CN117197521A - Simulation image preprocessing method, local specific absorption rate estimation method and device - Google Patents
Simulation image preprocessing method, local specific absorption rate estimation method and device Download PDFInfo
- Publication number
- CN117197521A CN117197521A CN202310727563.XA CN202310727563A CN117197521A CN 117197521 A CN117197521 A CN 117197521A CN 202310727563 A CN202310727563 A CN 202310727563A CN 117197521 A CN117197521 A CN 117197521A
- Authority
- CN
- China
- Prior art keywords
- electric field
- human body
- image
- absorption rate
- specific absorption
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004088 simulation Methods 0.000 title claims abstract description 83
- 238000010521 absorption reaction Methods 0.000 title claims abstract description 81
- 238000000034 method Methods 0.000 title claims abstract description 78
- 238000007781 pre-processing Methods 0.000 title claims abstract description 17
- 230000005684 electric field Effects 0.000 claims abstract description 156
- 238000012549 training Methods 0.000 claims abstract description 110
- 238000013136 deep learning model Methods 0.000 claims abstract description 78
- 230000005540 biological transmission Effects 0.000 claims abstract description 12
- 238000004590 computer program Methods 0.000 claims description 17
- 238000013507 mapping Methods 0.000 claims description 13
- 238000003860 storage Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 description 19
- 231100000817 safety factor Toxicity 0.000 description 15
- 238000009826 distribution Methods 0.000 description 14
- 230000000694 effects Effects 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000002595 magnetic resonance imaging Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000033228 biological regulation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008021 deposition Effects 0.000 description 2
- 230000005672 electromagnetic field Effects 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 101100112995 Schizosaccharomyces pombe (strain 972 / ATCC 24843) sim4 gene Proteins 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 125000004435 hydrogen atom Chemical group [H]* 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009885 systemic effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application relates to a simulation image preprocessing method, a local specific absorption rate estimation method and a device, which are characterized in that a simulation image is obtained by conducting a human body model to an electromagnetic transmission coil model, wherein the simulation image comprises an electric field image and a magnetic field image, the radio frequency transmission coil model comprises N channels, and N is more than or equal to 2; identifying the internal field of the human body in the simulation image, and setting the corresponding field intensity of the critical positions of the internal part of the human body and the air to zero; dividing the simulation image according to N channels to obtain a simulation image corresponding to each channel; extracting the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel to obtain a training set; the training set is used for training a deep learning model, and the deep learning model is used for predicting the radio frequency electric field. The method solves the problem that the accuracy of the local specific absorption rate predicted value of the object to be detected is low, and improves the accuracy of the local specific absorption rate predicted value of the object to be detected.
Description
Technical Field
The present application relates to the field of magnetic resonance imaging technology, and in particular, to a simulated image preprocessing method, a local specific absorption rate estimating method, an electronic device, and a storage medium.
Background
A magnetic resonance imaging (MRI, magnetic Resonance Imaging) system is a technique for imaging using magnetic resonance phenomena. Ultra-high field magnetic resonance may provide a higher signal-to-noise ratio relative to low field magnetic resonance systems. However, as the field intensity of the main magnetic field increases, the wavelength of the electromagnetic wave decreases, and a standing wave effect is generated after the electromagnetic wave is equivalent to the size of human tissue, so that the acquired image signals are uneven, and the deposition of radio frequency energy in the human tissue also increases correspondingly. The specific absorption rate (SAR, specific Absorption Rate), which is an index for evaluating the deposition of rf energy in body tissue, is the rf power absorbed per unit mass of tissue in units of kilowatts (W/kg). The radio frequency energy specific absorption rate can be divided into a systemic specific absorption rate (white-body SAR) and a Local specific absorption rate (Local-SAR), wherein the Local specific absorption rate refers to a specific absorption rate distribution of different locations of the human body. Typically, the specific absorption rate of the whole body can be estimated by monitoring the power recorded by a directional coupler in the magnetic resonance imaging system, while the local specific absorption rate is difficult to directly obtain by adopting an actual measurement method, and simulation software (such as CST, sim4Life, etc.) is typically used for simulation calculation. However, simulation has a few problems. Firstly, the human body model required during simulation is difficult to construct; secondly, the simulation requires a long time, and the requirement of acquiring the local specific absorption rate in real time cannot be met. Therefore, scholars have proposed using deep learning methods to estimate the local specific absorption rate of ultra-high fields.
