High-speed magnetic resonance imaging apparatus using k-space variable sampling

Abstract

A magnetic resonance imaging (MRI) apparatus according to the present invention comprises: a processor configured to acquire a radio frequency (RF) signal from the MRI apparatus, generate K-space data by performing a Fourier transform on the acquired RF signal, sample K-space data having a frequency magnitude less than a first frequency magnitude at a first sampling rate and sample K-space data having a frequency magnitude greater than or equal to the first frequency magnitude at a second sampling rate less than the first sampling rate, on the basis of a frequency magnitude of the K-space data, and when the first and second sampling rates are less than a predetermined sampling rate, apply the sampled data to a predetermined learned first artificial intelligence model to add and interpolate sampling data, and generate a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampling data; and a memory configured to store the generated magnetic resonance image.

Description

High-speed magnetic resonance imaging device using K-space variable sampling

[0001] The present invention relates to an apparatus and method for acquiring high-accuracy magnetic resonance images at high speed using K-space variable sampling.

[0002] Magnetic Resonance Imaging (MRI) is a state-of-the-art medical imaging diagnostic device that utilizes powerful magnetic fields and high frequencies to non-invasively observe anatomical structures within the human body. In particular, MRI devices can image soft tissues such as the brain, spine, and joints with high resolution, playing an essential role in the diagnosis of various diseases including cancer, stroke, and spinal disorders. Furthermore, because it poses no risk of radiation exposure and can provide cross-sectional images from various angles, it is widely utilized in modern medical systems and contributes to improving the accuracy and reliability of diagnoses through continuous technological advancements.

[0003] Current MRI devices acquire images by sampling all frequency bands evenly in a frequency space called K-space. While this uniform sampling method has the advantage of obtaining complete image information, it has a fundamental limitation in that data acquisition takes a considerable amount of time. In particular, since the patient must remain stationary without moving during the long measurement time, there is a problem in that the image quality can be significantly degraded by natural physiological movements such as breathing or heartbeats, or by involuntary movements of the patient.

[0004] To overcome these technical limitations, the present invention proposes a novel sampling method that takes into account the characteristics of K-space. The invention proposes an innovative approach that samples the low-frequency region of K-space, where key feature information of major parts of the human body and lesions is concentrated, at high density, while applying an interpolation technique utilizing undersampling and artificial intelligence to the high-frequency region, which contains relatively secondary information. This can be evaluated as a balanced solution that maximizes data acquisition efficiency while maintaining image quality.

[0005] The high-speed magnetic resonance imaging device and method using K-space variable sampling according to the present invention can drastically reduce measurement time compared to conventional methods. This has the advantage of significantly reducing the risk of image quality degradation caused by patient movement and greatly reducing the burden and discomfort of the patient during examination. In particular, by utilizing artificial intelligence technology for reconstruction of the high-frequency region, detailed information of the image can be accurately reproduced, thereby enabling the rapid provision of high-quality diagnostic images required clinically. This enables rapid diagnosis by medical staff and can greatly improve the efficiency of MRI equipment utilization in medical institutions.

[0006] The technical problem to be solved by the present invention is to provide a high-speed magnetic resonance imaging device using K-space.

[0007] Another technical objective of the present invention is to provide a method for acquiring magnetic resonance images at high speed using K-space.

[0008] Another technical objective of the present invention is to provide a computer-readable recording medium that records a program for executing a method of acquiring magnetic resonance images at high speed using K-space on a computer.

[0009] The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0010] A magnetic resonance imaging (MRI) device according to the present invention for achieving the above technical problem may include a processor that acquires a radio frequency (RF) signal in the magnetic resonance imaging device, generates K-space data by performing a Fourier transform on the acquired RF signal, samples K-space data with a frequency magnitude less than a first frequency magnitude based on the frequency magnitude of the K-space data at a first sampling rate, samples K-space data with a frequency magnitude greater than or equal to the first sampling rate at a second sampling rate smaller than the first sampling rate, and when the first and second sampling rates are less than a predetermined sampling rate, applies the sampled data to a predetermined learned first artificial intelligence model to add and interpolate the sampled data, and generates a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampled data, and a memory that stores the generated magnetic resonance image.

[0011] The processor may first generate a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are greater than or equal to the predetermined sampling rate, and generate a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are less than the predetermined sampling rate and on the interpolated sampling data.

[0012] The processor may apply at least one of a signal stabilization filter and a local averaging filter to K-space data of less than the first frequency magnitude, and apply at least one of a differential filter and a Gaussian filter to K-space data of greater than the first frequency magnitude.

[0013] The first and second sampling rates mentioned above may be determined based on an exponential function that decreases as the frequency magnitude increases, or a logarithmic function that is inversely proportional to the frequency magnitude.

[0014] The above-mentioned predetermined first artificial intelligence model may be a model based on SCNN (Spiking Convolutional Neural Network).

