Method, device and recording medium for providing low-dose training images
The method addresses the challenge of obtaining high-quality dental CBCT images at ultra-low doses by using AI-enhanced noise filtering and degradation techniques, resulting in efficient and effective image enhancement.
Patent Information
- Application Number
- KR1020240080909
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2024-06-21
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2044-06-21
AI Technical Summary
Existing dental CBCT imaging systems face challenges in obtaining high-quality images at ultra-low doses due to artifacts and dose management issues, limiting the effectiveness of dose reduction methods and data acquisition.
A method and device utilizing AI deep learning to process normal dose learning images, involving noise filtering, degradation, and enhancement learning models to convert ultra-low dose images into high-quality normal dose images, with techniques like blurring and dilation to improve image quality.
Enables the acquisition of high-quality images at ultra-low doses with improved efficiency by enhancing image quality through AI processing, allowing for larger quantities of usable images without increasing radiation exposure.
Smart Images

Figure 112024067217676-PAT00002_ABST
Abstract
Description
Technology Field
[0001] The present disclosure relates to a technology that enables the acquisition of a normal dose learning image based on a low dose learning image corresponding to an ultra-low dose CBCT. Background Technology
[0002] Artifacts (defects in an image resulting from improper image sampling) occur due to various factors during the process of acquiring Computerized Tomography (CT) images. The readability of CT images can be severely degraded by high-density objects within the subject. Objects with a density significantly higher than surrounding body tissues possess a relatively very large attenuation coefficient, which can cause beam hardening, beam scattering, starburst, and streak artifacts in the entire CT image as well as the surrounding area. Furthermore, dose management is becoming increasingly important in Dental CBCT due to the risk of patient exposure to CT (X-ray) imaging, and there is a need for methods to obtain high-quality images with lower radiation doses. While the general method for low-dose CBCT imaging is to reduce mA, Dental CBCT does not have a high mA compared to Medical CT, so there is a limit to the dose reduction achieved by lowering mA to achieve low-dose imaging. If a faster scan time is used to reduce mA for lowering doses, the dose exposure to the patient can be reduced more significantly than with conventional low-dose imaging. Efficiency can be improved if target images (images taken at normal or high dose) can be obtained using such ultra-low dose imaging. However, imaging conditions for ultra-low dose CBCT images do not exist in existing equipment, making it difficult to acquire them. Furthermore, even if humans are imaged under ultra-low dose conditions on equipment in the development stage, the number of data points that can be acquired is small, making it difficult to expect good results.Therefore, there is a need for a method to improve the quality of ultra-low dose CBCT images using AI (deep learning) with existing equipment. Prior art literature
[0003] Korean Registered Patent No. 10-2438068 (August 25, 2022) Method and apparatus for improving image quality using a non-local global strain reduction technique based on mutual information. The problem to be solved
[0004] One embodiment of the present disclosure aims to solve the problems of the aforementioned prior art and to provide a method, device, and recording medium that improve the quality of ultra-low dose CBCT images and enable the acquisition of normal dose learning images based on low dose learning images.
[0005] The technical problem to be solved is not limited to the technical problem described above, and may include various other technical problems within the scope obvious to a person skilled in the art. means of solving the problem
[0006] A method for a device according to a first aspect of the present disclosure to provide a low dose learning image may include: a receiving unit acquiring a normal dose learning image for a first object; a processor acquiring a denoise image by filtering noise from the normal dose learning image; the processor acquiring a noise component for the normal dose learning image based on a difference value between the normal dose learning image and the denoise image; the processor acquiring a degraded image with reduced sharpness for the normal dose learning image; and the processor providing a low dose learning image obtained by applying the noise component to the degraded image.
[0007] Additionally, the step of acquiring the degradation image may involve the processor performing blurring and / or dilation on the normal dose learning image to acquire the degradation image.
[0008] Additionally, the method may further include the step of the processor learning the normal dose learning image and the low dose learning image to obtain an enhancement learning model that acquires the normal dose learning image based on the low dose learning image.