In the related technology, a simulation image of a magnetic resonance imaging system is obtained through simulation, the simulation image is used as a training set, a deep learning model is trained, and the trained deep learning model is adopted to predict the local specific absorption rate of an object to be detected. However, with the deep learning model of the related art, the predicted value and the true value (simulation value) of the local specific absorption rate differ greatly.
Aiming at the problem of low accuracy of the local specific absorption rate estimated value in the related art, no effective solution is proposed at present.
Disclosure of Invention
In this embodiment, a simulated image preprocessing method, a local specific absorption rate estimating method, an electronic device and a storage medium are provided to solve the problem of low accuracy of a local specific absorption rate estimated value in the related art.
In a first aspect, in this embodiment, there is provided a simulated image preprocessing method, including:
the method comprises the steps of obtaining a simulation image, wherein the simulation image is obtained by conducting a human body model to an electromagnetic transmission coil model, the simulation image comprises an electric field image and a magnetic field image, and the radio frequency transmission coil model comprises N channels, wherein N is more than or equal to 2;
identifying the internal field of the human body in the simulation image, and setting the corresponding field intensity of the critical positions of the internal part of the human body and the air to zero;
Dividing the simulation image according to the N channels to obtain a simulation image corresponding to each channel;
extracting the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel to obtain a training set;
the training set is used for training a deep learning model, and the deep learning model is used for predicting a radio frequency electric field.
In some of these embodiments, said identifying the human body internal field in said simulated image comprises:
in the electric field image, determining the area of the first human body internal field according to the human body tissue medium parameter and the air medium parameter;
covering a first mask in the electric field image, wherein the area where the first human body internal field is located is not covered by the first mask; the method comprises the steps of,
in the magnetic field image, determining the region of the second human body internal place according to the human body tissue medium parameters and the air medium parameters;
and covering a second mask in the magnetic field image, wherein the second human body internal place is not covered by the second mask in a region.
In some of these embodiments, the zeroing the respective field strengths of the critical locations of the human body interior and the air comprises:
In the electric field image, determining a first critical position of the interior of the human body and air according to a human tissue medium parameter and an air medium parameter, and setting the electric field intensity of the first critical position to zero; and
in the magnetic field image, a second critical position of the interior of the human body and air is determined according to the human body tissue medium parameter and the air medium parameter, and the magnetic field intensity of the second critical position is set to zero.
In some of these embodiments, the radio frequency transmit coil model comprises a multi-channel array transmit coil model, a full-volume transmit coil model, or a partial-volume transmit coil model; and/or, the magnetic field intensity of the radio frequency transmitting coil model is not lower than 3T.
In a second aspect, in this embodiment, there is provided a local specific absorption rate estimating method, including:
acquiring real magnetic field data of an object to be detected after scanning;
inputting the real magnetic field data into a trained deep learning model, predicting to obtain a radio frequency electric field of the object to be detected, and obtaining a first local specific absorption rate of the object to be detected according to the radio frequency electric field; wherein training the deep learning model comprises:
Acquiring a training set, wherein the training set is obtained by processing the simulation image preprocessing method in the first aspect;
and training the deep learning model according to the training set until the deep learning model meets convergence conditions.
In some embodiments, the obtaining the first local specific absorption rate of the object to be measured according to the radio frequency electric field includes:
acquiring the conductivity and density of the object to be measured;
and determining a first local specific absorption rate of the object to be measured according to the conductivity of the object to be measured, the density and the radio frequency electric field, wherein the first local specific absorption rate is directly proportional to the conductivity, the first local specific absorption rate is inversely proportional to the density, and the first local specific absorption rate is directly proportional to the square of the radio frequency electric field.