[0015] A method for generating a magnetic resonance imaging (MRI) according to the present invention for achieving the above technical problem may include: a step in which a processor acquires a radio frequency (RF) signal from a magnetic resonance imaging device; a step in which the processor generates K-space data by performing a Fourier transform on the acquired RF signal; a step in which the processor samples K-space data with a frequency magnitude less than a first frequency magnitude at a first sampling rate based on the frequency magnitude of the K-space data; a step in which the processor samples K-space data with a frequency magnitude greater than or equal to the first frequency magnitude at a second sampling rate smaller than the first sampling rate; a step in which, if the first and second sampling rates are less than a predetermined sampling rate, the processor applies the sampled data to a predetermined learned first artificial intelligence model to add and interpolate the sampled data; a step in which the processor generates a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampled data; and a step in which a memory stores the generated magnetic resonance image.

[0016] The processor may include the step of first generating a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are greater than or equal to the predetermined sampling rate, and the step of generating a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are less than the predetermined sampling rate and on the interpolated sampling data.

[0017] The processor may include the step of applying at least one of a signal stabilization filter and a local averaging filter to K-space data of less than the first frequency magnitude, and the step of the processor applying at least one of a differential filter and a Gaussian filter to K-space data of greater than the first frequency magnitude.

[0018] The first and second sampling rates mentioned above may be determined based on an exponential function that decreases as the frequency magnitude increases, or a logarithmic function that is inversely proportional to the frequency magnitude.

[0019] The above-mentioned predetermined first artificial intelligence model may be a model based on SCNN (Spiking Convolutional Neural Network).

[0020] The high-speed magnetic resonance imaging device and method using K-space variable sampling according to the present invention can drastically reduce measurement time compared to conventional methods by using a variable sampling method based on the frequency magnitude of K-space data.

[0021] The high-speed magnetic resonance imaging device and method using K-space variable sampling according to the present invention can provide the effect of maintaining image quality while maximizing data acquisition efficiency by applying an undersampling technique and an interpolation technique utilizing artificial intelligence technology to a high-frequency region containing incidental information.

[0022] The high-speed magnetic resonance imaging device and method using K-space variable sampling according to the present invention can provide the effect of enabling medical staff to make accurate diagnoses by reducing the cost of re-imaging due to patient movement with a short measurement time compared to conventional methods and providing high image quality.

[0023] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below.

[0024] The accompanying drawings, which are included as part of the detailed description to aid in understanding the present invention, provide embodiments of the present invention and explain the technical concept of the present invention together with the detailed description.

[0025] Figure 1 is a diagram illustrating the layer structure of an artificial neural network.

[0026] Figure 2 is a diagram illustrating an example of a deep neural network.

[0027] Figure 3 is an example diagram illustrating the data processing process of a conventional magnetic resonance imaging device.

[0028] FIG. 4 is a diagram illustrating a block diagram for explaining the function of a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention.

[0029] FIG. 5 is an example diagram illustrating the data processing process of a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention.

[0030] FIG. 6 is a diagram illustrating a function of the sampling ratio according to the frequency magnitude used in a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention.

[0031] The present invention is capable of various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the invention to specific embodiments, and it should be understood that the invention includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention.

[0032] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0033] Terms such as "first," "second," etc., may be used to describe various components, but said components should not be limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another.

[0034] The terms used herein are merely for describing specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as “comprising” or “having” are intended to indicate the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0035] Additionally, terms such as “…part,” “…unit,” “…module,” and “…device” described in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0036] In some cases, to avoid obscuring the concept of the present invention, known structures and devices may be omitted or illustrated in the form of block diagrams focusing on the core functions of each structure and device. Additionally, throughout this specification, the same components are described using the same reference numerals.

[0037] Furthermore, the components of the embodiments described with reference to each drawing are not limited to the respective embodiments and may be implemented to be included in other embodiments within the scope of maintaining the technical spirit of the present invention. It is also obvious that multiple embodiments may be re-implemented as a single embodiment that integrates multiple embodiments, even if a separate description is omitted.

[0038] Before describing the present invention, we will explain artificial intelligence (AI), machine learning, and deep learning. The easiest way to understand the relationship between these three concepts is to visualize three concentric circles. Artificial intelligence is the largest circle, followed by machine learning, and deep learning, which is leading the current AI boom, can be considered the smallest circle.

[0039] The concept of artificial intelligence first emerged at the Dartmouth Conference hosted by Professor John McCarthy at Dartmouth College in the United States in 1956, and it has been growing explosively in recent years. This growth has been further accelerated, particularly since 2015, by the introduction of GPUs that provide rapid and powerful parallel processing capabilities. The advent of the Big Data era, characterized by explosively increasing storage capacity and a flood of data across all domains—including images, text, and mapping data—has also had a significant impact on this growth trend.

[0040] Artificial Intelligence - Realizing human intelligence in machines

[0041] In 1956, the pioneers of artificial intelligence dreamed of ultimately creating complex computers with characteristics similar to human intelligence. While artificial intelligence that thinks like a human, possessing human senses and thinking abilities, is called 'General AI,' the artificial intelligence achievable at the current level of technological development falls under the concept of 'Narrow AI.' Narrow AI is characterized by its ability to perform specific tasks with capabilities exceeding those of humans, such as image classification services on social media or facial recognition functions.