[0009] Additionally, the method may further include the step of a receiver acquiring a low dose image of a second target; and the step of the processor acquiring a normal dose image of the second target by applying the enhancement learning model to the low dose image.
[0010] Additionally, the step of acquiring the low dose image may acquire the low dose image according to a frame rate per second higher than the frame rate per second (FPS) required when the processor acquires the normal dose image.
[0011] In addition, the step of acquiring the low dose image may acquire the low dose image according to a scan time shorter than the scan time required when the processor acquires the normal dose image.
[0012] In addition, the step of acquiring the low dose images may acquire the same number of low dose images as the number of images acquired when the processor acquires the normal dose images.
[0013] Additionally, the step of providing the raw dose learning image allows the processor to obtain a degraded noise component in which the contour component is reduced from the noise component.
[0014] Additionally, the step of providing the low dose learning image may involve the processor applying the degradation noise component to the degradation image to obtain the low dose learning image.
[0015] A device for providing a low dose learning image according to a second aspect of the present disclosure may include: a receiver for acquiring a normal dose learning image for a first object; and a processor for acquiring a denoise image by filtering noise from the normal dose learning image, acquiring a noise component for the normal dose learning image based on a difference value between the normal dose learning image and the denoise image, acquiring a degraded image with reduced sharpness for the normal dose learning image, and providing a low dose learning image obtained by applying the noise component to the degraded image.
[0016] In addition, the processor can obtain the degraded image by performing blurring and / or dilation on the normal dose learning image.
[0017] In addition, the processor can learn the normal dose learning image and the low dose learning image to obtain an enhanced learning model that acquires the normal dose learning image based on the low dose learning image.
[0018] In addition, the receiver can acquire a low dose image for the second target, and the processor can acquire a normal dose image for the second target by applying the enhancement learning model to the low dose image.
[0019] Additionally, the processor can acquire the low dose image according to a frame rate per second higher than the frame rate per second (FPS) required to acquire the normal dose image, acquire the low dose image according to a scan time shorter than the scan time required to acquire the normal dose image, and acquire the same number of low dose images as the number of images acquired when acquiring the normal dose image.
[0020] A third aspect of the present disclosure may provide a computer-readable recording medium storing a program for executing the method according to the first aspect on a computer. Alternatively, a fourth aspect of the present disclosure may provide a computer program stored on a recording medium for implementing the method according to the first aspect. Effects of the invention
[0021] According to one embodiment, there is an advantage in that images of ultra-low dose CBCT can be obtained easily and in large quantities compared to conventional methods.
[0022] In addition, there is an effect that images of improved quality can be obtained from ultra-low dose CBCT through an AI (deep learning) method using the acquired ultra-low dose CBCT images.
[0023] The effects of the present disclosure are not limited to the effects described above, and should be understood to include all effects that can be inferred from the detailed description of the present disclosure or the configuration of the disclosure as described in the claims. Brief explanation of the drawing
[0024] FIG. 1 is a block diagram showing an example of the configuration of a device according to one embodiment. FIG. 2 is a flowchart illustrating each step in which a device according to one embodiment provides a low dose learning image. FIG. 3 is a diagram showing an example in which a device according to one embodiment acquires a noise component for a normal dose learning image. FIG. 4 is a diagram illustrating an example in which a device according to one embodiment obtains a degraded noise component in which the contour component is reduced from the noise component. FIG. 5 is a diagram illustrating an example in which a device according to one embodiment acquires a degradation image for a normal dose learning image. FIG. 6 is a diagram illustrating an example in which a device according to one embodiment applies a noise component to a degrade image to obtain a low dose learning image. FIG. 7 is a diagram illustrating an example in which a device according to one embodiment applies an enhancement learning model to a low dose image to acquire a normal dose image. FIG. 8 is an indicator schematically showing the efficiency when a device according to one embodiment acquires a normal dose image by applying an enhancement learning model to a low dose image. Specific details for implementing the invention
[0025] The terms used in the embodiments have been selected to be as widely used and general as possible, taking into account their functions; however, these may vary depending on the intentions of prior users in the field, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been arbitrarily selected by the applicant, and in such cases, their meanings will be described in detail in the relevant description of the invention. Therefore, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the overall content of this disclosure.
