Image space normalization and quantification system and method using the same

Abstract

The present invention provides a system and method for image space normalization and quantification using the same. The method includes the steps of receiving input image data, extracting a deformation field corresponding to the input image data based on a deformation field generator, and generating forward deformation data for the input image data using the deformation field to generate normalized data in which the input image data is spatially normalized, and iteratively learning the input image data to generate the learned deformation field generator, and the generating normalized data may include automatically generating the normalized data by performing spatial normalization using the input image data based on the learned deformation field generator corresponding to the input image data.

Description

[Technical Field]

[0001] The present invention relates to image space normalization and a quantification system and method using the same. [Background technology]

[0002] For statistical analysis of medical images, it is reasonable to normalize images obtained from individual subjects into a single space and compare them using 3D pixel values.

[0003] In particular, spatial normalization (SN or spatial normalization) is an essential procedure for statistical comparison and objective evaluation of brain positron emission tomography (PET) and single-photon emission computed tomography (SPECT) images. If spatial normalization is performed using only brain positron emission tomography (PET) images, many errors occur.

[0004] Specifically, positron emission tomography (PET), single photon emission computed tomography (SPECT), functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), and magnetoencephalography (MEG) are difficult to normalize spatially using individual information due to their characteristic of being functional images and limited spatial resolution.

[0005] Therefore, magnetic resonance imaging (MRI) or computed tomography (CT) is generally taken together, and the images are first spatially normalized. The resulting deformation vector field (or deformation field) is then applied to the functional image to perform spatial normalization. Although this method has the advantage of being able to obtain accurate data, it is time- and cost-intensive due to the need for expensive equipment.

[0006] There is also a method for spatially normalizing brain positron emission tomography (PET) images using an average template obtained from various samples, but this method is difficult to accurately analyze because it cannot accurately reflect the diverse characteristics between images of patients and normal individuals.

[0007] Meanwhile, in many fields, research is currently underway on using machine learning, which performs deep learning based on big data, to successfully solve complex and high-dimensional problems.

[0008] Therefore, there is a need for a technology that can perform accurate spatial normalization at low cost without using a separate magnetic resonance imaging (MRI) or computed tomography (CT) scan, using a machine learning technique or the like.

[0009] The matters described as the background art are merely intended to enhance understanding of the background of the present invention and should not be construed as acknowledging that they constitute prior art already known to those skilled in the art. Summary of the Invention [Problem to be solved by the invention]

[0010] The problem to be solved by the present invention is to provide an image space normalization and a quantification system and method using the same.

[0011] The problems to be solved by the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description. [Means for solving the problem]

[0012] To solve the above-mentioned problems, an image space normalization method and a quantification method using the same according to one embodiment of the present invention, performed by a management server, includes the steps of receiving input image data; extracting a deformation field corresponding to the input image data based on a deformation field generator; and generating forward deformation data for the input image data using the deformation field to generate normalized data in which the input image data is spatially normalized; and iteratively learning the input image data to generate the trained deformation field generator; and the generating the normalized data may include automatically generating the normalized data by performing spatial normalization using the input image data based on the trained deformation field generator corresponding to the input image data.

[0013] In one embodiment of the present invention, the deformation field generator may iteratively learn the input image data and different image data matched to the input image data.

[0014] In one embodiment of the present invention, the step of generating the deformation field generator may include: spatially normalizing the heterogeneous image data using the deformation field to generate deformed heterogeneous image data; calculating an error value that maximizes a similarity value between the deformed heterogeneous image data and template data corresponding to the deformed heterogeneous image data; and iteratively learning the deformation field generator to minimize the error value.

[0015] In one embodiment of the present invention, generating the normalized data includes repeatedly performing a process of generating at least one forward deformation data for the input image data using at least one deformation field based on at least one deformation field generator, thereby generating the normalized data in which the input image data is spatially normalized. In this embodiment, generating the normalized data includes extracting a first deformation field corresponding to the input image data based on a first deformation field generator; generating first forward deformation data for the input image data using the first deformation field; spatially normalizing the input image data using the first forward deformation data to generate spatially normalized first transformed input image data; extracting a second deformation field corresponding to the first transformed input image data based on a second deformation field generator; and generating second forward deformation data for the first transformed input image data using the second deformation field to generate the spatially normalized normalized data.