[0042] Machine Learning - A Specific Approach to Implementing Artificial Intelligence

[0043] Machine learning serves the role of automatically filtering spam from your inbox. Meanwhile, machine learning fundamentally uses algorithms to analyze data, learns through analysis, and performs judgments or predictions based on what it has learned. Therefore, its ultimate goal is not to directly code specific guidelines for decision criteria into the software, but rather to 'train' the computer itself through massive amounts of data and algorithms to learn how to perform tasks. Machine learning originated from concepts directly proposed by early artificial intelligence researchers, and its algorithmic methods include decision tree learning, inductive logic programming, clustering, reinforcement learning, and Bayesian networks. However, none of these have achieved general AI, which can be considered the ultimate goal, and it is true that early machine learning approaches often struggled to complete even narrow AI.

[0044] Currently, machine learning is achieving significant results in fields such as computer vision, but it has encountered a limitation in that a certain amount of coding work is involved throughout the entire process of implementing artificial intelligence, even without specific guidelines. For instance, when recognizing an image of a stop sign based on a machine learning system, the developer must directly code boundary detection filters that programmatically identify the start and end points of an object, shape detection systems that verify the surface of an object, and classifiers that recognize characters such as 'STO-P'. In this way, machine learning operates by recognizing images from 'coded' classifiers and 'learning' stop signs through algorithms.

[0045] Machine learning training methods find the most suitable model by adjusting the model parameters in a direction that minimizes the error between the target value and the predicted value. Here, the predicted value refers to the value produced when an input value is fed into the model, i.e., the output value. For example, a model composed of an arbitrary number of convolution layers, bidirectional LSTMs, and feedforward layers modifies each convolution layer, bidirectional LSTM, and feedforward layer as training progresses so that it becomes a model with the smallest error from the target value.

[0046] While machine learning achieves sufficient performance for commercialization in image recognition, the accuracy can drop in specific situations where signs are obscured by fog or trees. The reason computer vision and image recognition have not yet reached human levels until recently is due to these recognition rate issues and frequent errors.

[0047] Deep Learning - A technology that enables complete machine learning

[0048] The biological characteristics of the human brain, particularly the connection structure of neurons, inspired artificial neural networks, another algorithm created by early machine learning researchers. However, unlike the brain, where any physically adjacent neurons can be interconnected, artificial neural networks have fixed layer connections and data propagation directions.

[0049] For example, when an image is cut into numerous tiles and input into the first layer of a neural network, the neurons repeat the process of passing data to the next layer until a final output is generated at the last layer. Each neuron is assigned a weight representing the accuracy of the input based on the task performed, and the final output is determined by summing all the weights. In the case of a stop sign, the image's characteristics—such as its octagonal shape, red color, text, size, and movement—are finely cut and 'inspected' by the neurons, and the neural network's task is to identify whether it is a stop sign. Here, a 'probability vector' is utilized to predict the result based on weights derived from sufficient data.

[0050] Deep learning is a form of artificial intelligence that has evolved from artificial neural networks, utilizing information input and output layers similar to the neurons in the brain to learn data. However, because even basic neural networks require a massive amount of computation, the commercialization of deep learning faced obstacles from the beginning. Nevertheless, researchers continued their work and succeeded in parallelizing algorithms that prove the concept of deep learning based on supercomputers. Furthermore, the emergence of GPUs, which are optimized for parallel processing, dramatically accelerated the computational speed of neural networks, leading to the advent of true deep learning-based artificial intelligence.

[0051] Neural networks are highly likely to produce numerous incorrect answers during the 'learning' process. Returning to the example of the stop sign, to precisely adjust the weights of neuron inputs to always produce the correct answer regardless of weather conditions or day-night cycles, one might need to learn from hundreds, thousands, or perhaps even millions of images. Only when this level of accuracy is reached can the neural network be considered to have properly learned the stop sign. In 2012, Google and Stanford University Professor Andrew Ng implemented a 'Deep Neural Network' consisting of over 1 billion neural networks using 16,000 computers. Through this, they extracted and analyzed 10 million images from YouTube and succeeded in having the computer classify photos of people and cats. They enabled the computer to independently learn the process of recognizing and judging the shape and appearance of cats appearing in the videos.

[0052] The image recognition capabilities of systems trained with deep learning have already surpassed those of humans. Furthermore, the scope of deep learning extends to areas such as identifying cancer cells in the blood and tumors in MRI scans. Google's AlphaGo learned the fundamentals of Go and further strengthened its neural network through the process of repeatedly playing matches against AIs similar to itself. The emergence of deep learning has enhanced the practicality of machine learning and expanded the scope of artificial intelligence. Deep learning subdivides tasks in every way possible that can be supported by computer systems. Deep learning-based technologies, such as driverless cars, improved preventive medicine, and more accurate movie recommendations, are already being used in our daily lives or are on the verge of practical application. Deep learning is regarded as both the present and the future of artificial intelligence, possessing the potential to realize the general AI that once appeared in science fiction.

[0053] Below, we will take a closer look at deep learning.

[0054] Deep learning is a type of artificial neural network (ANN) based on human neural network theory. It is a set of machine learning models or algorithms that refer to a deep neural network (DNN) composed of a layer structure and having one or more hidden layers (hereinafter referred to as intermediate layers) between the input layer and the output layer. Simply put, deep learning can be described as an artificial neural network with deep layers.