[0026] When a part of the specification is described as “comprising” a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as “…part” as used in the specification refer to a unit that processes at least one function or operation, which may be implemented in hardware or software, or as a combination of hardware and software.
[0027] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein.
[0028] Throughout the specification, the term “region” may be interpreted as a concept that includes both two-dimensional and three-dimensional dimensions.
[0029] A plurality of embodiments will be described in detail below with reference to the drawings.
[0031] FIG. 1 is a block diagram showing an example of the configuration of a device (100) according to one embodiment.
[0032] Referring to FIG. 1, the device (100) may include a receiver (110), a processor (120), and a display (130).
[0033] A person skilled in the art will understand that, in addition to the components shown in FIG. 1, other general-purpose components may be further included in the device (100). For example, the device (100) may further include a memory (not shown). Or, according to other embodiments, a person skilled in the art will understand that some of the components shown in FIG. 1 may be omitted.
[0034] A receiver (110) according to one embodiment can acquire a normal dose learning image for a first object. In one embodiment, the normal dose learning image may be a CT image acquired through CBCT imaging of the oral cavity. The normal dose learning image may include 2D data and 3D data. That is, the CT image may be a concept that includes multiple images. The normal dose learning image may be a series of group images for generating an enhancement learning model to acquire a normal dose image through ultra-low dose CBCT imaging.
[0035] A processor (120) according to one embodiment can obtain a denoised image by filtering noise from a normal dose learning image. Additionally, the processor (120) can obtain a noise component for a normal dose learning image based on the difference value between the normal dose learning image and the denoised image. Additionally, the processor (120) can obtain a degraded image with reduced sharpness for a normal dose learning image. Additionally, the processor (120) can provide a low dose learning image obtained by applying a noise component to the degraded image.
[0036] In one embodiment, “providing” may include an example of a UI displayed on a user terminal. That is, although the processor (120) is shown as an example of providing a raw dose learning image, it is not limited thereto and may be provided through a display (130).
[0037] The display (130) can display a normal dose image, a low dose learning image obtained from a normal dose learning image, and / or a low dose image in 2D or 3D, and can display a normal dose image predicted to correspond to the low dose image through an enhancement learning model. Referring to FIG. 1, a wireless terminal (140) or a wired terminal (150) may be an example of a device corresponding to a server, device, camera, scanner, etc., that provides scan data for a patient's oral cavity. That is, a low dose image can be processed through the device (100) according to the present invention.
[0038] The embodiments described above with reference to FIG. 1 will be explained in more detail with reference to FIG. 2 to FIG. 8.
[0039] FIG. 2 is a flowchart illustrating each step in which a device (100) according to one embodiment provides a low dose learning image.
[0040] In step S210, the device (100) according to one embodiment acquires a normal dose learning image for a first object. In one embodiment, the normal dose image may represent an oral image of a patient acquired through a conventional dental CBCT imaging device. The normal dose learning image may be a plurality of candidate images to be used to acquire a low dose learning image. The first object may refer to a target object corresponding to an oral image representing a normal dose learning image, which is one of the plurality of candidate images. That is, the device (100) may acquire normal dose learning images corresponding to a plurality of first objects.
[0041] In step S220, the device (100) according to one embodiment filters noise from a normal dose learning image to obtain a denoised image. In one embodiment, the device (100) can filter noise from the normal dose learning image using a denoising deep learning algorithm, which is a known prior art. For example, noise filtering can be performed using a pre-trained AI image upscaler Real-ESRGAN. Thus, the device (100) can obtain a denoised image with the noise filtered.
