Device and method for converting contrast agent-free MRI into contrast agent-using MRI by using artificial intelligence

Abstract

The present invention relates to a device for converting contrast agent-free magnetic resonance imaging (MRI) to contrast agent-using MRI by using artificial intelligence. The device comprises: an MRI image selection unit for calling, when a patient identifier (ID) is received, a list of MRI images corresponding to a patient ID and selecting one MRI image from the list; a segmentation unit for segmenting a region of interest by applying a pre-trained algorithm to the selected contrast agent-free MRI image; a conversion unit for converting the segmented region of interest of the contrast agent-free MRI image into a contrast agent-using MRI image at the same anatomical position by applying a model trained to generate a contrast agent-using MRI image at the same anatomical position from a contrast agent-free MRI image; and an image quality improvement unit for improving the image quality of the converted contrast agent-using MRI image.

Description

Device and method for converting contrast-free MRI into contrast-enhanced MRI using artificial intelligence

[0001] The present invention relates to an apparatus and method for converting contrast-free magnetic resonance imaging (MRI) into contrast-enhanced MRI using artificial intelligence (AI).

[0002] MRI is a core medical imaging technology that visualizes the internal structures of the human body in high resolution to detect lesions and diagnose diseases, playing a crucial role in various clinical situations. In particular, it holds an unrivaled position in the field of medical diagnostic imaging because it can provide high-quality images without using radiation.

[0003] MRI using contrast agents enhances tissue contrast, enabling clear visualization of the location and characteristics of lesions; it is essential for the diagnosis and evaluation of diseases such as cancer, vascular disease, and inflammation, and offers the following clinical benefits.

[0004] Contrast-enhanced MRI clearly visualizes the size, boundaries, and blood flow patterns of tumors, contributing to cancer staging and treatment planning. It also diagnoses vascular abnormalities such as stenosis, aneurysms, and thrombi, and precisely analyzes blood flow conditions. Furthermore, by utilizing the characteristic of increased vascular permeability in areas of inflammation or infection, contrast-enhanced MRI can accurately determine the distribution and severity of lesions. It also aids in immediate treatment decisions by providing rapid and accurate information in emergency situations, such as stroke.

[0005] However, the use of contrast agents can cause allergic reactions in some patients and, in severe cases, may induce anaphylaxis; therefore, their use is restricted in specific patient groups, such as pregnant women, children, or patients with chronic diseases. Furthermore, the need for contrast agent injection procedures and additional imaging results in the disadvantage of increased overall examination time and costs. Consequently, to address these issues, a technological approach is required that overcomes the diagnostic limitations of non-contrast MRI and generates high-quality images comparable to those produced by contrast-enhanced MRI.

[0006] The present invention was devised to solve the above-mentioned problems, and the objective of the present invention is to provide an apparatus and method for converting an MRI without a contrast agent into an MRI with a contrast agent using artificial intelligence.

[0007] An apparatus for converting a contrast-free MRI into a contrast-enhanced MRI using artificial intelligence according to an embodiment of the present invention for achieving the above objective comprises: an MRI selection unit that, when a patient ID is input, retrieves a list of MRIs corresponding to the patient ID and selects one of the MRIs; a segmentation unit that segments a region of interest by applying a pre-learned algorithm to the selected contrast-free MRI; a conversion unit that converts the region of interest of the segmented contrast-free MRI into a contrast-enhanced MRI of the same anatomical location by applying a model learned to generate a contrast-enhanced MRI of the same anatomical location from the contrast-free MRI; and an image quality improvement unit that improves the image quality of the converted contrast-enhanced MRI.

[0008] A method for converting a contrast-free MRI into a contrast-enhanced MRI using artificial intelligence according to an embodiment of the present invention for achieving the above objective comprises: a step of, when a patient ID is input, retrieving a list of MRIs corresponding to the patient ID and selecting one of the MRIs; a step of segmenting a region of interest by applying a pre-trained algorithm to the selected contrast-free MRI; a step of converting the region of interest of the segmented contrast-free MRI into a contrast-enhanced MRI of the same anatomical location by applying a model trained to generate a contrast-enhanced MRI of the same anatomical location from the contrast-free MRI; and a step of improving the image quality of the converted contrast-enhanced MRI.

