KR102542008B1 - Method for generating ECV map by predicting hematocrit using Neural network - Google Patents

Abstract

The present invention relates to a method for generating an ECV map for determining myocardial disease, and is for generating a non-invasive based ECV map using a neural network, and the standard recovery rate of protons in heart tissue inverted with a radio frequency (RF) pulse. obtaining a T1 map generated by mapping the recovery time according to pixel units in a 2-dimensional space; extracting a blood T1 value from the T1 map; outputting an estimated hematocrit according to the extracted blood T1 value using a neural network trained to predict a blood hematocrit from the T1 value; and generating an extracellular volume index (ECV) map by using the output estimated red blood cell volume as a control variable of the T1 map. According to the present invention, it is possible to simplify the diagnosis process and increase the reliability of diagnosis by calculating the ECV index on a non-invasive basis.

Description

Method and apparatus for generating ECV map based on hematocrit prediction through neural network {Method for generating ECV map by predicting hematocrit using Neural network}

The present invention relates to a method for generating an ECV map for determining myocardial disease, and is for non-invasive generation of an ECV map using a neural network.
The present invention was derived from a study conducted as part of the Small and Medium Venture Business Startup Commercialization Support Project [B0080910000624, development of an automated model for diagnosing heart disease through comprehensive cardiac magnetic resonance imaging (MRI) analysis].

Among various molecular imaging techniques, magnetic resonance imaging (MRI) is based on the interaction between molecules and hydrogen atoms (protons) surrounding the tissue (lattice), and provides very good anatomical results. Because it can provide images, it is considered one of the most powerful and non-invasive diagnostic tools.

In recent magnetic resonance imaging, imaging techniques are being developed to quantitatively measure biophysical quantities as well as conventional anatomical tomography images. Biophysical quantities that can be measured by MRI include self-recovery time, blood flow rate, diffusion, and perfusion of hydrogen atoms following high-frequency pulses in human tissue, and can be measured safely and accurately non-invasively.

Among them, a contrast between tissues can be created through the characteristics of each component of self-healing, and this difference in recovery time is actively used as a biophysical quantity that can quantitatively evaluate tissue disease at the molecular level.

In addition, it is also possible to create images specialized for each disease by diversifying by reflecting additional characteristics to the difference in physical quantity at this time.

In particular, recently, an ECV (Extracellular Volume Index) index was calculated using the value before and after the injection of the contrast agent at T1 time, which defines the recovery time in the longitudinal direction after reversal magnetization, and the volume of red blood cells (hematocrit) in the blood, to improve fibrosis of the heart. It is used to predict various diseases.

However, in order to accurately determine the red blood cell volume used at this time, an additional blood test is required in addition to MRI imaging (Prior Patent Document Korea Patent Publication No. 10-2019-0117411 (2019.10.16)). The downside is that inspection is required.

In addition, the timing of the blood test may also affect the ECV index. In the case of outpatients, there may be a time difference between MRI imaging and blood collection, and therefore, this problem needs to be supplemented.

The purpose of the ECV map generation method according to the present invention is to propose a non-invasive index calculation method.

In addition, an object of the present invention is to propose a method for predicting features in blood using values generated by MRI imaging using a neural network.

In addition, an object of the present invention is to propose a method of more accurately predicting intra-blood characteristics through MRI imaging and patient characteristic information and generating a more accurate ECV map through this.

The method for generating an ECV map based on hematocrit prediction through a neural network according to an embodiment of the present invention to solve the above technical problem is a two-dimensional recovery time according to the reference recovery rate of protons in heart tissue reversed with RF (Radio Frequency) pulses. obtaining a T1 map generated by mapping in units of pixels in space; extracting a blood T1 value from the T1 map; outputting an estimated hematocrit according to the extracted blood T1 value using a neural network trained to predict a blood hematocrit from the T1 value; and generating an extracellular volume index (ECV) map by using the output estimated red blood cell volume as a control variable of the T1 map.