[0042] In step S230, the device (100) according to one embodiment acquires a noise component for the normal dose learning image based on the difference value between the normal dose learning image and the denoise image. The device (100) can acquire a noise component for all areas based on the difference value of sharpness between the normal dose learning image and the denoise image in which noise has been filtered. That is, the noise component may be a noise component generated in the normal dose learning image representing an oral image captured for the first object.
[0043] In step S240, the device (100) according to one embodiment obtains a degraded image with reduced sharpness for a normal dose learning image. In one embodiment, the device (100) may obtain a degraded image by performing blurring and / or dilation on the normal dose learning image. The device (100) may obtain a degraded image by performing blurring to blur the normal dose learning image or performing dilation to expand the image area. The device (100) may obtain a degraded image in which at least one area of the normal dose learning image containing original data is blurred. Accordingly, the device (100) may obtain a degraded image from a normal dose learning image corresponding to a normal dose CBCT image that exhibits a level of image quality (sharpness) corresponding to an ultra-low dose CBCT image.
[0044] In step S250, the device (100) according to one embodiment provides a low dose learning image obtained by applying a noise component to a degrade image. The device (100) may apply the noise component obtained in step S230 to the degrade image, which exhibits a quality level corresponding to the ultra-low dose CBCT image obtained in step S240. Since the noise component of the normal dose learning image can be removed together with the blurring or dilation performed to reduce sharpness in step S240, the noise component previously obtained for the normal dose learning image may be applied to the degrade image. Specifically, the device (100) may obtain a low dose learning image to which the noise component previously obtained for the degrade image is further applied. The device (100) may obtain a plurality of low dose learning images corresponding to each of the plurality of normal dose learning images. The low dose learning images corresponding to each normal dose learning image may be stored in a database as a single set of data. Accordingly, the device (100) can use the acquired set data for AI (deep learning) training. For example, the device (100) can learn a plurality of low dose learning images corresponding to each of a plurality of normal dose learning images, and acquire an enhanced learning model that acquires a normal dose learning image based on the low dose learning image. In this regard, the following will be explained with reference to FIGS. 3 to 8.
[0045] FIG. 3 is a diagram illustrating an example in which a device (100) according to an embodiment acquires a noise component for a normal dose learning image (310). The device (100) can acquire a normal dose learning image (310) representing an oral image of a first subject acquired through a shooting device, and can acquire a denoised image (320) from which the noise component has been removed in order to extract a noise component for the acquired normal dose learning image (310). The device (100) can acquire the denoised image (320) by performing noise filtering on the normal dose learning image (310) using a pre-trained AI image upscaler Real-ESRGAN. The device (100) can acquire a noise image (330) in which a difference occurs in an area through a difference value between data for all areas by comparing and analyzing the resolution, sharpness, etc. of the normal dose learning image (310) and the denoised image (320). Therefore, the device (100) can extract noise components from the noise image (330).
[0046] FIG. 4 is a diagram showing an example in which a device (100) according to one embodiment obtains a degraded noise component in which the contour component is reduced from the noise component.
[0047] In one embodiment, the device (100) can reduce the contour component from the noise component of the noise image (330). The device (100) can obtain a degraded noise component in which the contour component is reduced from the noise component. Additionally, the device (100) can obtain a raw dose learning image by applying the degraded noise component to the degraded image. For example, the device (100) can detect the contour component through edge component detection using the Canny Edge Detection algorithm. The device (100) can obtain a degraded noise image (410) in which the contour component is reduced from the noise image (330) in which the contour component is reduced. That is, the device (100) can obtain a degraded noise image (410) in which only the noise component other than the contour is reduced by reducing the contour component from the noise image (330). Accordingly, the device (100) can obtain a low dose learning image by applying a degraded noise image (410) to the degraded image. This will be explained in more detail with reference to FIGS. 5 and FIGS. 6.
[0048] FIG. 5 is a diagram illustrating an example in which a device (100) according to one embodiment acquires a degraded image (510) for a normal dose learning image (310). The device (100) may acquire a degraded image (510) by performing blurring or dilation on a normal dose learning image (310) for a first object. The normal dose learning image (310) may contain noise components along with original data. Therefore, as blurring and dilation are performed, the sharpness of the original data and noise components included in the normal dose learning image (310) may be reduced. That is, a degraded image (510) with reduced sharpness can be acquired by degrading according to the intensity of blurring and dilation set in the normal dose learning image (310).