In some embodiments, the training set includes simulated magnetic field data and simulated electric field data, wherein the simulated magnetic field data includes simulated magnetic field amplitude and simulated magnetic field phase, the simulated electric field data includes simulated electric field amplitude and simulated electric field phase, and training a deep learning model according to the training set includes:
Inputting the simulated magnetic field amplitude and the simulated electric field amplitude of the mth channel into the deep learning model to obtain an amplitude mapping relation between the magnetic field and the electric field in the mth channel;
and inputting the simulated magnetic field phase and the simulated electric field phase of the mth channel into the deep learning model to obtain a phase mapping relation between the magnetic field phase and the electric field phase in the mth channel.
In some of these embodiments, the method further comprises:
obtaining a second local specific absorption rate, wherein the second local specific absorption rate is obtained by electromagnetic simulation of a human body model guided-incidence frequency transmitting coil model corresponding to the object to be detected;
obtaining a safety factor according to the ratio of the first local specific absorption rate maximum value to the second local specific absorption rate maximum value in each channel;
evaluating the prediction accuracy of the deep learning model according to the safety factor;
and adjusting the weight of the deep learning model according to the prediction accuracy of the deep learning model.
In a third aspect, in this embodiment, there is provided an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method of the first aspect or the second aspect when executing the computer program.
In a fourth aspect, in this embodiment, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first or second aspect described above.
According to the simulation image preprocessing method, the local specific absorption rate estimation method, the electronic device and the storage medium, the human body internal field is identified in the simulation image, so that the deep learning model is more focused on the human body internal field in the training process, and the human body internal field is fully trained; the corresponding field intensity of the critical position of the inside of the human body and the air is set to zero, so that the deep learning model is easier to converge; the simulation image is divided according to the N channels, the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel are extracted, the relation between the internal field of the human body and the space position is deepened, and the problem of training distortion is solved. According to the application, the simulation image is preprocessed to construct a complete and accurate training set, so that the relation between the training set and the deep learning model is more intimate, and the trained deep learning model is more accurate, thereby solving the problem of low accuracy of the local specific absorption rate predicted value of the object to be tested and improving the accuracy of the local specific absorption rate predicted value of the object to be tested.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the other features, objects, and advantages of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:
FIG. 1 is a block diagram of the hardware architecture of a terminal in one embodiment;
FIG. 2 is a flow chart of a method for preprocessing a simulation image in one embodiment;
FIG. 3 is a schematic diagram of contrast of electric field images before and after identification in one embodiment;
FIG. 4 is a diagram showing training results obtained by training a deep learning model with simulated images before and after identification in one embodiment;
FIG. 5 is a schematic diagram of contrast of electric field images before and after zero in one embodiment;
figure 6 is a schematic diagram of learning effect evaluation before zero placement in one embodiment,
FIG. 7 is a schematic diagram of learning effect evaluation after zero placement in one embodiment;
FIG. 8 is a flow chart of a deep learning model training method in one embodiment;
FIG. 9 is a diagram showing training results of a conventional training method;
FIG. 10 is a diagram of training results of a training method according to one embodiment;
FIG. 11 is a second diagram of training results of a conventional training method;
FIG. 12 is a second diagram of training results of the training method according to one embodiment;
FIG. 13 is a flow chart of a method for estimating local specific absorption rate according to an embodiment;
fig. 14 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
The present application will be described and illustrated with reference to the accompanying drawings and examples for a clearer understanding of the objects, technical solutions and advantages of the present application.
Unless defined otherwise, technical or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," "these" and similar terms in this application are not intended to be limiting in number, but may be singular or plural. The terms "comprising," "including," "having," and any variations thereof, as used herein, are intended to encompass non-exclusive inclusion; for example, a process, method, and system, article, or apparatus that comprises a list of steps or modules (units) is not limited to the list of steps or modules (units), but may include other steps or modules (units) not listed or inherent to such process, method, article, or apparatus. The terms "connected," "coupled," and the like in this disclosure are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as used herein means two or more. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., "a and/or B" may mean: a exists alone, A and B exist together, and B exists alone. Typically, the character "/" indicates that the associated object is an "or" relationship. The terms "first," "second," "third," and the like, as referred to in this disclosure, merely distinguish similar objects and do not represent a particular ordering for objects.