[0055] The human brain is estimated to be composed of 25 billion nerve cells. The brain consists of nerve cells, and each nerve cell (neuron) refers to a single nerve cell that forms a neural network. A nerve cell contains a cell body, a single axon (or nurite) which is a projection of the cell body, and usually several dendrites (or protoplasmic processes). Information exchange between these nerve cells is transmitted through junctions between nerve cells called synapses. While a single nerve cell appears very simple when viewed in isolation, when these nerve cells come together, they are capable of possessing human intelligence. The dendrites are the part that receives signals sent by other nerve cells (Input), while the axon is the long extension from the cell body that transmits signals to other nerve cells (Output). There is a connection called a synapse that links the axon and dendrite, which transmit signals between nerve cells; however, the signal is not transmitted unconditionally, but is only transmitted when the signal strength exceeds a certain value (threshold). In other words, not only is the connection strength different for each synapse, but it also determines whether or not to transmit a signal.

[0056] Artificial neural networks (ANNs), a field of artificial intelligence, are mathematical models modeled by mimicking the structure of the biological (typically human) brain (neural networks). In other words, artificial neural networks are implemented by imitating the information processing and transmission processes of these biological neurons. As they are implemented similarly to how the human brain solves problems, neural networks possess excellent parallelism because each neuron operates independently. Furthermore, since information is distributed across numerous connections, problems in a few neurons do not significantly affect the entire network; consequently, they are resilient to a certain level of error and possess the ability to learn from a given environment.

[0057] Deep neural networks can be viewed as descendants of artificial neural networks. They are the latest version of artificial neural networks, having overcome existing limitations and achieved success in areas where numerous artificial intelligence technologies had previously failed. When examining the modeling of artificial neural networks that mimic biological neural networks, biological neurons are modeled as nodes in terms of processing units, and synapses are modeled as weights in terms of connections, as shown in Table 1 below.

[0058] Biological Neural Network Artificial Neural Network Cell Body Node Dendrite Input Axon Output Synapse Weight

[0059] Figure 1 is a diagram illustrating the layer structure of an artificial neural network. Just as human biological neurons are connected in multiple numbers to perform meaningful tasks, artificial neural networks also connect individual neurons to each other through synapses, creating multiple interconnected layers, and the connection strength between each layer can be updated using weights. In this way, the multi-layer structure and connection strengths are utilized in fields for learning and cognition.

[0060] Each node is connected by weighted links, and the entire model learns by repeatedly adjusting these weights. Weights represent the importance of each node as a fundamental means for long-term memory. Simply put, an artificial neural network trains the entire model by initializing these weights and updating and adjusting them with the data set to be trained. Once training is complete, when a new input is received, it infers an appropriate output value. The learning principle of an artificial neural network can be viewed as the process by which intelligence is formed from the generalization of experience, and it operates in a bottom-up manner. In Figure 1, when there are two or more intermediate layers (i.e., 5 to 10), the layers are considered to be deep, and it is called a Deep Neural Network; the learning and inference model achieved through such a Deep Neural Network can be referred to as Deep Learning.

[0061] Artificial neural networks can perform a certain role even with only one intermediate layer (commonly referred to as a 'hidden layer') in addition to inputs and outputs, but as the complexity of the problem increases, the number of nodes or layers must be increased. Among these, adopting a multi-layered model by increasing the number of layers is effective, but its scope of application is limited due to the limitations that efficient learning is impossible and the amount of computation required to train the network is large.

[0062] However, as the existing limitations mentioned above have been overcome, artificial neural networks have become capable of adopting deep structures. This has enabled the construction of complex and highly expressive models, leading to the 발표 of groundbreaking results in various fields such as speech recognition, face recognition, object recognition, and character recognition.

[0063] Figure 2 is a diagram illustrating an example of a deep neural network.

[0064] A Deep Neural Network (DNN) is an Artificial Neural Network (ANN) composed of multiple hidden layers between an input layer and an output layer. It is a set of machine learning models or algorithms referring to a Deep Neural Network (DNN) that has one or more hidden layers between an input layer and an output layer. Connections in a neural network are formed from the input layer to the hidden layer, and from the hidden layer to the output layer.

[0065] Deep neural networks, like general artificial neural networks, can model complex non-linear relationships. For example, in a deep neural network structure for an object identification model, each object can be represented as a hierarchical composition of the basic elements of an image. In this case, additional layers can combine features from progressively gathered lower layers. This characteristic of deep neural networks enables the modeling of complex data with fewer units (nodes) compared to similarly performed artificial neural networks.

[0066] Previous deep neural networks were typically designed as feedforward networks, but recent research has successfully applied deep learning structures to Recurrent Neural Networks (RNNs). Examples include the application of deep neural network structures in the field of language modeling. In the case of Convolutional Neural Networks (CNNs), not only have they been successfully applied in the field of computer vision, but their successful applications are also well-documented. More recently, CNNs have been applied to acoustic modeling for Automatic Speech Recognition (ASR) and are considered to have been more successful than existing models. Deep neural networks can be trained using the standard backpropagation algorithm. In this process, weights can be updated through stochastic gradient descent.