[0016] In one embodiment of the present invention, the method may further include: generating a backward deformation field for the template data using the deformation field; performing backward spatial normalization by applying the backward deformation field to the template data; and comparing the backward spatially normalized image with the input image data or the foreign image data; and may further include iteratively learning the backward spatially normalized image using the deformation field generator. In one embodiment of the present invention, the method may further include extracting and quantifying a quantitative value using the normalized data.

[0017] In addition, an image space normalization and quantification system using the same according to one embodiment of the present invention for solving the above-mentioned problems includes: an imaging device that acquires input image data and heterogeneous image data that matches the input image data; and a management server that extracts a deformation field corresponding to the input image data based on a deformation field generator, generates forward deformation data for the input image data using the deformation field, and generates normalized data in which the input image data is spatially normalized; wherein the management server iteratively learns the input image data to generate the trained deformation field generator, and the management server can automatically generate the normalized data by performing spatial normalization using the input image data based on the trained deformation field generator corresponding to the input image data.

[0018] A program according to an embodiment of the present invention is stored in a computer-readable recording medium so that the program can be combined with a computer, which is hardware, to perform the image space normalization and the quantification method using the image space normalization. Other specific details of the invention are included in the detailed description and drawings. [Effects of the Invention]

[0019] According to the present invention, it is possible to automatically perform spatial normalization using personal, business, and corporate images corresponding to medical, financial, construction, insurance, law, education, public services, and culture.

[0020] According to the present invention, automatic spatial normalization can be performed using a deformation field extracted using a medical image that mainly contains functional information without using a medical image that contains anatomical information.

[0021] According to the present invention, a deformation field generator is iteratively trained using both medical images that primarily contain functional information and medical images that primarily contain matched anatomical information, and then spatially normalized automatically using only the medical images that primarily contain functional information, thereby enabling spatial normalization to accurately reflect the various characteristics of medical images that contain individual functional information.

[0022] According to the present invention, by automatically quantifying only medical images that mainly contain functional information, errors can be minimized and highly accurate image interpretation results can be obtained. According to the present invention, the stage of a disease can be more clearly classified and determined using individualized images, and medical costs can be reduced through automated diagnosis. According to the present invention, automatically spatially normalized images can be used to quantify various analytical fields.

[0023] According to the present invention, various diseases can be detected early using automatically spatially normalized images, thereby providing opportunities for improved access to medical and support services and for establishing financial and care plans. The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description. [Brief explanation of the drawings]

[0024] [Figure 1] 1 is a conceptual diagram illustrating image space normalization and a quantification system using the same according to an embodiment of the present invention; [Figure 2] 2 is a diagram illustrating the detailed configuration of the image space normalization shown in FIG. 1 and a quantification system using the image space normalization; [Figure 3] 1 is a diagram illustrating image space normalization, a quantification method using the image space normalization, and a learning method using a deformation field generator according to an embodiment of the present invention. [Figure 4]4 is a diagram illustrating a concept of generating spatially normalized data shown in FIG. 3 according to an embodiment of the present invention. [Figure 5] 4 is a diagram illustrating a concept of generating spatially normalized data shown in FIG. 3 according to another embodiment of the present invention. [Figure 6] 4 is a diagram illustrating a concept of generating a learned deformation field generator shown in FIG. 3 according to an embodiment of the present invention. [Figure 7] 1 is a diagram illustrating an example according to an embodiment of the present invention and a comparative example; [Figure 8] 1 is a diagram illustrating an example according to an embodiment of the present invention and a comparative example; DETAILED DESCRIPTION OF THE INVENTION

[0025] The advantages and features of the present invention, as well as methods for achieving them, will become more apparent from the following detailed description of the embodiments in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, and may be embodied in various different forms. However, the present invention is defined only by the scope of the claims, and the present invention is not limited to the embodiments disclosed below.

[0026] The terms used in this specification are for the purpose of describing the embodiments and are not intended to limit the present invention. In this specification, the singular includes the plural unless otherwise specified in the context. The terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other elements in addition to the elements referenced. The same reference numerals refer to the same elements throughout this specification, and "and / or" includes each and every combination of one or more of the referenced elements. Although terms such as "first," "second," etc. are used to describe various elements, it should be understood that these elements are not limited by these terms. These terms are used merely to distinguish one element from another. Therefore, it should be understood that a first element referenced below may also be a second element within the technical spirit of the present invention.