The method provided by the application comprises a simulated image preprocessing method, a local specific absorption rate estimation method and a deep learning model training method, and the methods can be executed in a terminal, a computer or a similar computing device. Such as on a terminal, fig. 1 is a block diagram of the hardware architecture of the terminal according to an embodiment of the present application. As shown in fig. 1, the terminal may include one or more (only one is shown in fig. 1) processors 102 and a memory 104 for storing data, wherein the processors 102 may include, but are not limited to, a microprocessor MCU, a programmable logic device FPGA, or the like. The terminal may also include a transmission device 106 for communication functions and an input-output device 108. It will be appreciated by those skilled in the art that the structure shown in fig. 1 is merely illustrative and is not intended to limit the structure of the terminal. For example, the terminal may also include more or fewer components than shown in fig. 1, or have a different configuration than shown in fig. 1.
The memory 104 may be used to store a computer program, for example, a software program of application software and a module, such as a computer program corresponding to the method in the present embodiment, and the processor 102 executes the computer program stored in the memory 104, thereby performing various functional applications and data processing, that is, implementing the method described above. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located relative to the processor 102, which may be connected to the terminal via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission device 106 is used to receive or transmit data via a network. The network includes a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, simply referred to as NIC) that can connect to other network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is configured to communicate with the internet wirelessly.
It is important to detect the Local specific absorption rate (Local-SAR) of the human body, but it is difficult to obtain the Local specific absorption rate. A simulation software (such as Sim4 Life) includes multiple sets of typical mannequins, on which the electric and magnetic fields of each mannequin can be simulated separately. The electric field and the magnetic field of each human body model are divided into two parts, wherein the electric field and the magnetic field of one part of the human body model are used as training sets, and the electric field and the magnetic field of the other part of the human body model are used as test sets. The training set is directly adopted to train the deep learning model, and then the test set is used to test the trained deep learning model, so that the electric field and the magnetic field obtained by learning are found to have larger differences from the actual value (simulation value), and the predicted value and the actual value (simulation value) of the local specific absorption rate are caused to have larger differences. Through research and analysis, the following reasons are found:
(1) The human body model is arranged in the birdcage coil, the field amplitude around the coil excitation port is usually greatly higher than the field inside the human body, so that the deep learning model is excessively focused on the training of the near field attached to the coil excitation port, and the training of the field inside the human body is insufficient;
(2) At the critical position of human body and air, the electric field is often extremely large and is not consistent with the actual situation, so that the deep learning model is difficult to converge;
(3) The magnetic field and the electric field are closely related to the space position (the channel of the radio frequency transmitting coil), and the deep learning model does not distinguish the channel training magnetic field and the electric field, so that the training distortion is caused;
(4) During training, the phases of an untrained magnetic field and an electric field are caused by data volume limitation, so that training distortion is caused.
In view of the above analysis, in one embodiment, as shown in fig. 2, a simulated image preprocessing method is provided, and the method is applied to the terminal in fig. 1 for illustration, and includes the following steps:
step S201, a simulation image is obtained, wherein the simulation image is obtained by conducting a human body model to an electromagnetic simulation on an electromagnetic transmitting coil model, the simulation image comprises an electric field image and a magnetic field image, and the electromagnetic transmitting coil model comprises N channels, wherein N is more than or equal to 2.
The manikin can be obtained by simulation software such as CST, sim4 Life. The simulation software stores a typical human model library, and a proper digital human model can be selected from the typical human model library so as to cover a comprehensive body type range.
The radio frequency transmitting coil model can be obtained by constructing according to an actual radio frequency transmitting coil channel in simulation software. For example, according to the channel attribute of the actual radio frequency transmitting coil, corresponding channel parameters are configured in simulation software, and a radio frequency transmitting coil model is constructed according to the channel parameters. The actual radio frequency transmit coil channels comprise a multichannel array transmit coil, a full-volume transmit coil, or a partial-volume transmit coil, and the radio frequency transmit coil pattern comprises a multichannel array transmit coil pattern, a full-volume transmit coil pattern, or a partial-volume transmit coil pattern, respectively. The magnetic field intensity of the radio frequency transmitting coil model is not lower than 3T so as to simulate ultra-high field magnetic resonance.