The method may further include extracting a myocardial T1 value of the heart from the T1 map, wherein the neural network learns a correlation between the blood T1 value and the myocardial T1 value and the red blood cell volume.

Preferably, the blood and myocardial T1 values include a first T1 value extracted from a native T1 map before contrast medium injection and a second T1 value extracted from a post T1 map after injection.

The blood T1 value is obtained by fitting the recovery time of the heart tissue to a normalized curve in units of pixels using images obtained by taking multiple images of the recovery process of inverted protons in the heart tissue at arbitrary time intervals, and from the curve It is preferably extracted from the T1 map generated by calculating the recovery time according to the reference recovery rate.

Preferably, the myocardial T1 value is a T1 value corrected by using a parameter defining the curve of the T1 value of the myocardial region in the T1 map.

The method may further include receiving quantified patient information including the patient's gender and age, wherein the neural network learns the T1 value and the correlation between the patient information and the red blood cell volume.

In order to solve the above technical problem, the apparatus for generating an ECV map based on hematocrit prediction through a neural network according to an embodiment of the present invention calculates the recovery time according to the reference recovery rate of protons in heart tissue inverted by RF pulses in units of pixels in a two-dimensional space. T1 map generation unit for generating a T1 map by mapping to; a numerical prediction unit extracting a blood T1 value from the T1 map and outputting an estimated red blood cell volume according to the extracted blood T1 value by using a neural network trained to predict a blood red blood cell volume from the T1 value; and an ECV map generating unit generating an ECV map using the output estimated hematocrit as a control variable of the T1 map.

A diagnostic unit for diagnosing the degree of myocardial fiber using the generated ECV map is further included.

The method may further include extracting a myocardial T1 value of the heart from the T1 map, wherein the neural network learns a correlation between the blood T1 value and the myocardial T1 value and the red blood cell volume.

The blood and myocardial T1 values include a first T1 value extracted from a native T1 map before contrast medium injection and a second T1 value extracted from a post T1 map after injection.

According to the present invention, the patient's discomfort can be minimized by calculating the ECV index on a non-invasive basis.

In addition, the present invention can simplify the diagnosis process and increase the reliability of diagnosis by accurately predicting blood characteristics only with MRI scan results and patient characteristic information using a neural network.

1 is a diagram schematically illustrating an ECV map generating system according to an embodiment of the present invention.
2 is a flow diagram illustrating a method for generating an ECV map according to an embodiment of the present invention.
3 is a diagram illustrating a T1 map used in the method for generating an ECV map according to an embodiment of the present invention.
4 and 5 are diagrams illustrating the structure of a neural network for predicting red blood cell volume according to an embodiment of the present invention.
6 is a diagram illustrating the importance of input parameters of a neural network for predicting red blood cell volume according to an embodiment of the present invention.
7 is a diagram illustrating the structure of a neural network for predicting red blood cell volume according to an embodiment of the present invention.
8 is a diagram illustrating generation of a T1 map for predicting red blood cell volume according to an embodiment of the present invention.
9 is a diagram illustrating the configuration of an ECV generating device according to an embodiment of the present invention.

The following merely illustrates the principles of the invention. Therefore, those skilled in the art can invent various devices that embody the principles of the invention and fall within the concept and scope of the invention, even though not explicitly described or shown herein. In addition, all conditional terms and embodiments listed in this specification are, in principle, clearly intended only for the purpose of understanding the concept of the invention, and it should be understood that it is not limited to the specifically listed embodiments and conditions. .

The above objects, features and advantages will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art to which the invention belongs will be able to easily implement the technical idea of the invention. .

In addition, in describing the invention, if it is determined that a detailed description of a known technology related to the invention may unnecessarily obscure the subject matter of the invention, the detailed description will be omitted. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

First, MRI can calculate the recovery time according to the component by distinguishing the components in the recovery process that are recovered in the form of an exponential curve after inverting the atoms longitudinally magnetized by the magnetic field using the inversion recovery RF pulse. A signal generated from the subject is scanned in k-space to obtain an MRI signal, and an MRI image is acquired by converting the obtained MRI signal.