[0027] Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used in the sense that they can be commonly understood by a person of ordinary skill in the art to which the present invention belongs. Furthermore, terms defined in commonly used dictionaries should not be interpreted ideally or excessively unless they are clearly and specifically defined. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0028] FIG. 1 is a conceptual diagram illustrating image space normalization and a quantification system using the same according to one embodiment of the present invention, and FIG. 2 is a diagram illustrating the detailed configuration of the image space normalization and a quantification system using the same shown in FIG. 1.

[0029] 1 and 2, an image space normalization and quantification system 1 according to an embodiment of the present invention may include an imaging device 10, a management server 20, and an administrator terminal 30. In this case, the administrator terminal 30 may be omitted.

[0030] Here, the photographing device 10, the management server 20, and the administrator terminal 30 can transmit and receive data in real time in synchronization using a wireless communication network. The wireless communication network may support various long-distance communication methods, such as Wireless LAN (WLAN), DLNA (Digital Living Network Alliance), Wireless Broadband (Wibro), WiMAX (World Interoperability for Microwave Access), GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), CDMA2000 (Code Division Multi Access 2000), EV-DO (Enhanced Voice-Data Optimized or Enhanced Voice-Data Only), WCDMA (Wideband CDMA), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), IEEE 802.16, Long Term Evolution (LTE), LTEA (Long Term Evolution-Advanced), Wireless Mobile Broadband Service (WMBS), and Bluetooth Low Energy (BLE). Various communication methods such as, but not limited to, RF (Radio Frequency), Zigbee, RF (Radio Frequency), LoRa (Long Range), etc. may be applied, and various widely known wireless or mobile communication methods may also be applied.

[0031] The imaging device 10 may be a device capable of acquiring input image data such as medical image data containing functional information by imaging the state of a patient's brain. That is, the imaging device 10 may acquire PET (Positron Emission Tomography) images of beta-amyloid present in the patient's brain by imaging the brain of a patient injected with a radioactive tracer. In this case, the medical image data containing functional information may include, but is not limited to, all functional medical images that are difficult to spatially normalize using single information, such as SPECT, fMRI, EEG, and MEG.

[0032] For example, the input image data may include not only medical image data but also individual, business, and corporate image data corresponding to finance, construction, insurance, law, education, public services, and culture.

[0033] Such imaging device 10 may be, but is not limited to, PET, SPECT, CT, MRI, NIRS, etc., which can image the brain of a patient into which a radioactive tracer has been injected and obtain functional image data regarding the distribution of the radioactive tracer in the patient's brain.

[0034] Specifically, the imaging device 10 includes at least one of a flash camera, a gamma camera, and a PET / CT scanner, and can image at least a portion of the patient's brain, including the frontal cortex, posterior cingulate cortex, lateral temporal lobe, parietal lobe, occipital cortex, caudate nucleus, central temporal lobe, and anterior cingulate cortex. The radiotracer is a substance injected into the patient's body that binds to beta amyloid plaques, which are known to cause Alzheimer's disease, and enables PET imaging of the beta amyloid present in the patient's brain. The type of radiotracer may include, but is not limited to, at least one radioisotope selected from F-18, C-11, N-13, and O-15.

[0035] According to an embodiment, the imaging device 10 may acquire other modality image data such as, but not limited to, MRI (Magnetic Resonance Imaging) images or CT (Computed Tomography), i.e., other modality image data containing anatomical information. In this case, the other modality image data may be data matched to the input image data, but is not limited to this. Although the present embodiment has been described as being measured using the imaging device 10, the present invention is not limited to this and may be image data recorded in institutions such as companies, courts, hospitals, etc. Management Server 20 Data transmission / reception unit 200, a database unit 220, a monitoring unit 240 and a management control unit 260. Data transmission / reception unit The unit 200 can receive input image data from the image capture device 10 and different image data to be matched corresponding to the input image data. By way of example, Data transmission / reception unit The terminal 200 can transmit input image data and different image data to the administrator terminal 30 and receive normalized data from the administrator terminal 30 . The database unit 220 can store data transmitted and received between the photographing device 10, the management server 20, or the administrator terminal 30 through a wireless communication network.

