to the present inventors' investigation, it has been found that if 0.5 is adopted as α, an appropriate result can be obtained in many cases.

  In step 518, the pixel value or the pixel whose SUV calculated based on the pixel value is equal to or greater than the threshold value determined in step 516 is determined as a pixel group (bone tumor pixel group in which a bone tumor may exist). ).

  In step 520, the bone tumor pixel group extracted in step 518 is displayed. FIG. 6 shows an example of the display. In this figure, an appropriate cross section is cut out and displayed from the three-dimensional nuclear medicine image data 130, and a bone tumor pixel group located on the cross section is superimposed and displayed.

  In the figure, the portion indicated by a dark color indicated by the symbol T is the extracted bone tumor pixel group. The portion indicated by the symbol B is also shown in a dark color, but this is a pixel group corresponding to the bladder and is only displayed darkly due to physiological accumulation of radioactivity. Since the pixel group corresponding to the physiological accumulation site in the kidney, the bladder or the like is not included in the bone region pixel group extracted in step 506, it is not extracted as a tumor in step 518. In the embodiment actually used, it is desirable that the bone tumor pixel group T is displayed in a color that is clearly different from the physiologically integrated pixel group B so that the bone tumor pixel group can be clearly distinguished. FIG. 6 is also actually a color image, and the cross section of the nuclear medicine image data 130 is displayed in gray scale including the physiologically integrated pixel group B, and the bone tumor pixel group T is displayed in red. It was. FIG. 6 is a gray scale drawing of such a drawing due to limitations of the figures that can be used in the patent application.

  According to the above processing, since the bone tumor pixel group is statistically extracted using the histogram and the standard deviation, variation in the extraction result by the operator can be suppressed, and the tumor extraction can be performed objectively. On the other hand, since α is arbitrarily set, flexible setting according to the operator's request is also possible.

  Note that in the present embodiment, the processing from step 506 to step 518 is performed three-dimensionally. That is, the bone region pixel group extracted in step 506 has a three-dimensional spread. The standard deviation calculated in step 514 is a standard deviation of pixel values of a bone region pixel group having a three-dimensional spread. The bone tumor pixel group extracted in step 518 also has a three-dimensional spread. The cross-sectional view shown in FIG. 6 shows only one cross section.

  However, depending on the embodiment, the processing from step 506 to step 518 may be performed two-dimensionally using specific cross-sectional data of the three-dimensional nuclear medicine image data 130.

In some embodiments, the threshold value set in step 516 is determined using the standard deviation calculated in step 514 and the average value inevitably calculated in the calculation of the standard deviation.

Threshold = average value + β · standard deviation

You may set as follows. Further, depending on the embodiment, the above threshold value may be set by calculating the average value or the standard deviation using all the pixel groups extracted in step 506. Furthermore, depending on the embodiment, the threshold value may be set by calculating the average value or the standard deviation without changing each pixel to SUV and maintaining the original pixel value. It has been investigated by the present inventor that the tumor pixel group may be extracted satisfactorily even if the threshold value is set by these methods. Note that β is a positive number, and is a value that can be arbitrarily set by the user. However, it has been investigated by the present inventor that a tumor pixel group can be satisfactorily extracted in many cases when β = 1. .

  Depending on the embodiment, the bone tumor pixel group data extracted in step 518 may be stored. For example, the bone tumor pixel group data 138 may be stored, for example, in the mass storage device 106 (step 522). Step 524 indicates the end of the process.

  Next, a novel process for automatically extracting the outline of a tumor from nuclear medicine image data will be described with reference to FIG. 7A.

  FIG. 7A is a flowchart for illustrating a process 700 for extracting a tumor contour from nuclear medicine image data 130 disclosed herein. The processing introduced in this flowchart can be, for example, processing performed by the system 100 when the tumor contour extraction program module 126 of the nuclear medicine image analysis program 120 is executed by the CPU 102.

  Step 702 indicates the start of processing. In step 704, the nuclear medicine image data 130 to be processed 700 is read. In step 706, an appropriate cross-sectional image is cut out from the nuclear medicine image data 130 and displayed on the display device.