[0049] FIG. 6 is a diagram illustrating an example in which a device (100) according to one embodiment applies a noise component to a degrade image (510) to obtain a low dose learning image (610). Specifically, the device (100) can obtain a low dose learning image (610) by applying a degrade noise image (410) to the degrade image (510), which represents a degrade noise component in which the contour component is reduced from the noise component. The degrade image (510) is an image obtained by degrading from a normal dose learning image (310), and as blurring and / or dilation are performed, the noise component included in the original normal dose learning image (310) may disappear together during the degrade process. Accordingly, the device (100) can further apply a degrade noise image (410) containing a pure noise component in which the contour component is excluded from the noise component to the degrade image (510). For example, the device (100) can obtain a low dose learning image (610) by applying a degraded noise component in which the contour component is reduced from the original noise component of the normal dose learning image (310) in the degraded image (510). That is, as the degraded noise component is further applied to the degraded image (510) that has been blurred and / or dilated, a low dose learning image (610) with reduced sharpness and more severe light scattering can be obtained. A plurality of low dose learning images (610) corresponding to each of a plurality of normal dose learning images (310) can be obtained.
[0050] FIG. 7 is a diagram illustrating an example in which a device (100) according to one embodiment applies an enhancement learning model to a low dose image (710) to obtain a normal dose image (720). In one embodiment, the normal dose learning image (310) may be a series of group images for obtaining a normal dose image (720) using a low dose image (710) obtained through ultra-low dose CBCT imaging. The device (100) may obtain an enhancement learning model through AI learning using multiple set data that form a set between multiple low dose learning images (610) obtained through multiple normal dose learning images (310). As described above, the device (100) obtains a noise component by acquiring a denoise image (320) from a plurality of normal dose learning images (310), and obtains a lower dose learning image (610) and an enhanced learning model according to each sequence in which a lower dose noise component, in which the contour component is reduced from the noise component, is applied to a lower dose image (510) in which the plurality of normal dose learning images (310) have been blurred and / or dilated. Therefore, the matching accuracy between the normal dose learning images (310) and the lower dose learning images (610) stored in the database can be increased. The device (100) can obtain a lower dose image (710) for a second subject through ultra-low dose CBCT imaging. In one embodiment, the second object is an object to be used to obtain a normal dose image (720) from a raw dose image (710) captured by applying an enhancement learning model. That is, an enhancement learning model obtained through a plurality of normal dose learning images (610) for the first object can be applied to the raw dose image (710) obtained for the second object. The raw dose image (710) is obtained at a lower dose than the normal dose image (720).The device (100) can obtain a normal dose image (720) for the second object without requiring a higher dose by applying an enhancement learning model to a low dose image (710) for the second object. Specifically, the device (100) can obtain a normal dose image (720) corresponding to the second object by applying an enhancement learning model to a low dose image (710) obtained at a dose lower than the dose required to obtain a normal dose image (720) for the second object. According to this, there is an effect that an image of improved quality can be obtained by obtaining a normal dose image (720) from an ultra-low dose CBCT.
[0051] FIG. 8 is an indicator schematically showing the efficiency when a device (100) according to one embodiment acquires a normal dose image (720) by applying an enhancement learning model to a low dose image (710).