Specifically, the characteristics of each component of self-recovery can be emphasized by adjusting, for example, TR (Time to repetition) and TE (Time to echo) as variables for applied RF (Radio Frequency) pulses.

At this time, TR refers to the repetition time of the RF pulse. TR refers to a time interval for generating an RF pulse used to obtain a resonance signal, and mainly determines the amount of longitudinal relaxation (spin-lattice).

TE is the signal generation time, which means the time from outputting the RF pulse to acquiring the echo signal. TE determines the degree of dispersion (out-of-phase) of the recovered spin (Spin-Spin) onto the transverse plane.

At this time, T1 is defined as the T1 relaxation time from when the RF pulse is injected and then reversed until the average magnetization of 63% of the initial state is formed in the direction of the longitudinal axis.

T2 is defined as the time taken for the average magnetization of the transverse plane to decrease to 37% of the original value by dephasing.

That is, a T1 weighted image or a T2 weighted image may be generated by measuring the above time by varying the TR and TE of the RF pulse.

Specifically, the present invention proposes a method of predicting the hematocrit used as a control variable using a neural network while using the T1 image to generate an ECV map useful for quantitative analysis of diffuse myocardial fibrosis.

Hereinafter, it will be described in more detail with reference to FIG. 1 .

Referring to FIG. 1 , an ECV map generating device 100 performing a method of generating an ECV map based on prediction of a red blood cell count through a neural network according to the present embodiment may be configured to interwork with an MRI device 1000 .

As described above, the MRI apparatus 1000 acquires non-invasive images by injecting magnetic fields and RF pulses into the human body, extracting the characteristics of the magnetization and recovery process of hydrogen atoms, and generating MRI images of signals received during the recovery process. and make diagnosis possible.

In this case, the ECV map generator 100 may include a T1 map generator, a numerical prediction unit, and an ECV map generator.

That is, the T1 map generator may calculate a T1 recovery time according to a recovery rate in a spin-lattice recovery process in which an exponential curve is recovered after inverting an atom demagnetized by a magnetic field using an inversion recovery RF pulse. That is, a T1 map mapped in units of pixels may be generated by projecting the T1 value generated from the subject to the k-space as described above.

Next, to calculate the ECV map, a native T1 map of the heart before contrast medium injection and a post T1 map after contrast medium injection are required.

Therefore, the T1 map generation unit additionally creates a post T1 map by measuring the T1 time after the injection of the contrast medium.

Furthermore, in order to generate an ECV map, a blood hematocrit value is required. The hematocrit is a percentage occupied by red blood cells in total blood, and a conventional method is to collect and measure a blood sample.

However, the existing method is time consuming, invasive, and increases patient discomfort, and errors may occur on the ECV map depending on the difference between the measurement time and the MRI scan time, which may lead to misdiagnosis.

In other words, it is most accurate to measure blood sampling immediately before MRI imaging for ECV calculation, and if not, it is recommended to measure blood within 24 hours.

Therefore, in the present embodiment, the numerical prediction unit attempts to predict the hematocrit for generating the ECV map through the neural network using only the MRI results obtained by omitting the blood sampling step.

The ECV map generation unit generates an ECV map using the hematocrit predicted by the numerical prediction unit.

Hereinafter, a method for generating an ECV map based on prediction of red blood cell count through a neural network according to the present embodiment will be described in detail with reference to FIG. 2 .

First, a T1 map generated by mapping a recovery time according to a reference recovery rate of protons in heart tissue inverted by an RF pulse in units of pixels in a two-dimensional space is obtained (S100).

Specifically, in this embodiment, a T1 map of the heart before contrast medium injection and a T1 map after injection may be acquired.