  In step 708, the region including the tumor is highlighted on the cross-sectional image displayed in step 706. In order to perform this process, it is necessary to extract a region including a tumor from the nuclear medicine image data 130. This extraction process can be performed by several methods. One of them may be the process 500 itself described with reference to FIG. Therefore, in some embodiments, the bone tumor pixel group data 138 stored in step 522 of the process 500 may be read from the storage unit (eg, the mass storage device 106) in step 706. Then, from the bone tumor pixel group data 138, the pixel group located in the cross section displayed in step 706 may be displayed in an identifiable manner (for example, in a color different from the cross-sectional image displayed in step 706). .

  Another example of a technique for extracting a region including a tumor from the nuclear medicine image data 130 to be performed in step 708 includes a tumor having a pixel group having a pixel value equal to or larger than a predetermined threshold in the nuclear medicine image data 130. It is to be an area. Still another example of the technique is that the operator visually determines a region including a tumor and designates the region by, for example, surrounding the region with a mouse. Depending on the embodiment, in order to extract a region including a tumor from the nuclear medicine image data 130, any of the above methods may be implemented, and a tumor region may be extracted by a method different from these three methods. . In any case, the extracted tumor region is preferably displayed on the screen in a format that can be easily discriminated by the operator.

  FIG. 8A shows an example of the display made at step 708. As shown in the figure, on the cross-sectional image of the nuclear medicine image data 130, a region indicated by a dark color is displayed. This dark area indicated by reference numeral 802 is a tumor area. Originally, FIG. 8A is preferably a color image, and the color used for the tumor region is preferably a color that is clearly different from the color used for the cross-sectional image of the nuclear medicine image data 130. (FIG. 8A is also a color image, and the region indicated by reference numeral 802 is painted in red.) Note that this tumor region illustrated in the drawing is the process 500 described with reference to FIG. It is extracted by the embodiment. (At the same time, the physiologically accumulated site of the radiopharmaceutical is also removed by the embodiment of the process 200 described with reference to FIG. 2. This is why there is no accumulation at the site corresponding to the bladder.)

  Note that the cross-sectional image displayed in step 706 can be a cross-sectional image including a tumor region automatically extracted by the system 100. The technique for automatically extracting the tumor region can be, for example, as described above in connection with step 708. Therefore, depending on the embodiment, the processing performed in step 706 and the processing performed in step 708 may be performed in a partly reversed order (the order shown in FIG. 7A) or may be performed in an integral manner. May be broken. The system 100 is configured to automatically select, for example, a cross-section including the central coordinates of the automatically extracted tumor region or a cross-section including a pixel having the maximum pixel value in the tumor region as the cross-section to be displayed in step 706. May be. Further, the system 100 may be configured to automatically select a cross section including a tumor having the largest volume as a cross section to be displayed in step 706 when there are a plurality of automatically extracted tumor areas. In addition, when there are a plurality of automatically extracted tumor regions, the system 100 displays a list of each tumor region and allows the user to select the tumor region, and includes the tumor region in response to the user selecting a specific tumor region. It may be configured to display a cross section. Furthermore, the system 100 may be configured so that the user can freely select and change the cross-sectional image displayed in step 706. For example, a user interface element such as a scroll bar may be used to select and change freely.

  In step 710, a reference point is set within the displayed tumor region. This reference point is a reference point for pixel scanning in the next step. The reference point can be set by several methods depending on the embodiment. One of them is setting based on a selection operation by the operator. For example, when the operator clicks any point in the tumor area displayed in step 708 on the display device with the mouse, the system 100 selects the pixel of the nuclear medicine image data 130 corresponding to the clicked position. The reference point may be set. It is obvious that the same processing can be performed using other input means besides the mouse.

  In some embodiments, the system 100 may be configured to automatically set the reference point. For example, the center coordinate of the tumor region or a pixel having the maximum pixel value in the tumor region may be used as the reference point.

An embodiment is configured to perform the following processing in order to automatically set the reference point.
(1) The image data 130 is copied.
(2) In the copied image data, the tumor area is automatically extracted, and all the pixel values of the pixels other than the extracted tumor area are set to NULL. The tumor region extraction method can be, for example, as described above in connection with step 708.
(3) A cross section including a pixel having the maximum pixel value in the automatically extracted tumor region is cut out.
(4) Two end points are set in the tumor region included in the cross section. First, the first end point is a pixel having the smallest coordinate value on one axis (for example, the X axis). The second end point is a pixel having the largest coordinate value.
(5) A pixel located at an intermediate point between the first end point and the second end point is set as a reference point candidate.
(6) If the pixel value of a pixel that is a reference point candidate is not NULL, the pixel is determined as the final reference point.
(7) If the pixel value of the pixel set as the reference point candidate is NULL, the process returns to (4), and the first end point or the second end point is slightly shifted (for example, one pixel) within the tumor region, 5) Repeat the subsequent steps.