[0036] The database unit 220 can store data that supports various functions of the management server 20. That is, the database unit 220 can store a number of application programs (or applications) that are run by the management server 20, and data and commands for the operation of the management server 20. At least some of these applications can be downloaded from an external server via wireless communication.

[0037] Meanwhile, the data used in this embodiment and stored in the database unit 220 may be implemented in the form of a mapping table corresponding to each other, but is not limited thereto.

[0038] The monitoring unit 240 can monitor the operating status of the photographing device 10, the management server 20, or the administrator terminal 30, and data transmitted and received between the photographing device 10, the management server 20, or the administrator terminal 30 through a screen.

[0039] When the management control unit 260 receives input video data from the image capturing device 10, it can automatically perform spatial normalization using only the input video data to generate normalized data.

[0040] Specifically, when the management control unit 260 receives input image data, it extracts a deformation field corresponding to the input image data based on a deformation field generator, and generates forward deformation data for the input image data using the extracted deformation field, thereby generating normalized data in which the input image data is spatially normalized.

[0041] That is, the management controller 260 can extract a deformation field for the input image data to receive the input image data and perform spatial normalization. At this time, the input image data can be an image or video, as long as it is an image in which shaking, color bleeding, surrounding noise, etc. are automatically pre-processed and redundant data is removed.

[0042] Here, the deformation field can be extracted by vector field values ​​that allow each pixel included in the input image data to move in 3D directions around the x, y, and z axes for spatial normalization according to the template data.

[0043] For example, the management controller 260 may extract a deformation field from the pre-processed input image data in accordance with the template data, thereby generating forward deformation data for the input image data. In addition, the management controller 260 can generate template data for spatially normalizing the input image data.

[0044] For example, when the management controller 260 receives a plurality of input image data, it inputs the data to a deep learning architecture and generates spatially normalized data by applying a deformation field generator learned through deep learning. In this case, the deep learning architecture may include a convolutional neural network, which can represent an artificial neural network that performs calculations on the input image, understands it, extracts features, acquires information, or generates a new image.

[0045] In addition, when the management control unit 260 receives input image data, it extracts a deformation field corresponding to the input image data based on the deformation field generator, generates forward deformation data for the input image data using the extracted deformation field, generates a backward deformation field for the input image data using template data, and then generates normalized data in which the input image is spatially normalized using the generated forward deformation data.

[0046] That is, the management control unit 260 can generate a backward deformation field for the template data using the deformation field.

[0047] Specifically, the management controller 260 may apply the template data to the deformation field to generate a backward deformation field for the input image data so that the accuracy of the input image data and the template data is increased.

[0048] For example, the management controller 260 may generate a backward deformation field for template data using a deformation field, apply the generated backward deformation field to the template data to perform backward spatial normalization, and then compare the backward spatially normalized image with input image data or foreign image data. At this time, the management controller 260 may iteratively learn the backward spatially normalized image using a deformation field generator.

[0049] According to an embodiment, the management controller 260 may apply the forward deformation data to the deformation field to generate a backward deformation field for the input image data. Furthermore, the management control unit 260 can generate normalized data having continuity by using only the input video data.

[0050] Specifically, the management controller 260 applies the input image data to a plurality of deformation field generators to repeatedly perform spatial normalization on the input image data, thereby generating more accurate normalized data.

[0051] For example, when the management controller 260 receives input image data, it extracts a first deformation field corresponding to the input image data using a first deformation field generator, generates first forward deformation data for the input image data using the extracted first deformation field, spatially normalizes the generated first forward deformation data to generate spatially normalized first transformed input image data, extracts a second deformation field corresponding to the first transformed input image data using a second deformation field generator, and generates a spatially normalized first forward deformation data for the input image data using the extracted second deformation field. a third deformation field generator for generating a third deformation field corresponding to the second deformation input image data, a third deformation field generator for generating a third forward deformation data for the second deformation input image data, and a third forward deformation data generator for generating a third forward deformation data for the second deformation input image data.