[0009] According to one aspect of the present invention described above, by providing an apparatus and method for converting a contrast-free MRI into a contrast-enhanced MRI using artificial intelligence, it is possible to generate images similar to a contrast-enhanced MRI without using a contrast agent, and eliminate the risk of allergic reactions to the contrast agent and side effects in patients with impaired renal function.

[0010] Furthermore, it overcomes the diagnostic limitations of contrast-free MRI and provides high-quality images capable of clearly analyzing the location and characteristics of lesions such as cancer, vascular diseases, and inflammation, thereby improving the diagnostic accuracy of lesions that are difficult to detect with conventional contrast-free MRI.

[0011] Furthermore, since the procedure for injecting contrast agents and additional imaging are unnecessary, examination time is shortened and overall examination costs are reduced, providing efficiency to both patients and medical institutions.

[0012] FIG. 1 is a device diagram showing the internal blocks of a device for converting a contrast-free MRI into a contrast-enhanced MRI according to an embodiment of the present invention.

[0013] FIG. 2 is a diagram showing an example of a model that converts a contrast-free MRI into a contrast-enhanced MRI based on a conditional GAN ​​according to an embodiment of the present invention.

[0014] FIG. 3 is a diagram showing an example in which the image quality enhancement unit of FIG. 1 outputs an MRI using a contrast agent with improved image quality.

[0015] FIG. 4 is a diagram showing an example of a platform to which a conditional GAN-based model according to an embodiment of the present invention is applied, converting a contrast-free MRI into a contrast-enhanced MRI.

[0016] FIG. 5 is a diagram showing examples of segmented regions of interest from MRI without contrast agent and MRI with contrast agent,

[0017] And, FIG. 6 is a flowchart showing the operation of an MRI conversion device according to an embodiment of the present invention converting an MRI without a contrast agent into an MRI with a contrast agent.

[0018] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0019] The components according to the present invention are defined by functional distinction rather than physical distinction, and can be defined by the functions each performs. Each component may be implemented as hardware or as program code and processing units that perform each function, and the functions of two or more components may be included and implemented in a single component. Therefore, it should be noted that the names assigned to the components in the following embodiments are not intended to physically distinguish each component but are assigned to imply the representative function performed by each component, and that the technical concept of the present invention is not limited by the names of the components.

[0020] Before describing the embodiments of the present invention, the AI ​​algorithm applied in the present invention will be described.

[0021] Synthetic medical data is defined as 'plausible but non-existent fake data' created based on actual medical data such as clinical information, genomics, and lifelogs. It is generated to apply to specific environments where data cannot be directly obtained, or to protect the personal information and confidentiality of data. A representative AI algorithm applied to the generation of synthetic medical data is the Generative Adversarial Networks (GAN) algorithm. Unlike existing deep learning algorithms, it consists of a generator that creates fake data and a discriminator that distinguishes between real and fake data, allowing new medical image data to be generated through a competitive process between two opposing systems.

[0022] Preferred embodiments of the present invention will be described in more detail below with reference to the drawings.

[0023] FIG. 1 is a diagram showing the internal blocks of a device for converting a contrast-free MRI into a contrast-enhanced MRI according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of a model for converting a contrast-free MRI into a contrast-enhanced MRI based on a conditional GAN ​​according to an embodiment of the present invention, FIG. 3 is a diagram showing an example of an image quality improvement unit of FIG. 1 outputting a contrast-enhanced MRI with improved image quality, and FIG. 4 is a diagram showing an example of a region of interest segmented from a contrast-free MRI and a contrast-enhanced MRI.

[0024] The illustrated MRI conversion device includes an MRI selection unit (110), a segmentation unit (120), a conversion unit (130), and an image quality improvement unit (140).

[0025] When a patient identifier (ID) is entered, the MRI selection unit (110) retrieves a list of MRIs corresponding to the patient ID and selects one of the MRIs. At this time, the MRI list may be generated based on the MRIs of the patient collected from the medical imaging repository of each medical institution and includes MRIs of various sequences.

[0026] Additionally, it is assumed that the selected MRI is a T2FS (T2 Fat Saturation) sequence MRI, wherein the T2FS sequence is an MRI sequence designed for fat attenuation that suppresses high signals in adipose tissue to enable better visualization of signals in other tissues.