[0067] Various signals from the surrounding environment received by humans through their sensory organs can be expressed via a computer in the form of text, audio, images, and videos, and stored as data in the computer's internal storage device.

[0068] The high-dimensional data corresponding to the aforementioned text, audio, images, and videos stored in a computer consists of combinations of continuous '0's and '1's from a low-dimensional perspective, while from the perspective of a slightly higher-dimensional computer program, it consists of various structures, objects, or class instances defined in the programming language used by each program.

[0069] For existing AI technologies to learn, they must extract features—data capable of effectively representing high-dimensional data—from high-dimensional data such as text, audio, images, and videos that humans can process via computers; however, the methods for implementing and the terminology used to refer to this feature data differ across various AI models and the programming languages ​​capable of implementing them.

[0070] The MRI and key related technologies used in the present invention are described below.

[0071] Magnetic Resonance Imaging (MRI) device

[0072] Magnetic Resonance Imaging (MRI) is a state-of-the-art medical imaging diagnostic device that uses strong magnetic fields and high frequencies to image the inside of the human body. It utilizes the phenomenon where hydrogen nuclei within a patient's body resonate at specific frequencies within a magnetic field to precisely observe the structure and condition of tissues. Particularly useful for diagnosing conditions such as tumors, inflammation, and injuries by clearly visualizing soft tissues like the brain, spine, and joints, it is recognized as a harmless examination method as it does not use X-rays or radiation.

[0073] The main components of a magnetic resonance imaging (MRI) device consist of a superconducting magnet that generates a strong magnetic field, an RF coil that generates high-frequency pulses, a gradient coil that controls the direction of the magnetic field, and a computer system that receives and processes signals. The superconducting magnet is cooled with liquid helium to maintain an ultra-low temperature state, thereby generating a strong and uniform magnetic field. The gradient coil controls the gradient of the magnetic field to obtain accurate positional information in three-dimensional space, while the RF coil plays the role of exciting hydrogen nuclei and detecting the emitted signals.

[0074] MRI examinations are performed with the patient lying inside a cylindrical tunnel-shaped device, and various image acquisition methods are applied depending on the examination area and purpose. T1-weighted images are suitable for observing anatomical structures, T2-weighted images are useful for detecting lesions or inflammation, and diffusion-weighted imaging is essential for diagnosing acute diseases such as stroke. Additionally, the condition of blood vessels or tumors can be assessed more accurately by injecting contrast agents, and it is also possible to observe brain activity in real time through functional magnetic resonance imaging.

[0075] Magnetic resonance imaging technology is continuously advancing and evolving to reduce examination times and improve image quality. With the widespread adoption of high-field MRI of 3 Tesla or higher, it has become possible to obtain clearer images. In particular, advancements in real-time image acquisition technology have enabled dynamic examinations of the heart and joints, and the development of open MRI has allowed patients with claustrophobia and pediatric patients to undergo examinations more comfortably.

[0076] K-Space

[0077] K-space refers to a two- or three-dimensional frequency space in magnetic resonance imaging where original data is stored. Signals acquired during an MRI scan are measured in the time domain and recorded in K-space through Fourier transform. The center of K-space contains low-frequency components that determine image contrast and overall signal intensity, while the periphery contains high-frequency components that determine the detailed resolution of the image. Each point in K-space corresponds to a specific spatial frequency, and this frequency information combines to form the final MR image.

[0078] The K-space data acquisition method directly affects image quality and scan time. The most basic method is Cartesian sampling, which fills the K-space sequentially row by row. Radial sampling acquires data radially from the center outwards, while spiral sampling collects data along a spiral trajectory. Each sampling method has its own unique advantages and disadvantages, and the appropriate method is selected based on the purpose of the examination and the patient's condition. In particular, radial or spiral sampling can reduce image distortion caused by movement, making them useful for imaging areas with physiological movements such as breathing or heartbeats.

[0079] The processing of K-space data involves several stages of complex mathematical operations. The acquired signal first undergoes phase correction and amplitude correction, and is subsequently converted into a real image through a 2D or 3D inverse Fourier transform. During this process, various filtering techniques are applied to remove noise and improve image quality.

[0080] The low-frequency region of K-space is located at the center and contains important information that determines the overall contrast and signal intensity of the image. This region enables the differentiation of the brain's overall structure and major regions such as the cerebrum, cerebellum, and brainstem, while in abdominal imaging, it allows for the identification of the location and size of parenchymal organs such as the liver, kidneys, and spleen. Furthermore, it allows for the confirmation of the presence and approximate extent of large lesions, such as tumors or cysts, and enables the differentiation between normal tissue and lesions in T1 and T2-weighted images by reflecting the basic signal characteristics of the tissue. Particularly in images after contrast agent injection, signal changes in the low-frequency region accurately reflect the degree of contrast enhancement of tumors or inflammatory lesions, playing a key role in evaluating the characteristics and activity of the lesions.