Referring to FIG. 3, the T1 times of the tissues in the heart before (A. native T1) and after (B. post T1) contrast medium are output as different values. By doing so, an ECV map is created.

Specifically, the ECV indicator may be calculated by the following equation.

[Equation 1]

(Hematocrit = red blood cell volume, T1 _myo-post = myocardial post T1 time, T1 _myo-pre = myocardial native T1 time, T1 _blood-post = blood post T1 time, T1 _blood-pre = blood native T1 time)

Accordingly, the blood T1 value and the myocardial T1 value in the T1 map before and after the injection of the contrast medium are extracted (S200).

Subsequently, the predicted red blood cell volume according to the extracted blood T1 value is output using the neural network trained to predict the blood red blood cell volume from the extracted T1 value (S300).

Specifically, the neural network according to the present embodiment will be described in more detail with reference to the following drawings.

First, referring to FIG. 4 , the neural network 200 according to the present embodiment can learn a correlation between the input variable 210 composed of the blood T1 value and the quantified patient information value and the red blood cell volume 220 .

Each node formed between the input and output layers in the neural network 200 is interconnected, and by learning and updating the strength of the connection, the neural network can extract feature values that affect hematocrit.

Specifically, the connection strength can be defined as a weight (W) and a bias (B), and the output value can be passed to the next node through an operation such as y=Wx+b for the input value x of a specific node, and the final output neural network The output of is to update the weight or bias by back-propagating the error according to the difference from the actual value (Ground_truth).

After performing the above process, the neural network 200 learns to predict the hematocrit value from the patient information and the T1 value.

Further, in detail, the input value 210 for predicting red blood cells may include a value according to the anatomical division of the heart and personal information such as the patient's sex or age.

Referring to FIG. 5 , the neural network 200 according to the present embodiment uses age, gender, native T1, post T1, native blood T1, post blood T1, T2, and left ventricular ejection fraction (EF) as input data 210. available.

In addition, T1 can be divided into 16 regions according to standardized criteria (eg, the American Heart Association (AHA) model) for classifying detailed tissues of the heart, and the T1 values of the divided regions can be used as input.

In addition, in the case of blood or myocardium T1, the heart can be divided into three height units and extracted for each T1 map (base, mid, and apex).

Among the above input values, values that affect hematocrit can be emphasized according to learning of the neural network as described above, and through this, the neural network can predict the hematocrit more accurately.

Referring to FIG. 6 , as a result of training of the neural network, blood T1 in the upper image (NT1_B3), middle image (NT1_B2), and lower portion (NT1_B1), gender, and age were determined to be the most important variables affecting the hematocrit. Therefore, a node of a neural network that operates with these values as inputs passes a more emphasized output to the next node.

For example, in the case of native blood T1, the weight of the node in the neural network may be determined to be high in the order of a blood T1 value on an upper T1 map, a blood T1 value on a lower T1 map, and a blood T1 value on a middle T1 map.

In addition, although the aforementioned neural network 200' is exemplified as using tabulated input values, the neural network itself may additionally be trained to directly output the hematocrit 220 using the plurality of T1 maps 210' as input.

That is, by directly recognizing values of pixels in the T1 map, T1 values of blood T1 and detailed tissue of the heart are extracted on the map, and the hematocrit value may be extracted through the T1 values.

Referring to FIG. 7 , in this embodiment, the neural network 200' may be implemented in the form of a Convolution Neural Network (CNN). That is, a structure in which a plurality of convolution layers performing a convolution operation on an image input as an input layer may be stacked, and the convolution layer is learned to emphasize feature values in an image through a weight in a filter.

The feature map generated through the convolution layer may be flattened through the fully connected layer and generated as a final output value.

In this embodiment, a native T1 map and a post T1 map can be input as input images, and gender and age can be additionally input as patient information. The error is calculated by comparing the hematocrit predicted as the final output with the correct value, and the weights in the convolution layer are updated by back-propagating the error.