[0027] The segmentation unit (120) segments the region of interest by applying a pre-learned algorithm to the MRI without contrast agent selected through the MRI selection unit (110). Here, the region of interest refers to an area from which organs (small intestine, large intestine, lungs, etc.) in which movement over time is detected are excluded in the MRI without contrast agent and the MRI with contrast agent. For example, assuming a sagittal MRI containing the vertebral body, the region of interest refers to the remaining area in the sagittal MRI excluding the anterior region of the spine containing the organs.

[0028] In addition, the applied algorithm is trained based on data in which a professional labeler manually separates regions of interest from multiple MRIs, automatically segmenting fixed structures excluding organ areas where movement is detected from MRIs without contrast agents and MRIs with contrast agents, thereby preventing temporal and spatial distortion.

[0029] As another example, the segmentation unit (120) may apply a dynamic time-space based segmentation algorithm to increase the conversion accuracy between MRI without contrast agent and MRI with contrast agent. In this case, the segmentation unit (120) analyzes continuous time-series data of the same patient during the segmentation process to strengthen the algorithm to reflect the characteristics of tissues that change over time and space, thereby allowing for more precise segmentation of the region of interest by reflecting the minute movements of the lesion. This is differentiated from single-image based segmentation and improves the precision and reliability of the converted MRI by considering changes along the time axis.

[0030] The conversion unit (130) applies a model trained to generate a contrast-enhanced MRI of the same anatomical location from a contrast-free MRI, and converts the region of interest of the contrast-free MRI segmented through the segmentation unit (120) into a contrast-enhanced MRI of the same anatomical location. The converted MRI is assumed to be a T1CFS (T1 Contrast-enhanced Fat Saturation) sequence MRI, where the T1CFS sequence is a sequence that uses a contrast agent and suppresses fat signals to highlight specific lesions.

[0031] In addition, the applied model is a generative model based on Pix2Pix among conditional GAN ​​algorithms, and is trained by a generator and a discriminator.

[0032] As explained in more detail through FIG. 2, the generator (210) learns the contrast-free MRI (201) and the contrast-use MRI (203) of the same anatomical location and generates the contrast-use MRI of the same anatomical location as the contrast-free MRI.

[0033] That is, the generator (210) learns the source data, the MRI without contrast agent (201), and the target data, the MRI with contrast agent (203), simultaneously to identify the 1:1 relationship between pixels in the two images. At this time, the MRI without contrast agent used for the generator (210)'s learning is an MRI in which the region of interest has been segmented through a pre-learned algorithm, and the MRI with contrast agent used for the generator (210)'s learning is an MRI in which the same region as the region of interest of the segmented MRI without contrast agent has been segmented from the MRI with contrast agent. The reason for performing segmentation first on the MRI without contrast agent in this way is that the distinction between structures appears more clearly in the MRI without contrast agent when segmentation is performed.

[0034] In addition, the MRI without contrast agent with the region of interest segmented is like the right image in Fig. 4(a), and the MRI with the region of interest segmented is like the right image in Fig. 4(b).

[0035] The discriminator (220) performs learning to determine authenticity by comparing the contrast-enhanced MRI (205) generated through the generator (210) with the actual contrast-enhanced MRI (203) used for learning, and performs repeated learning for the process until the similarity between the generated contrast-enhanced MRI (205) and the actual contrast-enhanced MRI (203) reaches a maximum value. Here, the maximum value refers to a threshold value used to measure the similarity between the generated contrast-enhanced MRI (205) and the actual contrast-enhanced MRI (203), which is a value indicating that the similarity between the two images is very high.

[0036] Additionally, the discriminator (220) measures the error value (loss) (207) between the generated contrast-enhanced MRI (205) and the actual contrast-enhanced MRI (203) and feeds it back to the generator (210). That is, the discriminator (220) updates the model by feeding back the error value between the two images, and through this, the generator (210) can generate a contrast-enhanced MRI that is more similar to the actual contrast-enhanced MRI (203).

[0037] The generator (210) generates a contrast-enhanced MRI of the same anatomical location as the contrast-unenhanced MRI selected through the MRI selection unit (110) by reflecting the error value (207).

[0038] As another example, the model applied in the transformation unit (130) can be designed to maintain structural consistency by comparing anatomical structures with predefined label data, and to this end, the conditional GAN-based generative model may include an additional structure-preserving loss function. This can increase similarity with contrast-enhanced MRI while simultaneously resolving structural distortion issues that may occur during MRI transformation.