[0081] The high-frequency region of K-space is located in the periphery and determines the detailed structure and boundary sharpness of the image. This region contains information regarding abrupt signal changes between tissues and fine structures, and is directly related to image resolution. Data in the high-frequency region has relatively low signal intensity and is susceptible to noise; therefore, various filtering techniques can be applied to correct for this. Particularly in brain imaging, information from the high-frequency region plays a crucial role in identifying the boundaries between white and gray matter, small vascular structures, and microscopic lesions, thereby enabling accurate diagnosis.

[0082] Fourier Transform and Inverse Fourier Transform

[0083] The Fourier Transform is a mathematical process that converts time-domain signals acquired from MRI into the frequency domain. It serves to decompose the time-dependent signals emitted by hydrogen nuclei in the patient's body, excited by RF pulses, into frequency components and map them to K-space. During this process, data in the form of complex numbers is generated, while both the amplitude and phase information of the signal are preserved. Particularly when acquiring 2D or 3D images, the Fourier Transform is applied sequentially to each direction, thereby accurately recording spatial frequency information at each point in K-space. This transformation process is a critical step in converting spatial information provided by the gradient magnetic field into frequency information.

[0084] The Inverse Fourier Transform is a process that transforms frequency domain data stored in K-space into actual anatomical images. It reconstructs the signal intensity distribution in the spatial domain by combining frequency components stored at each point in K-space; in this process, low-frequency components determine the overall contrast of the image, while high-frequency components determine the detailed resolution. Various filtering techniques can be applied to remove noise and artifacts that may occur during the transformation process.

[0085] Spiking Convolution Neural Network (SCNN)

[0086] Spiking Convolutional Neural Networks (SCNNs) are a type of artificial neural network that mimics the operation of biological neurons. Unlike conventional CNNs, they utilize a method where neurons transmit information through binary spike signals. They process the temporal patterns of spikes entering through synapses and are characterized by generating spikes only when the neuron's membrane potential exceeds a threshold. Due to these characteristics, they are highly energy-efficient and suitable for real-time processing, demonstrating superior performance particularly in processing time-series data such as visual and auditory signals. Furthermore, they consume significantly less power than conventional CNNs, making them highly applicable in mobile devices and edge computing environments. By incorporating the characteristics of biological neural networks, they enable more natural pattern recognition and learning.

[0087] The present invention proposes an apparatus and method for extracting high-quality magnetic resonance images at high speed by applying variable sampling according to the K-space frequency domain.

[0088] Figure 3 is an example diagram illustrating the data processing process of a conventional magnetic resonance imaging device.

[0089] Referring to Fig. 3, RF signal acquisition is the first process performed in a magnetic resonance imaging device, which involves acquiring high-frequency signals generated as hydrogen nuclei inside the human body resonate within a magnetic field through a receiving coil. Since the signal acquired during this process is very weak, sophisticated amplification and filtering processes may be included.

[0090] The Fourier Transform is a process that converts RF signals measured in the time domain into the frequency domain, enabling the analysis of the signal's frequency components and the extraction of spatial location information. In this transformation process, the Fast Fourier Transform algorithm is generally used, preserving both the phase and amplitude information of the signal.

[0091] The K-space data generation step is the process of mapping Fourier-transformed data into a two-dimensional or three-dimensional frequency space. In this step, phase encoding and frequency encoding information are combined to assign data to each point in K-space, which is an important factor in determining the resolution and quality of the final image.

[0092] The sampling process is a step for extracting K-space data, aiming to preserve important information while reducing data size. However, since conventional magnetic resonance imaging (MRI) devices perform sampling at a fixed rate, imaging takes a significant amount of time.

[0093] The Inverse Fourier Transform (IFF) step is the process of converting sampled K-space data back into the spatial domain. Typically, the Fast Inverse Fourier Transform algorithm is used in this process, and the converted data is represented as grayscale values ​​that reflect actual anatomical structures.

[0094] The magnetic resonance imaging (MRI) generation stage is the process of reconstructing inverse Fourier transformed data into an actual image. In this stage, various image processing techniques, such as contrast adjustment, noise removal, and resolution enhancement, can be applied, generating images of a quality suitable for diagnosis.

[0095] The memory storage step is the process of storing the finally generated magnetic resonance image in digital form.

[0096] FIG. 4 is a diagram illustrating a block diagram for explaining the function of a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention.

[0097] Referring to FIG. 4, a high-speed magnetic resonance imaging device (400) using K-space variable sampling may include a processor (410) and a memory (420).

[0098] FIG. 5 is an example diagram illustrating the data processing process of a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention.

[0099] The processor (410) can acquire a radio frequency (RF) signal from a magnetic resonance imaging device and generate K-space data by performing a Fourier transform on the acquired RF signal, which corresponds to the steps of "acquiring RF signal," "fourier transform," and "generating K-space data" in FIG. 5.

[0100] Based on the frequency magnitude of the K-space data, the processor (410) may sample K-space data with a frequency magnitude less than a first frequency magnitude at a first sampling rate, and K-space data with a frequency magnitude greater than the first frequency magnitude at a second sampling rate smaller than the first sampling rate. This corresponds to the step of comparing "frequency magnitude and first frequency magnitude" in FIG. 5, and the "sampling at the first sampling rate" step and the "sampling at the second sampling rate" step (510) according to the comparison result.