Next, an ECV map is generated by using the estimated red blood cell volume output through the neural network as a control variable of the T1 map (S400).

The ECV map may be generated by applying the predicted hematocrit value to the difference between the native T1 map and the post T1 map as an adjustment variable using the above equation and mapping the calculated value in pixel units.

In addition, in this embodiment, some values of the T1 map may be mathematically predicted or the predicted values may be corrected and used. In general, the T1 map may be generated by fitting the recovery time of hydrogen atoms inverted according to the continuously provided RF pulse after the inversion pulse to a normalized curve.

That is, a T1 map is created according to the curve generated by fitting the recovery time according to the characteristics of hydrogen atoms constituting heart tissue in an exponential function form, and the T1 recovery time is extracted from this curve to be used for predicting red blood cell volume and generating an ECV map. can

Specifically, referring to FIG. 8, using an inversion recovery (IR) pulse, hydrogen atoms are rotated 180 degrees from +M0, which is in an equilibrium state, to be reversed to -M0, and then an image can be obtained through a signal received during the recovery process. there is. At this time, TI (inversion time) is adjusted according to the time interval of image acquisition. Depending on the TI time, the degree of recovery of hydrogen atoms in each tissue to +M0 according to the surrounding characteristics varies, and according to this difference, the contrast of MRI images It will be different.

That is, a geometric curve can be generated by aligning images (T1-weighted images) by mapping a plurality of signals obtained according to the TI time into K-space and observing a signal change in units of pixels in the image.

Additionally, in this embodiment, in order to minimize the effect of the heartbeat, it is also possible to determine the TI as one heartbeat cycle using an electrocardiogram (ECG) signal of the heart.

The curve for ideal curve fitting can be defined as a three-parameter (A, B, T1 ^* ) model. That is, the curve has the time value t according to T1 ^* as an input of the exponential function, and the state S(t) of the hydrogen atom according to three parameters and time t is defined by Equation 2 below.

[Equation 2]

S(t)= A-Bexp(-t/T ₁ ^* )

However, since the curve at this time is calculated mathematically without reflecting the state of actual atoms according to continuous addition of pulses or accumulation of magnetized states, the T1* value used in this embodiment can be corrected and used.

That is, the mathematically calculated T1 ^* can be corrected using the parameters A and B, and the corrected T1 can be calculated to be close to the actual one through the following equation.

[Equation 3]

T ₁ =T ₁ ^* (B/A-1)

However, in this embodiment, in predicting the hematocrit, T1 in the T1 map generated by Equation 2 above. Values or calibrated T1* values can optionally be used. Specifically, in the case of blood T1, the mathematically calculated value T1 ^* on the curve before correction may be used.

That is, referring to FIG. 5, the T1 value for myocardium among the T1 values used for prediction uses the value corrected according to Equation 3, but the T1 value for blood is the T1 ^* value on the curve according to Equation 2 before correction. make available.

As described above, the ECV map generation method according to the present embodiment extracts the intracardiac T1 value and uses it to predict the hematocrit.

Hereinafter, referring to FIG. 9 , an ECV generating device 100 that performs a method of generating an ECV map based on prediction of a red blood cell number according to an embodiment of the present invention will be described.

In this embodiment, the ECV generator 100 may include a T1 map generator 110, a numerical prediction unit 120, an ECV map generator 130, and a diagnosis unit 140.

As described above, the T1 map generating unit 110 generates an image by mapping the time T1 taken for the reference recovery rate according to the RF pulse of the hydrogen atoms in the tissue of the subject, preferably the heart, in units of pixels.

In this embodiment, in order to generate a T1 map, a plurality of acquired images according to TIs may be mapped into curves to calculate T1 ^* , and corrected T1 may be extracted by correcting the calculated T1 ^* .

Next, the numerical predictor 120 predicts the hematocrit using the T1 value of the T1 map. Specifically, the T1 values of the native T1 map before the injection of the contrast agent and the T1 value of the post T1 map after the injection of the contrast agent are used, but the segmented T1 values may be used by dividing the heart tissue.