[0052] Referring to FIG. 8, the modes for acquiring images may include a normal mode, a low-dose mode, and an ultra-low-dose mode. In one embodiment, the low-dose image (710) may be an image acquired in the ultra-low-dose mode. In the same detector with a matrix of 750x750x450 and a pitch of 0.2 mm, the normal mode and the low-dose mode do not use binning, but the ultra-low-dose mode uses binning. That is, in the ultra-low-dose mode, 2x2 pixels can be binning into one. Additionally, when acquiring an image in the ultra-low-dose mode according to one embodiment, if the same X-ray tube voltage of 95kV is applied to the normal mode, the low-dose mode, and the ultra-low-dose mode, the X-ray tube current must be 6mA in the normal mode, but 1mA may be applied in the low-dose mode and the ultra-low-dose mode. Additionally, the scan time may correspond to 22 seconds for normal mode and low dose mode, and a shorter 10 seconds for ultra-low dose mode. Accordingly, the dose levels for normal mode, low dose mode, and ultra-low dose mode are 100, 83.3, and 7.6, respectively, so that the dose level in low dose mode is reduced by 16.7 percent compared to normal mode, and the dose level in ultra-low dose mode is reduced by 92.4 percent compared to normal mode. One embodiment corresponds to ultra-low dose mode. Therefore, the device (100) can acquire a low dose image (710) according to a frame rate per second higher than the frame rate per second (FPS) required to acquire a normal dose image (720). That is, in the case of ultra-low dose mode, a high frame rate per second can be acquired because binning is used. Additionally, the device (100) can acquire a low dose image (710) according to a scan time shorter than the scan time required to acquire a normal dose image (720). The device (100) can acquire a number of low dose images (710) equal to the number of images acquired when acquiring a normal dose image (720).That is, according to one embodiment, when the device (100) acquires a low dose image (710) through an ultra-low dose mode, it can acquire the same number of images by applying a higher number of frames per second and a shorter scan time than when acquiring a normal dose image (720) through a normal mode. According to one embodiment, when using an ultra-low dose mode to acquire the same number of images, a higher number of frames per second and a shorter scan time can be applied, so the efficiency of acquiring images can be improved. In addition, when the device (100) acquires a low dose image (710) using an ultra-low dose mode, it acquires a normal dose image (720) by applying an enhancement learning model, so there is an effect that high-quality images can be acquired even at an ultra-low dose.
[0054] The embodiments described above are merely examples and are not to be interpreted as being limited thereto.
[0055] The sequence and combination of steps described above are examples of embodiments, and it can be understood that the sequence, combination, branching, functions, and entities performing them may be implemented in various forms with additions, omissions, or modifications, provided that the essential characteristics of each component described in the specification are not deviated from. Furthermore, throughout the specification, "provision" may be interpreted to include the process of a subject acquiring specific information or directly or indirectly transmitting or receiving it to a specific subject, and to comprehensively include the performance of related operations required in such processes.
[0056] Various embodiments of the present disclosure may be implemented as software comprising one or more instructions stored in a storage medium (e.g., memory) readable by a machine (e.g., a display device or a computer). For example, a processor (120) of the machine (e.g., processor (120)) may call at least one of the one or more instructions stored from the storage medium and execute it. This enables the machine to be operated to perform at least one function according to at least one called instruction. The one or more instructions may include code generated by a compiler or code that can be executed by an interpreter. The storage medium readable by the machine may be provided in the form of a non-transitory storage medium. Here, 'non-temporary' merely means that the storage medium is a tangible device and does not contain a signal (e.g., electromagnetic waves), and this term does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily.
[0057] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be distributed in the form of a device-readable storage medium, or distributed online (e.g., download or upload) through an application store (e.g., Play Store™) or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily created in a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.
