Simple CT image diagnostic support system

Abstract

The problem addressed by the present invention is to provide an aid system that makes it possible to accurately perform diagnosis based on plain CT imaging even when the given CT value difference is small in cases such as immediately after the onset of a disease.　This plain CT imaging aid system is characterized by comprising: (1) a means for obtaining a plain CT image; and (2) a means for reconstructing the CT image by nonlinearly transforming the CT values of pixels of the CT image.

Description

[Technical Field]

[0001] The present invention relates to a simple CT image diagnostic assistance system and a simple CT image diagnostic assistance method. [Background technology]

[0002] MRI is the standard test for diagnosing conditions such as cerebral infarction, but demand for the test is high and it is difficult to make an appointment. Furthermore, because the examination takes a long time, examination rooms are often unavailable and even acute illnesses cannot be examined immediately. In such cases, a plain CT scan, which does not use contrast agents, may be performed first. However, a definitive diagnosis of acute cerebral infarction is rarely made with a plain CT scan alone; the diagnosis is confirmed by the results of an MRI. For example, in the case of hyperacute cerebral infarction, prompt treatment has a significant impact on the patient's quality of life.

[0003] For example, early CT signs and hyperdense MCA signs are crucial for diagnosing hyperacute cerebral infarction. However, the difference in CT values ​​between early CT signs and normal areas immediately after onset is only about 2-10, and even experienced doctors often have difficulty making a diagnosis. Summary of the Invention [Problem to be solved by the invention]

[0004] Therefore, in simple CT image diagnosis, an auxiliary system that can accurately diagnose even with slight differences in CT values, such as immediately after the onset of symptoms, is desired. [Means for solving the problem]

[0005] Therefore, the inventors have tried various image processing methods to clarify CT images, and have found, quite unexpectedly, that if the CT value of each pixel in a CT image is raised to a power and the CT image is reconstructed based on the CT value obtained by raising the power, slight differences in CT values, such as those immediately after onset, are displayed as clear differences, thereby assisting simple CT image diagnosis, and have completed the present invention.

[0006] That is, the present invention provides the following [1] to

[14] . [1] A simple CT image diagnostic support system comprising: (1) a means for obtaining a simple CT image; and (2) a means for reconstructing a CT image by nonlinearly transforming the CT value of each pixel in the CT image. [2] The simple image diagnosis support system according to [1], wherein the nonlinear transformation means is a means for reconstructing a CT image by raising the CT value of each pixel in the CT image to a power. [3] The simple CT image diagnostic support system according to [1] or [2], further comprising a means for displaying the reconstructed CT image as a color map. [4] A simple CT image diagnostic support system according to any one of [1] to [3], wherein, when the CT values ​​are nonlinearly converted, at least one of the following preprocessing steps is performed: replacing CT values ​​below 0 with values ​​of 0 or close to 0; replacing CT values ​​that would overflow due to nonlinear conversion with values ​​that do not overflow; or uniformly shifting all CT values ​​to a range that does not overflow. [5] A simple CT image diagnostic support system according to any one of [1] to [4], wherein the simple CT image is a simple CT image of a patient selected from hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and detailed examination of the vascular wall of a patient with arterial dissection. [6] The plain CT image diagnostic support system according to any one of [1] to [4], wherein the plain CT image is a plain CT image of a patient selected from hyperacute cerebral infarction and brain tumor. [7] A simple CT image diagnosis support system according to any one of [1] to [6], comprising: an artificial intelligence constructed by learning using disease information corresponding to a simple CT image as teacher data and CT images reconstructed by the reconstructing means as training data; and a prediction means for reconstructing the simple CT image obtained by the obtaining means using the reconstructing means and inputting the reconstructed image into the artificial intelligence, thereby predicting disease information corresponding to the simple CT image. [8] A method for assisting diagnosis using simple CT images, comprising: (1a) obtaining a simple CT image; and (2a) performing nonlinear transformation on the CT value of each pixel in the CT image to reconstruct the CT image. [9] The simple image diagnosis support method according to [8], wherein the nonlinear conversion step is a step of reconstructing a CT image by raising the CT value of each pixel in the CT image to a power.

[10] The method for assisting diagnosis with simple CT images according to [8] or [9], further comprising a means for displaying the reconstructed CT image as a color map.