[0039] The image quality improvement unit (140) improves the image quality of the contrast-enhanced MRI converted through the conversion unit (130). That is, the image quality improvement unit (140) inputs the contrast-enhanced MRI converted as shown in FIG. 3, removes noise and blurring, applies super-resolution to adjust the resolution, and finally outputs the contrast-enhanced MRI with improved image quality.

[0040] As another example, the image quality improvement unit (140) can improve image quality by adding a user-customized algorithm to optimize noise removal and resolution adjustment methods considering the age, gender, physical condition, etc. of a specific patient. For example, when processing MRI data of a pediatric patient, an improvement algorithm suitable for tissue size and density can be applied, thereby generating more accurate MRI data by reflecting the characteristics of each patient.

[0041] FIG. 5 is a diagram showing an example of a platform to which a conditional GAN-based model according to an embodiment of the present invention is applied, converting an MRI without contrast agent into an MRI with contrast agent.

[0042] When the driving program starts, the illustrated platform (510) receives an identifier (ID) from the user for a patient for whom a contrast agent-based MRI is to be generated, and retrieves the patient's MRI from the medical image repository of each medical institution and displays it on the screen. At this time, the MRI displayed on the screen may be composed of a list of MRIs by sequence.

[0043] The user selects the desired MRI sequence from the MRI list, and the region of interest of the selected MRI is segmented through an automatic segmentation algorithm.

[0044] When a region of interest image (501) of a non-contrast MRI is input to the platform (510), a conditional GAN-based model according to an embodiment of the present invention is applied to generate a contrast-enhanced MRI (503) of the same anatomical location as the region of interest image (501) of the non-contrast MRI and display it on a new screen.

[0045] FIG. 6 is a flowchart illustrating the operation of an MRI conversion device according to an embodiment of the present invention converting an MRI without a contrast agent into an MRI with a contrast agent.

[0046] When a patient ID is entered, the MRI conversion device retrieves a list of MRIs corresponding to the patient ID and selects one of them (S601), and applies a pre-trained algorithm to the selected MRI without contrast agent to segment the region of interest (S603).

[0047] Then, the MRI converter applies a model trained to generate a contrast-enhanced MRI of the same anatomical location from a contrast-free MRI to convert the region of interest of the contrast-free MRI segmented in S603 into a contrast-enhanced MRI of the same anatomical location. (S605)

[0048] Subsequently, the MRI converter improves the image quality of the contrast-enhanced MRI converted from S605. (S607)

[0049] Meanwhile, the apparatus and method proposed in this invention for converting contrast-free MRI into contrast-enhanced MRI can also be applied to embodiments that convert contrast-free MRI by combining other medical imaging data, such as CT (Computed Tomography), PET (Position Emission Tomography), and ultrasound, in addition to MRI. In this case, an artificial intelligence model is trained by fusing multimodal image data captured on the same patient and applied to the conversion and segmentation units. Through this, anatomical and pathological information that is difficult to identify with MRI alone can be additionally reflected, thereby further enhancing the quality and diagnostic value of the converted MRI.

[0050] Furthermore, the apparatus and method proposed in this invention for converting a contrast-free MRI into a contrast-enhanced MRI can introduce a lesion detection algorithm that effectively detects lesions in the contrast-free MRI and links them with the converted MRI. That is, by adding a lesion region matching function between the converted contrast-enhanced MRI and the original MRI, the conversion results can be directly utilized for lesion detection and diagnosis; and by expanding this into a medical diagnostic support system linked to lesion detection, the technology can be differentiated.

[0051] In addition, algorithm optimization and GPU acceleration technology can be applied to process the MRI conversion process in real time, and the computational efficiency of the conversion unit (130) and the image quality improvement unit (140) can be maximized so that the conversion results can be checked in real time within the hospital. This real-time processing technology can provide high-quality images quickly without the use of contrast agents in emergency situations, thereby greatly improving clinical utility.

[0052] The method of the present invention for converting a contrast-free MRI into a contrast-enhanced MRI can be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination.

[0053] The program instructions recorded on the above-mentioned computer-readable recording medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of computer software.

[0054] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, and flash memory.

[0055] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.

[0056] Although various embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. It is understood that various modifications can be made by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention.