[0101] FIG. 6 is a diagram illustrating a function of the sampling ratio according to the frequency magnitude used in a high-speed magnetic resonance imaging device (400) using K-space variable sampling according to the present invention. Here, e is the natural constant, a is the decay coefficient, which is a constant that controls the decay rate of the function, f is the frequency magnitude, and log is the logarithmic function (natural logarithm or common logarithm).

[0102] Referring to FIG. 6, the processor (410) can be determined based on an exponential function that decreases as the frequency magnitude increases or a logarithmic function that is inversely proportional to the frequency magnitude, and e^(-af) is shown as an example of an exponential function and 1 / log(f+1) is shown as an example of a logarithmic function that is inversely proportional to the frequency magnitude. When the processor (410) uses a sampling rate according to the frequency magnitude according to the e^(-af) function corresponding to the exponential function, a sampling rate of about 75% is used when the frequency magnitude is 5 and a sampling rate of about 50% is used when the frequency magnitude is 10, so the sampling rate applied to K-space data with a frequency magnitude of 10 is about 25% less than that applied to K-space data with a frequency magnitude of 5.

[0103] When the first and second sampling rates are less than a predetermined sampling rate, the processor (410) can apply the sampled data to a predetermined learned first artificial intelligence model to add and interpolate the sampled data, which corresponds to the step of comparing the "sampling rate and the reference sampling rate" in FIG. 5, and the step of "data interpolation through the AI ​​model" and the step of "sampling data retention" (520) according to the comparison result.

[0104] Referring to FIG. 6, the processor (410) uses a sampling rate according to the frequency magnitude according to the e^(-af) function corresponding to the exponential function, and when the predetermined sampling rate is 25%, K-space data with a frequency magnitude of 20 uses a sampling rate of less than 25%, so additional sampling data can be generated and interpolated by applying it to a predetermined trained first artificial intelligence model. Here, the predetermined trained first artificial intelligence model may be a model based on SCNN (Spiking Convolutional Neural Network).

[0105] The number of sampling data interpolated through a predetermined first artificial intelligence model may be the difference between the first and second sampling rates and a predetermined sampling rate, or it may be another predetermined sampling rate. For example, if the predetermined sampling rate is 25% and the sampling rate corresponding to a frequency magnitude of 19 is 21%, the number of sampling data interpolated through the first artificial intelligence model may be the number corresponding to a sampling rate of 4%, or if the predetermined interpolated sampling rate is 30%, it may be the number corresponding to 9%, which is the difference from the sampling rate corresponding to a frequency magnitude of 19 (21%).

[0106] The processor (410) can generate a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampled data, which corresponds to the "inverse Fourier transform" step and the "generated magnetic resonance image" step of FIG. 5.

[0107] The memory (420) can store the generated magnetic resonance image, which corresponds to the "store in memory" step of FIG. 5.

[0108] In the process of generating a magnetic resonance image by inverse Fourier transforming the sampled data and the interpolated sampled data, the processor (410) may first generate a magnetic resonance image by inverse Fourier transforming the sampled data in which the first and second sampling ratios are greater than or equal to the predetermined sampling ratio, and then generate a magnetic resonance image later by inverse Fourier transforming the sampled data and the interpolated sampled data in which the first and second sampling ratios are less than the predetermined sampling ratio.

[0109] When the processor (410) generates a magnetic resonance image by first performing an inverse Fourier transform on low-frequency size sampling data that does not use interpolated sampling data, multi-resolution-based data processing is performed, thereby providing the effect of rapidly outputting a magnetic resonance image in real time while reconstructing detailed images step by step.

[0110] The processor (410) can apply at least one of a signal stabilization filter and a local averaging filter to K-space data of less than a first frequency magnitude, and can apply at least one of a differential filter and a Gaussian filter to K-space data of greater than or equal to a first frequency magnitude.

[0111] The processor (410) can distinguish between K-space data to which at least one of a signal stabilization filter and a local averaging filter can be applied and K-space data to which at least one of a differential filter and a Gaussian filter can be applied based on a predetermined sampling rate or a predetermined frequency magnitude.

[0112] In this way, when the processor (410) applies a signal stabilization filter or a local averaging filter to K-space data corresponding to the low-frequency region, it can provide the effect of preventing distortion of the overall image structure and improving overall image uniformity. Additionally, when the processor (410) applies at least one of a differential filter or a Gaussian filter to K-space data corresponding to the high-frequency region, it can provide the effect of improving the visibility of microstructures and reducing unnecessary variability in the high-frequency region.

[0113] Referring to FIG. 6, for example, the processor (410) may apply at least one of a signal stabilization filter and a local averaging filter to K-space data with a sampling rate of 50% or more, and may apply at least one of a differential filter and a Gaussian filter to K-space data with a sampling rate of less than 50%.

[0114] Referring to FIG. 6, as another example, the processor (410) may apply at least one of a signal stabilization filter and a local averaging filter to K-space data with a frequency magnitude of 10 or less, and may apply at least one of a differential filter and a Gaussian filter to K-space data with a frequency magnitude greater than 10.