For example, the T1 value of blood within the heart and the T1 value for each unit area according to standardized criteria for classifying the heart may be used as variables for numerical prediction.

Furthermore, the numerical predictor 120 may receive input of body information such as gender or age as basic information of the patient in addition to the T1 map, and may utilize the information for prediction.

In detail, numerical prediction can be performed using the neural network 200, and the neural network 200 can learn the importance of parameters that affect the determination of the hematocrit among input variables and reflect it as a weight.

In addition, in this embodiment, the neural network 200' can receive a directly digitized value, but it is also learned to receive the T1 map itself as an input, extract feature values through CNN-based convolution operation, and output the volume of red blood cells. possible.

When the hematocrit is predicted through the above process, the ECV map generator 130 generates an ECV map using the T1 value and the hematocrit.

Additionally, the diagnosis unit 140 may perform analysis of myocardial fibrosis using features in the image of the ECV map.

Specifically, it is possible to extract a region in which the myocardium is expected to be fibrotic by using the feature of the pixel of the ECV map, quantify the degree of progression thereof through the feature value of the pixel, and provide it to the user.

According to the present invention, the patient's discomfort can be minimized by calculating the ECV index on a non-invasive basis.

In addition, the present invention can simplify the diagnosis process and increase the reliability of diagnosis by accurately predicting blood characteristics only with MRI scan results and patient characteristic information using a neural network.

Furthermore, various embodiments described herein may be implemented in a recording medium readable by a computer or a device similar thereto using, for example, software, hardware, or a combination thereof.

According to hardware implementation, the embodiments described herein include application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), It may be implemented using at least one of processors, controllers, micro-controllers, microprocessors, and electrical units for performing other functions. The described embodiments may be implemented in the control module itself.

According to software implementation, embodiments such as procedures and functions described in this specification may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein. The software code may be implemented as a software application written in any suitable programming language. The software code may be stored in a memory module and executed by a control module.

The above description is merely an example of the technical idea of the present invention, and those skilled in the art can make various modifications, changes, and substitutions without departing from the essential characteristics of the present invention. will be.

Therefore, the embodiments disclosed in the present invention and the accompanying drawings are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings. . The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the equivalent range should be construed as being included in the scope of the present invention.

Claims (13)