[0058] Those skilled in the art related to the embodiments described above will understand that they may be implemented in modified forms without departing from the essential characteristics of the description. Therefore, the disclosed methods should be considered in an illustrative rather than a restrictive sense. The scope of the disclosure is defined by the claims, not by the foregoing description, and all variations within the scope of equivalence should be interpreted as being included in the disclosure. Explanation of the symbols
[0059] 100: Device 110: Receiver 120: Processor 130: Display 140: Wireless terminal 150: Wired terminal 310: Normal dose learning image 320: Denoise image 330: Noise image 410: Degraded noise image 510: Degraded Image 610: Raw dose learning image 710: Low Dose Image 720: Normal Dose Image
Claims
Claim 1 A method for a device to provide a low dose learning image, comprising: a receiving unit acquiring a normal dose learning image for a first object; a processor acquiring a denoise image by filtering noise from the normal dose learning image; the processor acquiring a noise component for the normal dose learning image based on a difference value between the normal dose learning image and the denoise image; the processor acquiring a degradation noise component in which the contour component is reduced from the noise component; the processor acquiring a degradation image in which the sharpness is reduced for the normal dose learning image; and the processor providing a low dose learning image obtained by applying the degradation noise component to the degradation image. Claim 2 A method according to claim 1, wherein the step of acquiring the degradation image comprises: a step in which the processor performs blurring and / or dilation on the normal dose learning image to reduce the sharpness; and a step in which, if the noise component included in the normal dose learning image is reduced by the blurring and / or dilation, the processor applies a degradation noise image containing a pure noise component with the contour component excluded to the degradation image. Claim 3 A method according to claim 1, further comprising the step of the processor learning the normal dose learning image and the low dose learning image to obtain an enhancement learning model that obtains the normal dose learning image based on the low dose learning image. Claim 4 A method according to claim 3, further comprising: a step in which the receiver acquires a low dose image of a second target; and a step in which the processor applies the enhancement learning model to the low dose image to acquire a normal dose image of the second target. Claim 5 In claim 4, the step of acquiring the low dose image is a method of acquiring the low dose image according to a frame rate per second higher than the frame rate per second (FPS) required when the processor acquires the normal dose image. Claim 6 In claim 4, the step of acquiring the low dose image is a method in which the processor acquires the low dose image according to a scan time shorter than the scan time required when acquiring the normal dose image. Claim 7 In claim 4, the step of acquiring the low dose image is a method of acquiring the same number of low dose images as the number of images acquired when the processor acquires the normal dose image. Claim 8 In claim 1, the step of obtaining the degradation noise component comprises: a step in which the processor detects the contour component by detecting the edge component of the noise component using a Canny Edge Detection algorithm, and reduces the detected contour component to obtain the degradation noise component. Claim 9 delete Claim 10 A device for providing a low dose learning image, comprising: a receiver for acquiring a normal dose learning image for a first object; and a processor for acquiring a denoise image by filtering noise from the normal dose learning image, acquiring a noise component for the normal dose learning image based on a difference value between the normal dose learning image and the denoise image, acquiring a degradation noise component in which a contour component is reduced from the noise component, acquiring a degradation image in which sharpness is reduced for the normal dose learning image, and providing a low dose learning image acquired by applying the degradation noise component to the degradation image. Claim 11 A device according to claim 10, wherein the processor performs blurring and / or dilation on the normal dose learning image to reduce the sharpness, and, when the noise component included in the normal dose learning image is reduced by the blurring and / or dilation, applies a degraded noise image containing a pure noise component with the contour component excluded to the degraded image. Claim 12 A device according to claim 10, wherein the processor learns the normal dose learning image and the low dose learning image, and obtains an enhancement learning model that obtains the normal dose learning image based on the low dose learning image. Claim 13 A device according to claim 12, wherein the receiver acquires a low dose image for a second target, and the processor acquires a normal dose image for the second target by applying the enhancement learning model to the low dose image. Claim 14 A device according to claim 13, wherein the processor acquires the low dose image according to a frame rate per second (FPS) higher than the frame rate per second required to acquire the normal dose image, acquires the low dose image according to a scan time shorter than the scan time required to acquire the normal dose image, and acquires a number of low dose images equal to the number of images acquired when acquiring the normal dose image. Claim 15 A computer-readable recording medium having a program stored on it for executing the method of any one of claims 1 to 8 on a computer.
Citation Information
Patent Citations
CT ultra-low dose automatic three-dimensional positioning scanning method and system
CN111839574A
System, method and computer-accessible medium for ultralow dose computed tomography image reconstruction
US20200126271A1
Method and apparatus for generating medical image denoising model
KR1020220131809A
Ai-enabled ultra-low-dose CT reconstruction
WO2022266406A1