The electric field image is an image carrying electric field distribution information, the electric field intensity can be expressed by the degree of color shade, if the color of a certain position in the image is darker, the electric field intensity of the position is larger, otherwise, if the color of the certain position in the image is lighter, the electric field intensity of the position is smaller. The magnetic field image is an image carrying magnetic field distribution information, the magnetic field intensity can be expressed by the degree of color shade, if the color of a certain position in the image is darker, the magnetic field intensity of the position is larger, otherwise, if the color of the certain position in the image is lighter, the magnetic field intensity of the position is smaller.
Step S202, identifying the internal field of the human body in the simulation image, and setting the corresponding field intensity of the critical positions of the internal and air of the human body to zero.
Identifying the human body internal field in the simulated image refers to identifying the human body internal electric field in the electric field image and identifying the human body internal magnetic field in the magnetic field image. Taking the example of identifying the internal electric field of a human body in the electric field image, fig. 3 provides an electric field image comparison group before and after identification, wherein 3-1 represents the electric field image before identification; 3-2 represents the identified electric field image, the large area coverage area represents air, and the middle exposed portion represents the human body internal field. The abscissa represents the lateral distance of the two-dimensional image, and the ordinate represents the longitudinal distance of the two-dimensional image. The light and dark color represents the intensity of field intensity, and the darker the color of a certain position in the image, the larger the field intensity of the position, and vice versa.
For convenience of description, a deep learning model trained according to a simulation image before identification is referred to as a first deep learning model, and a deep learning model trained according to a simulation image after identification is referred to as a second deep learning model. FIG. 4 provides a training result comparison group obtained by training the deep learning model with the simulation images before and after the identification, wherein 4-1 represents the simulation electric field image before the identification, 4-2 represents the learning electric field image output by the first deep learning model, 4-3 represents the simulation electric field image after the identification, and 4-4 represents the learning electric field image output by the second deep learning model. The abscissa represents the lateral distance of the two-dimensional image, and the ordinate represents the longitudinal distance of the two-dimensional image.
Therefore, after the human body internal field is marked in the simulation image, the deep learning model is more focused on training of the human body internal field, the bright spot position with higher amplitude is more accurate, and the learning effect is better.
Setting the respective field strengths of the critical positions of the inside of the human body and the air to zero refers to setting the field strengths of the electric field of the critical positions of the inside of the human body and the air to zero and setting the field strengths of the magnetic field of the critical positions of the inside of the human body and the air to zero. Taking the field intensity of the electric field at the critical position of the inside of the human body and the air as an example, fig. 5 provides a comparison group of electric field images before and after the zeroing, wherein 5-1 represents the electric field image before the zeroing, and 5-2 represents the electric field image after the zeroing. At the critical point of human tissue and air, the electric and magnetic fields tend to be unusually large, which is not practical.
The similarity of the learning image and the real (simulation) image can be described using a normalized correlation coefficient, and an expression of the normalized correlation coefficient will be given below:
wherein I is 1 、I 2 Image pixels, u, of the learning image and the real image, respectively 1 、u 2 The average values of pixels of the learning image and the real image are ncc, respectively, and the normalized correlation coefficients of the learning image and the real image are obtained.
And dividing the pixel value of the learning image with the pixel value of the real image point by point, and displaying the divided ratio in the form of a histogram to obtain an error histogram. In the present embodiment, the learning effect before and after zero setting is evaluated using the normalized correlation coefficient and the error histogram. FIG. 6 provides a schematic diagram of learning effect evaluation before zeroing, wherein 6-1 represents a simulated electric field image before zeroing; 6-2 represents a learning electric field image before zero setting; 6-3 represents an error electric field image before zero setting; 6-4 represents the error histogram before zero setting, the normalized correlation coefficient cor=73.5%, the abscissa represents the ratio of the learning electric field to the real electric field, the ordinate represents the frequency of occurrence of the ratio, and the more the frequency of the ratio of 1, the more accurate the prediction is explained. FIG. 7 provides a schematic diagram of learning effect evaluation after zeroing, wherein 7-1 represents a simulated electric field image after zeroing; 7-2 represents the learning electric field image after zero setting; 7-3 represents the error electric field image after zero setting; 7-4 represents the error histogram after zero setting, the normalized correlation coefficient cor=81.7%, the abscissa represents the ratio of the learning electric field to the real electric field, the ordinate represents the frequency of occurrence of the ratio, and the more the frequency of the ratio of 1, the more accurate the prediction is explained. From this, training the training set after zero setting treatment trains the deep learning model, and the normalized correlation coefficient of the obtained learning image and the real image is improved, and the center of the error histogram is closer to 1.