  In the process 700, where the reference point is set does not significantly affect the quality of the final tumor contour extraction result. However, it is desirable that the reference point is set within the tumor region.

  If a plurality of tumor regions are displayed in the displayed cross-sectional image, a reference point is set for each.

  An example of the reference point set in step 710 is shown in FIG. 8B. The dark region indicated by reference numeral 802 is the same as the tumor region of FIG. 8A. In this example, the reference point is set at the position of the black point indicated by reference numeral 804.

  In step 712, the nuclear medicine image data 130 is scanned three-dimensionally from the reference point as a starting point, the change in the pixel value is examined in each scanning direction, and a pixel value profile is created. This is shown in FIG. 8C. Four arrows 806 are drawn radially from the reference point 804, each of which indicates the scanning direction. Although only four scanning directions are shown in the figure, the scanning should actually be performed in all directions in a three-dimensional radial pattern. In the contour extraction example to be introduced later, in the same plane, scanning is performed in 360 ° in all directions around the reference point in 1 ° increments, and the plane to be scanned is 180 ° in 1 ° increments around the axis passing through the reference point. Rotation is performed, and omnidirectional scanning is performed on each plane to extract the contour. Therefore, 360 × 180 = 64800 pixel value profiles were created in total. Of course, these numerical values are merely examples, and it is a matter of course that different values may be used for the angle increments.

  A circle represented by reference numeral 810 represents a scanning range in a specific plane. That is, pixel values are scanned in a range from the reference point to the circle 810 in each scanning direction on the plane. By limiting the scanning range in this way, it is possible to prevent unnecessary resources from being scanned by wasting unnecessary calculation resources. In this example, pixel values are sampled at 26 points in each scanning direction until reaching the circle 810 from the reference point. The scanning range is desirably set so as to include the entire tumor area. Of course, this sampling number is merely an example.

  An example of a pixel value profile created in one of the scanning directions is shown in FIG. 8D. The value on the horizontal axis corresponds to the position of the sampling point. As described above, since sampling is performed at 26 points in this example, 26 points are shown in the figure. The value on the vertical axis is the pixel value.

  In step 714, each pixel value profile created in the previous step is smoothed. Any smoothing method may be used. For example, smoothing by the well-known three-point moving average method may be used. That is, smoothing may be performed by a method of replacing a pixel value at a certain sampling point with a pixel average value of three points including that point and an adjacent sampling point. In addition, various smoothing processes may be applied depending on the embodiment. The pixel value profile shown in FIG. 8D obtained by performing the smoothing process using the above-described three-point moving average method is shown in FIG. 8E (in FIG. 8E, the average of three points cannot be obtained for the values at both ends. Data values are left intact).

In step 716, a gradient intensity profile is created for the smoothed pixel value profile. When there is a point 1 (x ₁ , y ₁ ) and a point 2 (x ₂ , y ₂ ), the gradient strength between them is {(x ₂ -x ₁ ) ² It is defined by the square root of + (y ₂ -y ₁ ) ² }. In this step, for each point of the pixel value profile subjected to the smoothing process in the previous step, the gradient strength is calculated between the adjacent points. Therefore, in this example, a gradient intensity profile having 25 gradient intensity values is created for each pixel value profile having 26 pixel values. FIG. 8F shows a graph with an example of a (smoothed) pixel value profile and a corresponding example of a gradient intensity profile.

  In step 718, smoothing is applied to each of the gradient intensity profiles created in the previous step. As in the case of smoothing the pixel value profile, any smoothing method may be used, for example, the well-known three-point moving average method may be used. In this example as well, the gradient intensity profile was smoothed using this method.