[11] A simple CT image diagnostic assistance method according to any one of [8] to

[10] , wherein, when the CT values ​​are nonlinearly converted, at least one of the following preprocessing steps is performed: replacing CT values ​​of 0 or less with values ​​of 0 or close to 0; replacing CT values ​​that would overflow due to nonlinear conversion with values ​​that do not overflow; or uniformly shifting all CT values ​​to a range that does not overflow.

[12] The method for assisting diagnosis with simple CT images according to any one of [8] to

[11] , wherein the simple CT images are those of a patient selected from those of patients with hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and detailed examination of the vascular wall of a patient with arterial dissection.

[13] The plain CT image diagnostic support system according to any one of [8] to

[11] , wherein the plain CT image is a plain CT image of a patient selected from hyperacute cerebral infarction and brain tumor.

[14] A simple CT image diagnosis assistance method according to any one of [8] to

[13] , comprising: an artificial intelligence constructed by learning using disease information corresponding to a simple CT image as teacher data and the CT image reconstructed by the reconstructing process as training data; and a prediction process for predicting disease information corresponding to the simple CT image by reconstructing the simple CT image obtained by the obtaining process in the reconstructing process and inputting the reconstructed image into the artificial intelligence. [Effects of the Invention]

[0007] The simple CT image diagnostic support system and method of the present invention can clearly display slight differences in CT values ​​in simple CT images of hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumors, brain metastasis of cancer, liver metastasis of cancer, and detailed examination of vascular walls in patients with arterial dissection, thereby enabling early detection of these diseases and appropriate treatment. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 10 is a diagram showing CT values ​​in an original CT image and CT values ​​after exponentiation. [Figure 2] The original CT image, the CT image after exponentiating the CT value, and the CT image in which the exponentiated CT image is displayed in a color map are shown. [Figure 3] FIG. 1 is a diagram illustrating an example of the configuration of a simple CT image diagnosis support system. DETAILED DESCRIPTION OF THE INVENTION

[0009] The present invention is a simple CT image diagnostic support system and a simple CT image diagnostic support method. A simple CT (Computed Tomography) image is created without the use of contrast agents by detecting transmitted X-rays irradiated from various angles, collecting data, and then using a computer to visualize the collected data using back projection. In other words, X-rays are irradiated and an image is displayed based on the degree of absorption of the X-rays. Specifically, areas with high X-ray absorption (high absorption areas) appear white on the image (bone, calcification, blood clots, metals, etc.), while areas with low absorption (low absorption areas) appear black on the image (cerebrospinal fluid (ventricles), infarction, fat, etc.). The presence and size of infarction can be diagnosed based on the shade of this color. The system and method for assisting diagnosis with simple CT images of the present invention are a system and a method for clarifying the color shading in the CT images to assist diagnosis.

[0010] One aspect of the present invention is a simple CT image diagnostic support system comprising: (1) a means for acquiring a simple CT image; and (2) a means for reconstructing a CT image by nonlinearly transforming the CT value of each pixel in the CT image. Another aspect of the present invention is a simple CT image diagnostic support method, characterized by comprising: (1a) a step of obtaining a simple CT image; and (2a) a step of nonlinearly converting the CT value of each pixel in the CT image to reconstruct the CT image. The above means and steps will be collectively described below.

[0011] The means (1) (step (1a)) is a means (method) for obtaining a plain CT image. Means (1) is a means for obtaining CT images using a conventional CT device, in which the radiation source and detector rotate around the object to be examined, the object receives X-rays from all directions, the irradiated X-rays pass through the object, are partially absorbed and attenuated by the object, and then reach the X-ray detector located on the opposite side of the radiation source and are recorded. After recording the degree of absorption in each direction, the image can be obtained by reconstructing it using a Fourier transform on a computer.