[0052] Although this embodiment discloses that the final normalized data is generated based on the first to third deformation field generators, this is not limited thereto, and the normalized data can be generated based on at least the first deformation field generator.

[0053] In addition, the management controller 260 may generate a trained deformation field generator by repeatedly training the deformation field generator using input image data and heterogeneous image data matched to the input image data. In this case, the management controller 260 may train the deformation field generator using a deep learning or machine learning technique, but is not limited thereto.

[0054] Specifically, the management controller 260 may generate deformed heterogeneous image data by spatially normalizing the heterogeneous image data using the deformation field, and then calculate an error value that maximizes the similarity between the generated deformed heterogeneous image data and template data corresponding to the deformed heterogeneous image data. The management controller 260 may iteratively train the deformation field generator to minimize the calculated error value, thereby generating a trained deformation field generator.

[0055] In other words, the management and control unit 260 generates spatially normalized transformed input image data based on the input image data, calculates an error value between the spatially normalized transformed heterogeneous image data and the corresponding template data based on heterogeneous image data matched and paired with the input image data, determines whether the generated transformed input image data is authentic, and generates normalized data. That is, the management and control unit 260 can automatically perform spatial normalization using only medical images containing functional information by iteratively learning input image data containing functional information and heterogeneous image data matched thereto containing anatomical information.

[0056] For example, the management and control unit 260 can calculate an error value that maximizes the cross-correlation between the MRI deformation data and the MRI template data. That is, iterative learning can be performed using the deformation field generator to minimize the calculated error value.

[0057] In addition, the management control unit 260 can automatically generate normalized data by spatially normalizing the forward deformation data based on the learned deformation field generator.

[0058] According to an embodiment, the management controller 260 can verify the compatibility of the input image data and the normalized data matched to the input image data through iterative learning based on a convolutional neural network (CNN) algorithm.

[0059] In addition, the management control unit 260 can extract and quantify a quantitative value for a specific field using the normalized data. The management server 20 having such a structure can generate spatially normalized normalized data by performing spatial normalization when only the input image data is provided using the deformation field extracted corresponding to the input image data based on a deformation field generator that is trained to have a minimum error value by comparing spatially normalized data for heterogeneous image data corresponding to the input image data with template data.

[0060] The management server 20 may be implemented using hardware circuits (e.g., CMOS-based logic circuits), firmware, software, or a combination thereof. For example, it may be implemented using transistors, logic gates, and electronic circuits in the form of various electrical structures.

[0061] The administrator terminal 30 is a terminal owned by a separate administrator, and can transmit and receive data in real time by synchronizing with the photographing device 10 and the management server 20 through a wireless communication network. At this time, the administrator terminal 30 can transmit and receive data using an application program.

[0062] The administrator terminal 30 can automatically quantify the input image data received from the imaging device 10 and / or the management server 20 through spatial normalization.

[0063] The administrator terminal 30 may include various portable electronic communication devices that support communication with the photographing device 10 and the management server 20. For example, the photographing device 10 may include various portable terminals such as a smartphone, a personal digital assistant (PDA), a tablet, a wearable device, a smartwatch, a smart glass, a head mounted display (HMD), and various Internet of Things (IoT) terminals as separate smart devices, but may also include electronic communication devices such as a non-portable desktop computer and a workstation computer.

[0064] The operation of the image space normalization and quantification system using the same according to an embodiment of the present invention having the above structure is as follows: Figure 3 is a diagram illustrating image space normalization and a quantification method using the same, and a learning method using a deformation field generator, according to an embodiment of the present invention, Figure 4 is a diagram illustrating a concept of generating spatially normalized data shown in Figure 3 according to an embodiment of the present invention, Figure 5 is a diagram illustrating a concept of generating spatially normalized data shown in Figure 3 according to another embodiment of the present invention, and Figure 6 is a diagram illustrating a concept of generating a learned deformation field generator shown in Figure 3 according to an embodiment of the present invention. First, as shown in FIG. 3, the management server 20 can receive input image data from the image capturing device 10 (S100).

[0065] At this time, the input image data may include not only medical image data including functional data, but also individual, business, and corporate image data corresponding to finance, construction, insurance, law, education, public services, and culture. Next, the management server 20 can determine whether or not spatial normalization using the input image data is progressing (S110).