  In step 720, a range (contour determination range) in which a tumor contour point is determined is determined in the pixel value profile. The contour determination range is performed based on a point (extreme value point) that is an extreme value in the pixel value profile (or the pixel value profile smoothed in step 714). The search for the extreme points is performed by, for example, taking a difference between adjacent points in the pixel value profile (or the pixel value profile smoothed in step 714) and changing the sign to 0 or changing the sign of positive / negative. It can be performed as an extreme point. The contour determination range can be determined, for example, from the start point of the pixel value profile (or the pixel value profile subjected to the smoothing process in step 714) to the point where the minimum value is obtained.

  An example of a contour determination range determination process that can be performed in step 720 will be introduced using FIG. 7B. Step 742 indicates the start of processing. In step 744, a search is made for a point (extreme point) at which an extreme value is obtained for the pixel value profile smoothed in step 714. In this search, as described above, for each point constituting the smoothed pixel value profile, a difference in pixel value from an adjacent point is taken and the difference becomes 0, or the sign of positive or negative is reversed. Can be used as extreme points.

  In step 746, the pixel value difference is calculated between the searched extreme points, and the maximum value is obtained.

  In step 748, a minimum value point satisfying a predetermined condition (that is, an extreme value point having a minimum value) is searched from among the plurality of extreme value points searched in step 744. The minimum value point searched here is, for example, a minimum value in which the difference in pixel value from the immediately preceding maximum value point (or maximum value point) is, for example, 20% or more of the maximum value calculated in step 746, for example. It can be a local minimum point closest to the reference point (the point set in step 710) (that is, the local minimum point having the smallest value on the horizontal axis).

  In step 750, a contour determination range is determined. The contour determination range can be, for example, a range from the start point of the pixel value profile (that is, the point closest to the reference point) to the minimum value point specified in step 748.

  FIG. 9 shows an example of the contour determination range determined in step 720.

  Note that the above-described processing for determining the contour determination range is an example. Depending on the embodiment, the contour determination range may be determined by another method. For example, in some embodiments, a point closest to the reference point is set as a start point, and a point away from the tumor area extracted in step 708 by a certain range outside (for example, a point separated by 5 pixels outside) is set as an end point. A range between the end point and the end point may be configured as the contour determination range.

  In some embodiments, for example, the local minimum point identified in step 748 is the local minimum point at which the difference in pixel values between the extreme points is greatest (ie, the local minimum point that gives the maximum value determined in step 746). ).

  In some embodiments, for example, the starting point of the contour determination range is a local maximum point that is closer to the reference point than the local minimum point specified in step 748, and the difference in pixel value from the local minimum point is determined in step 746. For example, it may be a local maximum point that is 20% or more of the maximum value calculated in (1).

  It should be noted that the specific processing for setting the contour determination range may vary depending on the embodiment. In the above description, a numerical value of 20% is used, but it is only an example, and other values may be used.

Returning to FIG. 7A, the description of the process 700 will be continued. In step 722, a tumor contour point in the pixel value profile is determined using the smoothed gradient intensity profile created in step 718 for the pixel profile whose contour determination range has been determined in step 720. This step can proceed as follows, for example.
(Sub-step 1) In the determination range set in step 720, the vertex of the smoothed gradient intensity profile is searched. That is, the extreme point that gives the maximum value is searched.
(Sub-step 2) When only one vertex is searched in sub-step 1, a point on the pixel value profile corresponding to the vertex is determined as a tumor contour point.
(Sub-step 3) When a plurality of vertices are searched in sub-step 1, the kurtosis is calculated for each vertex, and the point on the pixel value profile corresponding to the vertex having the maximum kurtosis is determined as the tumor contour point. And decide. Here, the kurtosis is a cosine of an angle (θ) between two vectors extending from a vertex (local maximum point) to an adjacent local minimum point (extreme point giving a local minimum) on the gradient intensity profile ( cos θ). That is, the direction vector from the vertex to the adjacent local minimum point is:
, Kurtosis cos (θ) is defined as follows.

  FIGS. 10A and 10B exemplify how contour points are determined. FIG. 10A corresponds to the situation of the sub-step 2 described above, and only one vertex of the gradient intensity profile is searched within the determination range. Therefore, a point on the pixel value profile corresponding to this vertex is extracted as a contour point of the tumor.

  On the other hand, FIG. 10B corresponds to the situation of the above-mentioned sub-step 3, and a plurality of vertices of the gradient intensity profile are searched within the determination range. In this case, the kurtosis is calculated for each vertex, and the vertex giving the maximum kurtosis (that is, the vertex having the maximum value of cos θ) is extracted as the contour point of the tumor.