[0012] The means (2) (step (2a)) is a means (step) for nonlinearly transforming the CT value of each pixel in the CT image to reconstruct the CT image. The unit of CT value is Hounsfield unit (hereafter abbreviated as HU). CT value is expressed as a relative value, with air being -1000 HU and water being 0 HU. Negative values ​​(0 to -100) represent fat. The original CT value is converted into a nonlinear value by nonlinear transformation. The nonlinear transformation can be in the form of a monotonically increasing function (however, in implementation, a conversion table or the like can be used instead of a function). This maintains the magnitude relationship between the values ​​before and after transformation. Furthermore, the larger the original value, the larger the function (a function whose first derivative is monotonically increasing) can be selected, which increases the magnitude of "transformed value - original value." A specific example is a function that performs exponentiation (a function that multiplies a number X by X n n may be an integer of 2 or more, or a real number greater than 1), exponential function (applying a to a number X) X The following calculations are performed: a is a real number greater than or equal to 1. In the present invention, it is preferable to employ a means for exponentiating the CT value of each pixel in a CT image.

[0013] Figure 1 shows an example of exponentiating the CT value of each pixel in a CT image. For example, an original image CT value of 30 is converted to 900, and an original image CT value of 35 is converted to 1225. Here, the power is usually a square, but it may also be a cube if necessary.

[0014] Many DICOM CT images are 16-bit, so exponentiation can result in overflow. Also, raising a negative CT value to a power can sometimes result in a positive number. Because the range of unsigned 16-bit gradation is 0 to 65,535, squaring the original CT value can cause overflow when the original CT value is 256 or greater. Because the range of signed 16-bit gradation is -32,768 to 32,767, squaring the original CT value can cause overflow when the original CT value is 182 or greater. Therefore, when performing the nonlinear conversion of the CT values, it is preferable to perform at least one of the following preprocessing steps: replacing CT values ​​of 0 or less with a value of 0 or close to 0 (any value that does not affect the diagnosis of the area not to be replaced, for example, 1 or less); replacing CT values ​​that overflow due to the nonlinear conversion with a value that does not overflow (any value that does not affect the diagnosis of the area not to be replaced, for example, a value that does not overflow but is sufficiently large, or 0 or close to 0); or uniformly shifting all CT values ​​to a range that does not overflow (which can be achieved by adding or subtracting a fixed value). Specifically, when raising the CT value to a power, it is preferable to replace CT values ​​below 0 with 0 as necessary, and replace CT values ​​that overflow when raised to a power with 0, or if the target overflows when raised to a power, add or subtract from all CT values ​​and shift them within the target.

[0015] A CT image reconstructed by this procedure is shown in FIG. As shown in Figure 2, the reconstructed CT image can clearly display slight differences in CT values ​​compared to the original CT image.

[0016] The simple CT image diagnosis support system (method) of the present invention may further include a means (step) for displaying the reconstructed CT image as a color map. The color map display may be implemented as a function of a workstation such as VINCENT or ZIOstation. As shown in Figure 2, color map display makes it possible to more clearly display the color shading in CT images, allowing for more accurate diagnosis.

[0017] The system of the present invention can be a simple CT image diagnosis support system having: an artificial intelligence constructed by learning using disease information corresponding to a simple CT image as teacher data and the CT image reconstructed by the reconstructing means as training data; and a prediction means for predicting disease information corresponding to the simple CT image by reconstructing the simple CT image obtained by the obtaining means using the reconstructing means and inputting the reconstructed image into the artificial intelligence. The system of the present invention can also be realized by, for example, having one or more processors, which perform various necessary processes such as the construction by learning described above, reconstruction of simple CT images, and prediction of disease information. Furthermore, the method of the present invention can be a simple CT image diagnosis assistance method comprising: an artificial intelligence constructed by learning using disease information corresponding to the simple CT image as teacher data and the CT image reconstructed in the reconstructing step as training data; and a prediction step of predicting disease information corresponding to the simple CT image by reconstructing the simple CT image obtained in the obtaining step in the reconstructing step and inputting the reconstructed image into the artificial intelligence.

[0018] A specific example of the simple CT image diagnostic system of the present invention will be described using a block diagram. FIG. 3 is a block diagram showing an example of the configuration of the simple CT image diagnosis support system 10. As shown in FIG. The simple CT image diagnosis support system 10 can be implemented using a computer such as a PC or a workstation. The computer includes an arithmetic unit such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), a storage device such as a memory, and an input / output device such as a touch panel display. The simple CT image diagnosis support system 10 is constructed by controlling the computer with software.