[0066] When spatial normalization is performed using only the input image data (S110), the management server 20 can extract a deformation field corresponding to the input image data using the deformation field generator (S120).

[0067] Next, the management server 20 can generate forward deformation data for the input image data using the deformation field (S130). Next, the management server 20 can generate a backward deformation field using the template data (S140). Next, the management server 20 can generate normalized data that is spatially normalized with the forward deformation data (S150).

[0068] Specifically, referring to FIG. 4, when the management server 20 receives input image data, it extracts a deformation field corresponding to the input image data based on a deformation field generator, generates forward deformation data for the input image data using the extracted deformation field, applies the deformation field to the template data to generate a backward deformation field, and then generates normalized data that is spatially normalized using the generated forward deformation data. In accordance with an embodiment, the management server 20 can generate normalized data having continuity.

[0069] For example, as shown in FIG. 5, when the management server 20 receives input image data, it extracts a first deformation field corresponding to the input image data using a first deformation field generator, generates first forward deformation data for the input image data using the extracted first deformation field, and generates spatially normalized first transformed input image data; extracts a second deformation field corresponding to the first transformed input image data using a second deformation field generator, generates second forward deformation data for the first transformed input image data using the extracted second deformation field, and generates spatially normalized second transformed input image data; extracts a third deformation field corresponding to the second transformed input image data using a third deformation field generator, generates third forward deformation data for the second transformed input image data using the extracted third deformation field, and generates spatially normalized final normalized data using the generated third forward deformation data. Finally, the management server 20 can perform quantification to extract quantitative values ​​for specific regions from the normalized data (S160).

[0070] Alternatively, if spatial normalization is not performed using only the input image data (S110), the management server 20 can generate a learned deformation field generator using deep learning or machine learning techniques using the input image data and heterogeneous image data (S170-S200).

[0071] Specifically, referring to FIG. 6, the management server 20 receives input image data and heterogeneous image data to be matched to the input image data (S170), then spatially normalizes the heterogeneous image data using a deformation field to generate deformed heterogeneous image data (S180), calculates an error value that maximizes the similarity value between the generated deformed heterogeneous image data and template data corresponding to the deformed heterogeneous image data (S190), and performs iterative learning using a deformation field generator to minimize the error value (S200).

[0072] Hereinafter, the accuracy of spatial normalization will be described in detail by comparing a spatial normalization method using only input image data according to an embodiment of the present invention with a spatial normalization method using input image data and heterogeneous image data matched to the input image data, with reference to Figures 7 and 8. Figures 7 and 8 are diagrams for explaining an example and a comparative example according to an embodiment of the present invention, where Figure 7 is a diagram showing the comparison results of spatial normalization using an example and a comparative example when amyloid is negative, and Figure 8 is a diagram showing the comparison results of spatial normalization using an example and a comparative example when amyloid is positive.

[0073] 7 and 8 illustrate examples in which normalized data is generated by spatially normalizing only input image data based on a deformation field generator trained to have a minimum error value by comparing spatial normalized data for heterogeneous image data corresponding to the input image data with template data, and first and second comparative examples illustrate examples in which normalized data is generated by spatially normalizing input image data and heterogeneous image data corresponding to the input image data. In this case, the first input image data may be a first input medical image mainly including functional information and may be input image data for a case in which cerebral atrophy is severe and amyloid is negative, and the second input image data may be a first input medical image mainly including functional information and may be input image data for a case in which cerebral atrophy is severe and amyloid is positive. In this embodiment, the first and second comparative examples can be generated using the SPM12 program and an MRI matched to the input image, but are not limited thereto.

[0074] 7, when spatial normalization is performed using only input image data as in Example 1 when the brain is severely atrophied and amyloid is negative, the spatial normalized data for Example 1 is more accurate than that for Comparative Example 1. That is, when spatial normalization is performed using only input image data according to an embodiment of the present invention, a result with high accuracy of normalized data can be obtained.

[0075] For example, in the examples and comparative examples, the accuracy of spatial normalization can be analyzed by calculating the true value or gold standard value, which is calculated from the standard uptake value ratio (SUVR) value of the PET image. In this case, the true value can be, but is not limited to, a value obtained by finely segmenting the brain region on the MRI using a program called Freesurfer.