  Note that the processing in steps 714 to 722 is performed individually for all the pixel value profiles created in step 712. For example, as described above, in order to create the contour extraction example shown in FIG. 11, a total of 64800 pixel value profiles are created. Steps 714 to 722 are individually performed for all these pixel value profiles.

  In step 724, the extracted contour points are displayed. Depending on the embodiment, the extracted contour points may be connected and displayed as a closed curve. FIG. 11 shows a display example of the extracted contour points. Depicted in this figure are contour points 1102 and contours 1104 extracted for the tumor region 802 shown in FIGS. 8A and 8B. Basically, one contour point 1102 should be determined for each of the pixel value profiles created in step 712 (it may not be determined due to an error or the like). Although only 13 contour points 1102 are drawn in the figure, of course, in practice, far more contour points 1102 are determined. A closed curve 1104 connects the extracted contour points 1102.

  Step 726 indicates the end of the process.

  According to step 700, since the outline of the tumor is automatically extracted, the variation in the extraction result by the operator is suppressed, and the tumor outline can be objectively extracted. In this embodiment, since the contour extraction is automatically performed except for selecting the reference point in step 710, the operator's labor required for the contour extraction is very small.

  The preferred embodiments of the technical idea disclosed in the present specification have been described above, but it is needless to say that the embodiments of the technical idea are not limited to those introduced. The above-described example is merely an example of an embodiment, and various other embodiments are possible.

  As an example, one usage example of the process 200 introduced using the flowchart of FIG. 2 will be introduced using FIG. The process 1200 introduced in FIG. 12 is a method for analyzing and displaying a nuclear medicine image, which is related to a method for displaying an accumulation by a bone lesion so that an operator can identify it.

  Step 1202 indicates the start of processing. In step 1204, the nuclear medicine image data 130 to be processed 1200 is read. In step 1206, the bone region data 134 stored in step 212 of the process 200 of FIG. 2 is read.

  In step 1208, an appropriate cross section is cut out from the nuclear medicine image data 130 and displayed as an image. The section to be cut out may be a section including a pixel having the maximum pixel value in the bone region, for example. In step 1210, a pixel having a pixel value equal to or higher than the threshold is extracted from the same cross-sectional image using an appropriate threshold for distinguishing accumulation due to a tumor or the like.

  In the loop shown in steps 1212 to 1218, each extracted pixel is compared with the bone region data read in step 1206 to determine whether or not the pixel overlaps the bone region (step). 1214). If it is determined that the pixel overlaps the bone region, the location corresponding to the pixel is displayed on the image displayed in step 1208 so that the operator can identify it visually. For example, it is displayed in red or a clearly clear color. On the other hand, when it is determined that the pixel does not overlap the bone region, such display is not performed, and the determination for the next pixel is performed.

  When the determination process of step 1214 is performed for all the pixels extracted in step 1210, the process 1200 exits the loop and ends (step 1220).

  The display that the process 1200 provides will be described with reference to FIG. FIG. 13A shows in white a pixel having a pixel value equal to or greater than the threshold value used in step 1210 in a cross-sectional image of the nuclear medicine image data 130. It is shown that there are pixels having pixel values equal to or greater than the threshold at locations indicated by reference numerals 1302 and 1304. However, the reason why the pixel value indicated by the reference numeral 1304 is high is due to physiological accumulation, not due to accumulation due to a tumor or the like. On the other hand, FIG. 13B shows a region having a pixel value equal to or greater than a predetermined threshold value by the process 1200 and displaying only a region that overlaps the bone region so as to be identifiable and having a pixel value equal to or greater than the threshold value. However, the display is created by the process of not displaying the region that does not overlap the bone region in an identifiable manner. Compared to FIG. 13A, the accumulation 1302 observed in the bone region is highlighted as in FIG. 13A, but the physiological accumulation 1304 is not highlighted. In other words, according to the process 1200, only the accumulation due to the bone lesion can be clearly observed without the influence of physiological accumulation.

  In FIG. 13, white is used for highlighting, but this is not appropriate in nature. This is because white is also used to draw other areas. In practice, it is desirable to use color for this highlighting. The reason why the color is not used in FIG. 13 or other drawings of the present application is only due to the limitation that the color cannot be used in the drawings of the patent application.