A method for generating an ECV map performed by a computing device,
obtaining a T1 map generated by mapping a recovery time according to a reference recovery rate of protons in cardiac tissue inverted by a radio frequency (RF) pulse in units of pixels in a two-dimensional space;
extracting myocardial T1 values and blood T1 values of the heart from the T1 map;
Using a neural network learned to predict the blood hematocrit from the T1 value, the myocardial T1 value input to one node in the neural network and the blood T1 value according to the weight obtained by learning the connection strength between the nodes outputting an estimated red blood cell volume as a weighted calculation result of the myocardial T1 value and the blood T1 value;
Generating an extracellular volume index (ECV) map through calculation of the output estimated red blood cell volume and the T1 value on the T1 map;
The neural network,
The blood T1 value calculated by curve fitting the recovery rate of the protons over time in the blood region with a curve and the T1 value calculated by the curve fitting of the protons in the myocardial region are at least one value used for the curve fitting. learning the correlation between the myocardial T1 value corrected with the parameter and the hematocrit as a weight;
The myocardial T1 value is a value for each region in which a plurality of regions of the heart are divided. According to claim 1,
Further comprising extracting a myocardial T1 value of the heart in the T1 map,
Wherein the neural network learns a correlation between the blood T1 value and the myocardial T1 value and the red blood cell volume. According to claim 2,
Wherein the blood and myocardial T1 values include a first T1 value extracted from a native T1 map before contrast agent injection and a second T1 value extracted from a post T1 map after administration of the contrast agent. A method for generating ECV maps based on numerical prediction. According to claim 2,
The blood T1 value is obtained by fitting the recovery time of the heart tissue to a normalized curve in units of pixels using images obtained by taking multiple images of the recovery process of inverted protons in the heart tissue at time intervals, and the reference recovery rate from the curve. A method for generating an ECV map based on prediction of red blood cell count through a neural network, characterized in that it is extracted from the T1 map generated by calculating the recovery time according to. According to claim 4,
The myocardial T1 value is a T1 value corrected by using a parameter defining the curve of the T1 value of the myocardial region in the T1 map. According to claim 2,
Further comprising the step of receiving quantified patient information including the patient's gender and age,
Wherein the neural network learns a correlation between the T1 value, the patient information, and the red blood cell volume. According to claim 2,
The blood T1 value is a value according to the curve calculated according to curve fitting, and the myocardial T1 value is a value obtained by correcting the value according to the curve calculated according to the curve fitting. How to create. T1 map generation unit that generates a T1 map by mapping the recovery time according to the reference recovery rate of protons in the heart tissue reversed by RF pulses in pixel units in a 2-dimensional space
Using a neural network trained to extract the blood T1 value in the T1 map and predict the blood hematocrit from the T1 value, for the myocardial T1 value and the blood T1 value input to one node in the neural network, the node a numerical prediction unit outputting an estimated red blood cell volume as a weight calculation result of the myocardial T1 value and the blood T1 value according to the weight obtained by learning the connection strength between the cells;
An ECV map generating unit generating an ECV map by calculating the output estimated hematocrit and the T1 value on the T1 map;
The neural network,
The blood T1 value calculated by curve fitting the recovery rate of the protons over time in the blood region with a curve and the T1 value calculated by the curve fitting of the protons in the myocardial region are at least one value used for the curve fitting. Learning the correlation between the myocardial T1 value corrected by the parameter and the red blood cell volume as a weight,
The myocardial T1 value is an ECV map generating device based on prediction of red blood cell count through a neural network, characterized in that the heart muscle T1 value is a value for each region divided into a plurality of regions of the heart. According to claim 8,
The apparatus for generating an ECV map based on prediction of red blood cell count, characterized in that it further comprises a diagnostic unit for diagnosing the degree of myocardial fiber using the generated ECV map. According to claim 8,
Further comprising extracting a myocardial T1 value of the heart in the T1 map,
Wherein the neural network learns a correlation between the blood T1 value and the myocardial T1 value and the red blood cell volume. According to claim 10,
Wherein the blood and myocardial T1 values include a first T1 value extracted from a native T1 map before contrast agent injection and a second T1 value extracted from a post T1 map after administration of the contrast agent. Numerical prediction-based ECV map generation device. According to claim 10,
The blood T1 value is a value according to the curve calculated according to curve fitting, and the myocardial T1 value is a value obtained by correcting the value according to the curve calculated according to the curve fitting. generating device. obtaining a T1 map generated by mapping a recovery time according to a reference recovery rate of protons in cardiac tissue inverted by a radio frequency (RF) pulse in units of pixels in a two-dimensional space;
extracting myocardial T1 values and blood T1 values of the heart from the T1 map;
Using a neural network learned to predict the blood hematocrit from the T1 value, the myocardial T1 value input to one node in the neural network and the blood T1 value according to the weight obtained by learning the connection strength between the nodes outputting an estimated red blood cell volume as a weighted calculation result of the myocardial T1 value and the blood T1 value;
Generating an extracellular volume index (ECV) map through calculation of the output estimated red blood cell volume and the T1 value on the T1 map;
The neural network,
The blood T1 value calculated by curve fitting the recovery rate of the protons over time in the blood region with a curve and the T1 value calculated by the curve fitting of the protons in the myocardial region are at least one value used for the curve fitting. Learning the correlation between the myocardial T1 value corrected by the parameter and the red blood cell volume as a weight,
The myocardial T1 value is a value for each region in which the heart region is divided into a plurality of regions.

Images