[0019] The CT image diagnosis support system 10 includes a control unit 11, a power processing unit 12, a preprocessing unit 14, a reconstruction processing unit 16, a color map display unit 18, and an artificial intelligence (AI) 20. The control unit 11 controls the operation of the CT image diagnosis support system 10 based on user input. Under the control of the control unit 11, a plain CT image is obtained. The power processing unit 12 is an example of a processing unit (nonlinear transformation processing unit) that performs nonlinear transformation, such as exponentiation, of the CT values ​​of the input and output simple CT images. Prior to the power processing, a preprocessing unit 14 is performed. This preprocessing is performed to prevent the exponentiated CT image from being misleading to the user or to prevent the display of the exponentiated CT image from appearing unnatural or abnormal. Specifically, preprocessing can be performed to replace CT values ​​below 0 with values ​​of 0 or near 0, to replace CT values ​​that would overflow when powered with a value of 0 that does not overflow, or to shift all CT values ​​to a range that does not overflow. The reconstruction processing unit 16 reconstructs a CT image based on the exponentiated CT values. The color map display unit 18 displays the reconstructed CT image as a color image by comparing it with a color map.

[0020] The artificial intelligence 20 includes a model 22, a prediction unit 24, and a learning processing unit 26. The model 22 may be, for example, a neural network, but is not limited to this. The model 22 is trained by the learning processing unit 26. The learning processing unit 26 trains parameters of the model 22 using training data and corresponding teacher data. By using reconstructed CT images as the training data and disease information corresponding to these CT images as the teacher data, the model 22 can predict disease information from the reconstructed CT images. The prediction unit 24 inputs the reconstructed CT images into the trained model 22, thereby predicting disease information corresponding to these CT images. Examples of disease information to be predicted include hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and patients with arterial dissection.

[0021] The simple CT image diagnostic support system (method) of the present invention is particularly useful for assisting in the diagnosis of diseases that require diagnosis based on slight differences in CT values. For example, it is useful for simple CT image diagnosis of patients selected from hyperacute cerebral infarction, identification of the source of bleeding in subarachnoid hemorrhage, the extent of brain tumor spread, brain metastasis of cancer, liver metastasis of cancer, and detailed examination of the vascular wall in patients with arterial dissection. Furthermore, it is useful for simple CT image diagnosis of patients selected from hyperacute cerebral infarction and brain tumor. [Example]

[0022] The present invention will now be described in more detail with reference to examples, but the present invention is not limited to these examples.

[0023] Example 1 Each pixel data of the CT image is read. If the CT value of the area to be analyzed is exponentiated and the 16-bit gradation overflows, the same value is added or subtracted from all pixels before the exponentiation. If the loaded pixel data type is signed 16-bit gradation, pixel values ​​with a CT value of 182 or greater are replaced with 182; if the loaded pixel data type is unsigned 16-bit gradation, pixel values ​​with a CT value of 256 or greater are replaced with 256. In addition, pixel values ​​with a CT value of 0 or less are replaced with 0. When displaying images of pixel data that has been exponentiated, a window level and window width are set, just like in normal CT. The window level indicates the center position of the gradation for image display, and the window width is a method of displaying an area that is half the upper and lower window widths centered on the window level. For example, if the window level is 100 and the window width is 80, the area displayed will be one with a CT value between 60 and 140. CT values ​​below the lower limit are displayed as the lowest density, and CT values ​​above the upper limit are displayed as the highest density. When diagnosis is difficult using grayscale, color map display is expected to further improve diagnostic ability. If the displayed image is rough, it is effective to use a smoothing filter.

[0024] (1) Obtaining plain CT images We use CT images taken on patients suspected of hyperacute cerebral infarction due to symptoms such as dysarthria, motor disorders, and impaired consciousness. The CT devices used were Canon Aquilion ONE, Canon Aquilion Lightning, Siemens Definishon Flash, etc., and head CT scans were performed using these devices.

[0025] (2) Raise the CT value of each pixel in the CT image The array on the left in Figure 1 is the pixel values ​​of the CT image. The array on the right is the array obtained by exponentiating the pixel data on the left. For example, raising the data on the left, 30, to a power (in this case, squaring) makes the data 900. Similarly, raising the data to the power of 35 makes the value 1225. Before the power was raised, the signal value difference was 5, but after the power is raised, the signal value difference becomes 325. For example, if the signal value of normal tissue is 35 and the signal value of cerebral infarction lesion is 30, the data before exponentiation can only be displayed as a difference of at most 5 gradations, but the data after exponentiation can be displayed as a difference of at most 325 gradations.