[0115] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include a plurality of processing elements and / or a plurality of types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0116] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computing devices and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0117] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. 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. 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 devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0118] The embodiments described above are combinations of the components and features of the present invention in a specific form. Each component or feature should be considered optional unless otherwise explicitly stated. Each component or feature may be implemented in a form not combined with other components or features. Additionally, it is possible to construct embodiments of the present invention by combining some components and / or features. The order of operations described in the embodiments of the present invention may be changed. Some components or features of one embodiment may be included in another embodiment, or may be replaced with corresponding components or features of another embodiment. It is obvious that embodiments may be constructed by combining claims that do not have an explicit citation relationship in the claims, or that new claims may be included by amendment after filing.

[0119] In the present invention, the processor (410) may be implemented by hardware, firmware, software, or a combination thereof. When implementing an embodiment of the present invention using hardware, the processor (410) may be equipped with ASICs (application specific integrated circuits) or DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), etc., configured to perform the present invention. It may also be implemented as a computer-readable recording medium that records a program for executing a method to prevent user information leakage during user authentication according to the present invention on a computer.

[0120] It is obvious to those skilled in the art that the present invention may be embodied in other specific forms without departing from the essential features of the invention. Accordingly, the foregoing detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

[0121] The high-speed magnetic resonance imaging device using K-space variable sampling according to the present invention can be used in industries such as the medical industry and the bio industry.

Claims

1. In a Magnetic Resonance Imaging (MRI) device, An RF (Radio Frequency) signal is acquired from the above magnetic resonance imaging device, and The above-mentioned acquired RF signal is Fourier transformed to generate K-Space data, and Based on the frequency magnitude of the above K-space data, K-space data with a frequency magnitude less than the first frequency magnitude is sampled at a first sampling rate, and K-space data greater than the first frequency magnitude is sampled at a second sampling rate smaller than the first sampling rate, and If the above first and second sampling rates are less than a predetermined sampling rate, the sampled data is applied to a predetermined trained first artificial intelligence model to add the sampled data and interpolate, and A processor that generates a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampled data; and A magnetic resonance imaging device comprising a memory for storing the magnetic resonance image generated above.

2. In Paragraph 1, The above processor is, First, a magnetic resonance image is generated by inverse Fourier transforming the sampling data in which the first and second sampling rates are greater than or equal to the predetermined sampling rates, and A magnetic resonance imaging device that generates a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are less than the predetermined sampling rates and the interpolated sampling data.

3. In Paragraph 1, The above processor is, At least one of a signal stabilization filter and a local averaging filter is applied to K-space data with a first frequency magnitude or less, and A magnetic resonance imaging device that applies at least one of a differential filter and a Gaussian filter to K-space data with a first frequency magnitude or greater.

4. In Paragraph 1, The above first and second sampling rates are, A magnetic resonance imaging device characterized by being determined based on an exponential function that decreases as the frequency magnitude increases or a logarithmic function that is inversely proportional to the frequency magnitude.

5. In Paragraph 1, The above-mentioned predetermined trained first artificial intelligence model is, A magnetic resonance imaging device based on an SCNN (Spiking Convolutional Neural Network) model.

6. In a method for generating Magnetic Resonance Imaging (MRI), A step in which a processor acquires an RF (Radio Frequency) signal from a magnetic resonance imaging device; A step in which the processor performs a Fourier transform on the acquired RF signal to generate K-Space data; The above processor samples K-space data with a frequency magnitude less than a first frequency magnitude at a first sampling rate based on the frequency magnitude of the K-space data; The processor samples K-space data greater than the first frequency magnitude at a second sampling rate smaller than the first sampling rate; If the first and second sampling rates are less than a predetermined sampling rate, the processor applies the sampled data to a predetermined learned first artificial intelligence model to interpolate by adding the sampled data; The step of the processor generating a magnetic resonance image by performing an inverse Fourier transform on the sampled data and the interpolated sampled data; and A method for generating a magnetic resonance image, comprising the step of storing the generated magnetic resonance image in memory.

7. In Paragraph 6, The above processor first generates a magnetic resonance image by performing an inverse Fourier transform on sampling data in which the first and second sampling rates are greater than or equal to the predetermined sampling rates, and A method for generating a magnetic resonance image, comprising the step of the processor generating a magnetic resonance image by performing an inverse Fourier transform on the sampling data in which the first and second sampling rates are less than the predetermined sampling rates and the interpolated sampling data.

8. In Paragraph 6, The step of the processor applying at least one of a signal stabilization filter and a local averaging filter to K-space data with a first frequency magnitude or less; and A method for generating a magnetic resonance image, comprising the step of the processor applying at least one of a differential filter and a Gaussian filter to K-space data with a first frequency magnitude or greater.

9. In Paragraph 6, The above first and second sampling rates are, A method for generating magnetic resonance images, characterized by being determined based on an exponential function that decreases as the frequency magnitude increases or a logarithmic function that is inversely proportional to the frequency magnitude.

10. In Paragraph 6, The above-mentioned predetermined trained first artificial intelligence model is, A method for generating magnetic resonance images using an SCNN (Spiking Convolutional Neural Network) based model.

11. A computer-readable recording medium storing a program for executing on a computer the method for generating a magnetic resonance image described in any one of paragraphs 6 through 10.

Images