  Various other embodiments exist in the technical idea disclosed in this specification.

  Individual features included in the various embodiments introduced in the specification or drawings may only be used in conjunction with the embodiments where it is directly described that the features are included, and other features described herein. The present invention can also be used in combination in the embodiments and various embodiments not described. In particular, the order of the processes introduced in the flowchart does not necessarily have to be executed in the order in which they are introduced. Depending on the preference of what is to be implemented, the order may be changed or executed concurrently in parallel. May be implemented inseparably, or may be implemented as an appropriate loop. All of these variations are included in the scope of the invention disclosed in the present application, and the scope of the invention is not limited by the implementation of processing. The description order of the processes specified in the claims does not necessarily specify the essential order of the processes. For example, an embodiment in which the order of the processes is different, an embodiment in which the processes are executed including a loop, etc. Is also included in the scope of the claimed invention. Regardless of whether the current claims are claimed or not, the applicant claims to have the right to obtain a patent for all forms that do not depart from the spirit of the invention. Keep it.

100 system 102 CPU
104 Main storage device 106 Mass storage device 107 Display interface 108 Peripheral device interface 109 Network interface 110 Operating system 120 Nuclear medicine image analysis program 130-138 Data

Claims (8)

A computer program for analyzing nuclear medicine image data, the computer program being executed by a processing means of the system,
Obtaining a pixel group of a region corresponding to a bone from a pixel group included in the nuclear medicine image data;
Obtaining a standard deviation based on at least some pixel values of the acquired pixel group;
Extracting a bone tumor pixel group based on the standard deviation;
Comprising a composed program instructions so as to perform,
Further, the computer program is executed by the processing means, so that the system
Creating a histogram based on the pixel values of the acquired pixel group;
Identifying a peak value of the histogram;
Determining, in the histogram, two class values whose frequency is a predetermined percentage of the peak value;
Wherein at least a part of the acquired pixel group is a group of pixels having pixel values included in an interval determined by the two class values. A computer program.   The extracting a bone tumor pixel group based on the standard deviation includes extracting a bone tumor pixel group based on the larger one of the two class values and the standard deviation. The computer program described.   Extracting a bone tumor pixel group based on the standard deviation is based on an average value calculated based on the at least some pixel values of the acquired pixel group and the standard deviation. The computer program according to claim 1, comprising extracting a pixel group. The computer program is executed by the processing means of the system, so that the system
Displaying the extracted bone tumor region in an identifiable manner on the nuclear medicine image data;
Further configured program instructions so as to perform the provided computer program according to any one of claims 1 to 3. Acquiring a pixel group in an area corresponding to the bone
Extracting a bone region from the CT image data aligned with the nuclear medicine image data;
Extracting a group of pixels in the nuclear medicine image data in a region overlapping the extracted bone region;
The containing computer program according to any one of claims 1 to 4. Processing means;
Storage means for storing the computer program according to any one of claims 1 to 5 ;
Having a system. A method of analyzing nuclear medicine image data performed by the processing means of the system performed by the processing means of the system:
Obtaining a pixel group of a region corresponding to a bone from a pixel group included in the nuclear medicine image data;
Obtaining a standard deviation based on at least some pixel values of the acquired pixel group;
Extracting a bone tumor pixel group based on the standard deviation;
Along with including,
Creating a histogram based on the pixel values of the acquired pixel group;
Identifying a peak value of the histogram;
Determining, in the histogram, two class values whose frequency is a predetermined percentage of the peak value;
Wherein the at least part of the acquired pixel group is a group of pixels having pixel values included in a section determined by the two class values .   A system for analyzing nuclear medicine image data,
  Means for acquiring a pixel group in a region corresponding to a bone from a pixel group included in the nuclear medicine image data;
  Means for obtaining a standard deviation based on at least some of the pixel values of the acquired pixel group;
  Means for extracting a bone tumor pixel group based on the standard deviation;
With
  Means for creating a histogram based on the pixel values of the acquired pixel group;
  Means for identifying a peak value of the histogram;
  Means for determining, in the histogram, two class values whose frequency is a predetermined ratio of the peak value;
Wherein the at least part of the acquired pixel group is a group of pixels having pixel values included in a section determined by the two class values.