[0057] [Explanation of the symbol]

[0058] 110 : MRI selection section

[0059] 120: Segmentation section

[0060] 130: Conversion unit

[0061] 140: Image Quality Improvement Section

Claims

1. A device that converts contrast-free Magnetic Resonance Imaging (MRI) into contrast-enhanced MRI using artificial intelligence, An MRI selection unit that, when a patient identifier (ID) is entered, retrieves a list of MRIs corresponding to the patient ID and selects one of the MRIs; A segmentation unit that segments a region of interest by applying a pre-trained algorithm to the selected contrast-agent-free MRI; A conversion unit that applies a model trained to generate a contrast-enhanced MRI of the same anatomical location from a contrast-free MRI to convert the region of interest of the segmented contrast-free MRI into a contrast-enhanced MRI of the same anatomical location; An MRI conversion device comprising: an image quality improvement unit that improves the image quality of an MRI using the above-described converted contrast agent.

2. In Paragraph 1, The above model is trained by a generator and a discriminator, and An MRI conversion device characterized by the above-mentioned generator learning a contrast-free MRI and a contrast-enhanced MRI of the same anatomical location to generate a contrast-enhanced MRI of the same anatomical location as the selected contrast-free MRI, the above-mentioned discriminator comparing the generated contrast-enhanced MRI with the contrast-enhanced MRI used in the learning and feeding back the error value between the comparison MRIs to the above-mentioned generator, and the above-mentioned generator generating a contrast-enhanced MRI of the same anatomical location as the selected contrast-free MRI by reflecting the error value.

3. In Paragraph 1, The contrast-free MRI used for training the above model is an MRI in which the region of interest is segmented through the above-mentioned pre-trained algorithm, and An MRI conversion device characterized in that the contrast-enhanced MRI used for training the above model has a region identical to the region of interest of the segmented contrast-unenhanced MRI, which is the MRI segmented from the contrast-enhanced MRI.

4. In Paragraph 3, An MRI conversion device characterized in that the above region of interest is an area in which organ regions where movement over time is detected in the above contrast agent-free MRI and contrast agent-used MRI are excluded.

5. In Paragraph 1, The above image quality improvement unit is, An MRI conversion device that removes noise and blurring from the above-described converted contrast agent MRI and improves image quality by adjusting the resolution through the application of super resolution.

6. A method for converting contrast-free Magnetic Resonance Imaging (MRI) into contrast-enhanced MRI using artificial intelligence, When a patient identifier (ID) is entered, a step of retrieving a list of MRIs corresponding to the patient ID and selecting one of the MRIs; A step of segmenting the region of interest by applying a pre-trained algorithm to the selected contrast-agent-free MRI; A step of applying a model trained to generate a contrast-enhanced MRI of the same anatomical location from a contrast-free MRI to convert the region of interest of the segmented contrast-free MRI into a contrast-enhanced MRI of the same anatomical location; MRI conversion method comprising the step of improving the image quality of the MRI using the converted contrast agent.

7. In Paragraph 6, The above model is trained by a generator and a discriminator, and An MRI conversion method characterized by the above-mentioned generator learning a contrast-free MRI and a contrast-enhanced MRI of the same anatomical location to generate a contrast-enhanced MRI of the same anatomical location as the selected contrast-free MRI, the above-mentioned discriminator comparing the generated contrast-enhanced MRI with the contrast-enhanced MRI used in the learning, feeding back the error value between the comparison MRIs to the above-mentioned generator, and the above-mentioned generator generating a contrast-enhanced MRI of the same anatomical location as the selected contrast-free MRI by reflecting the error value.

8. In Paragraph 6, The contrast-free MRI used for training the above model is an MRI in which the region of interest is segmented through the above-mentioned pre-trained algorithm, and An MRI conversion method characterized in that the contrast-enhanced MRI used for training the above model has a region identical to the region of interest of the segmented contrast-unenhanced MRI, which is the MRI segmented from the contrast-enhanced MRI.

9. In Paragraph 8, An MRI conversion method characterized in that the above-mentioned region of interest is an area in which organ regions where movement over time is detected in the above-mentioned MRI without contrast agent and MRI with contrast agent are excluded.

10. In Paragraph 6, The step of improving the image quality of the MRI using the converted contrast agent described above is, An MRI conversion method that removes noise and blurring from the above-described converted contrast agent MRI and improves image quality by adjusting the resolution through the application of super-resolution.

Images