[0026] (3) Reconstruct the CT image based on the CT values ​​obtained by exponentiation. Figure 2 shows the data from Figure 1 displayed as a CT image. The image in the upper left is a diffusion-weighted MRI image of the right frontal lobe. The white area indicates the lesion site of acute cerebral infarction. In the CT images, there is no noticeable difference when compared to the left brain. However, in the power image of the center image, a decrease in signal value in the cortex of the right brain can be clearly observed. In addition, the lentiform nucleus is also more clearly blurred compared to the left brain. The image on the far right is a color map representation of the power image. A color map is a color palette that varies in ratio from specific color to specific color, and assigns colors to values. In both the power image and the color map display of the power image, early CT signs are clearly observable. CT image Window level 40 Window width 120 Power image Window level 1200 Window width 600 Power image color map Window level 1200 Window width 600 [Explanation of symbols]

[0027] 10 Simple CT image diagnostic support system 11 Control section 12 Power processing section 14 Pretreatment section 16 Reconstruction processing unit 18 Color map display 20. Artificial Intelligence 22 models 24 Prediction Department 26 Learning processing unit

Claims

1. A simple CT image diagnostic support system comprising: (1) a means for acquiring a simple CT image; and (2) a means for reconstructing a CT image by squaring the CT value of each pixel in the CT image, A simple CT image diagnostic support system that performs at least one of the following preprocessing steps when squaring the CT values: replacing CT values ​​below 0 with values ​​of 0 or close to 0; replacing CT values ​​that would overflow when squared with values ​​that do not overflow; or uniformly shifting all CT values ​​to a range that does not overflow.

2. 2. The simple CT image diagnostic support system according to claim 1, further comprising means for displaying the reconstructed CT image in a color map.

3. 2. The simple CT image diagnostic support system according to claim 1, wherein the simple CT image is a simple CT image of a patient selected from patients with hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and detailed examination of vascular walls of patients with arterial dissection.

4. 2. The plain CT image diagnostic support system according to claim 1, wherein the plain CT image is a plain CT image of a patient selected from the group consisting of hyperacute cerebral infarction and brain tumor.

5. 5. The simple CT image diagnosis support system according to claim 1, comprising: an artificial intelligence constructed by learning using disease information corresponding to the simple CT image as teacher data and the CT image reconstructed by said reconstructing means as training data; and a prediction means for predicting disease information corresponding to the simple CT image by reconstructing the simple CT image obtained by said obtaining means using said reconstructing means and inputting the reconstructed image to said artificial intelligence.

6. 1. A method for assisting diagnosis using simple CT images, comprising: (1a) obtaining a simple CT image; and (2a) squaring the CT value of each pixel in the CT image to reconstruct the CT image, A simple CT image diagnostic assistance method that performs at least one of preprocessing steps when squaring the CT values: replacing CT values ​​below 0 with values ​​of 0 or close to 0; replacing CT values ​​that overflow when squared with values ​​that do not overflow; or uniformly shifting all CT values ​​to a range that does not overflow.

7. 7. The method for assisting diagnosis with simple CT images according to claim 6, wherein the simple CT images are those of a patient selected from those for detailed examination of vascular walls of patients with hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and arterial dissection.

8. 7. The method for assisting diagnosis with simple CT images according to claim 6, wherein the simple CT images are those of a patient selected from those for detailed examination of vascular walls of patients with hyperacute cerebral infarction, subarachnoid hemorrhage, brain tumor, brain metastasis of cancer, liver metastasis of cancer, and arterial dissection.

9. 7. The plain CT image diagnostic support system according to claim 6, wherein the plain CT image is a plain CT image of a patient selected from the group consisting of hyperacute cerebral infarction and brain tumor.

10. an artificial intelligence constructed by learning using disease information corresponding to the plain CT image as teacher data and the CT image reconstructed in the reconstructing step as training data; and a prediction step of reconstructing the plain CT image obtained in the obtaining step in the reconstructing step and inputting the reconstructed image into the artificial intelligence to predict disease information corresponding to the plain CT image; The simple CT image diagnosis auxiliary method according to any one of claims 6 to 9, comprising:

Images