[0076] In other words, Comparative Examples 1 and 2 perform spatial normalization using input image data and heterogeneous image data corresponding to the input image data, while Examples 1 and 2 can perform spatial normalization using only the input image data based on a learned deformation field generator that has the smallest error value by comparing spatial normalized data for heterogeneous image data corresponding to the input image data with template data.

[0077] Accordingly, the spatial normalization method using the first and second embodiments according to the present invention can perform accurate spatial normalization of an input image without heterogeneous image data while minimizing the execution time.

[0078] The image space normalization and the quantification system and method using the same according to each embodiment of the present invention are technologies developed through the "Platform Development for Functional Brain Image AI-Based Analysis of Degenerative Brain Diseases" project, a 2020 Biomedical Technology Commercialization Support Project (BT200151) by the Seoul Industrial Development Agency.

[0079] The steps of a method or algorithm described in connection with the embodiments of the present invention may be embodied directly in hardware, in a software module executed by hardware, or in a combination thereof. The software module may reside in Random Access Memory (RAM), Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), Flash Memory, a hard disk, a removable disk, a CD-ROM, or any other form of computer-readable storage medium commonly known in the art to which the present invention pertains.

[0080] Although the embodiments of the present invention have been described above with reference to the accompanying drawings, those skilled in the art will understand that the present invention may be embodied in other specific forms without changing the technical spirit or essential characteristics thereof. Therefore, it should be understood that the above-described embodiments are illustrative in all respects and are not limiting. [Explanation of symbols]

[0081] 1. Quantification System 10 Imaging equipment 20 Management Server 200 Data transmission / reception unit 220 Database Department 240 Monitoring Department 260 Management and Control Unit 30 Administrator terminal

Claims

1. In a video space normalization method performed by a management server, receiving input video data; extracting a deformation field corresponding to the input image data based on a deformation field generator; and generating forward deformation data for the input image data using the deformation field, thereby generating normalized data in which the input image data is spatially normalized; Iteratively learning a combination of the input video data and the heterogeneous video data associated with the input video data to generate the learned deformation field generator; generating the normalized data includes: performing spatial normalization on only the input image data based on the learned deformation field generator to automatically generate the normalized data.

2. The step of generating a deformation field generator includes: generating deformed heterogeneous image data by spatially normalizing the heterogeneous image data using the deformation field; calculating an error value that maximizes a similarity value between the transformed heterogeneous image data and template data corresponding to the transformed heterogeneous image data; and 2. The method of claim 1, further comprising the step of iteratively training the deformation field generator so as to minimize the error value.

3. generating the normalized data includes: repeatedly performing a process of generating at least two or more forward deformation data for the input image data using at least two or more deformation fields based on at least two or more deformation field generators, and generating the normalized data by spatially normalizing the input image data; generating the normalized data includes: extracting a first deformation field corresponding to the input image data based on a first deformation field generator; generating first forward deformation data for the input image data using the first deformation field; spatially normalizing the input image data using the first forward deformation data to generate spatially normalized first transformed input image data; extracting a second deformation field corresponding to the first modified input image data based on a second deformation field generator; and 2. The image space normalization method of claim 1, further comprising: generating second forward deformation data for the first transformed input image data using the second deformation field to generate the spatially normalized normalized data.

4. generating a backward deformation field for the template data using the deformation field; applying the backward deformation field to the template data to perform backward spatial normalization; and comparing the backward spatially normalized image with the input image data or the foreign image data; The image space normalization method of claim 2 , further comprising: iteratively learning the backward space normalized image using the deformation field generator.

5. an imaging device that acquires input image data and different image data that is matched to the input image data; and a management server that extracts a deformation field corresponding to the input image data based on a deformation field generator, generates forward deformation data for the input image data using the deformation field, and generates normalized data in which the input image data is spatially normalized; The management server Iteratively learning a combination of the input video data and the different video data associated with the input video data to generate the learned deformation field generator; The management server An image space normalization system that automatically generates the normalized data by performing space normalization on only the input image data based on the learned deformation field generator.

6. A computer program stored on a computer-readable recording medium so that the computer program can be combined with a computer that is hardware to perform the method according to claim 1.

Images