Step S203, dividing the simulation image according to the N channels to obtain a simulation image corresponding to each channel.
Dividing the simulation image according to the N channels to obtain a simulation image corresponding to each channel, namely determining an electric field image and a magnetic field image obtained by simulation of each channel so as to establish an association relationship between each channel and the simulation image.
Step S204, extracting the amplitude and the phase of the human body internal field in the simulation image corresponding to each channel to obtain a training set, wherein the training set is used for training a deep learning model, and the deep learning model is used for predicting a radio frequency electric field.
And extracting the amplitude and the phase of the electric field in the human body from the electric field image of each channel to obtain the amplitude of the electric field in the human body of each channel and the phase of the electric field in the human body of each channel. And extracting the amplitude and the phase of the magnetic field in the human body from the magnetic field image of each channel to obtain the amplitude of the magnetic field in the human body of each channel and the phase of the magnetic field in the human body of each channel. Thus, the training set includes the human body internal electric field amplitude of each channel, the human body internal electric field phase of each channel, the human body internal magnetic field amplitude of each channel, the human body internal magnetic field phase of each channel, the human body internal electric field amplitude of each channel and the human body internal electric field phase of each channel constitute simulated electric field data, and the human body internal magnetic field amplitude of each channel and the human body internal magnetic field phase of each channel constitute simulated magnetic field data.
The deep learning model comprises a mapping relation between an electric field and a magnetic field, when the deep learning model is used, simulation magnetic field data of an object to be detected are input into the trained deep learning model, a corresponding radio frequency electric field is output, and then the local specific absorption rate of the object to be detected is obtained according to the radio frequency electric field. The local specific absorption rate calculation formula is as follows:
wherein SAR represents the local specific absorption rate of the object to be measured, sigma represents the conductivity of the object to be measured, ρ represents the density of the object to be measured, and E represents the radio frequency electric field. It can be seen that the accuracy of the prediction of the radio frequency electric field is a key factor affecting the accuracy of the local specific absorption rate.
Step S201 to step S204 are described above, in which the human body internal field is identified in the simulation image, so that the deep learning model focuses more on the human body internal field in the training process, so as to fully train the human body internal field; the corresponding field intensity of the critical position of the inside of the human body and the air is set to zero, so that the deep learning model is easier to converge; the simulation image is divided according to the N channels, the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel are extracted, the relation between the internal field of the human body and the space position is deepened, and the problem of training distortion is solved. According to the embodiment, the simulation image is preprocessed to construct the complete and accurate training set, so that the relation between the training set and the deep learning model is more intimate, the trained deep learning model is more accurate, the problem that the accuracy of the local specific absorption rate predicted value of the object to be detected is low is solved, and the accuracy of the local specific absorption rate predicted value of the object to be detected is improved.
In one embodiment, identifying the human body internal field in the simulated image may be accomplished by:
in the electric field image, determining the area of the first human body internal field according to the human body tissue medium parameters and the air medium parameters; the first mask is covered in the electric field image, wherein the region of the first human body internal field is not covered by the first mask. In the magnetic field image, determining the region of the second human body internal place according to the human body tissue medium parameters and the air medium parameters; and covering the second mask in the magnetic field image, wherein the second human body internal space is not covered by the second mask in the region. In the simulated image, there are two media, human tissue and air, respectively. Wherein the field intensity of the position of the human tissue medium is greater than 0, and the field intensity of the position of the air medium is 0. Therefore, the human tissue medium and the air medium can be distinguished according to whether the field intensity is 0 (or more than 0), so that the location inside the human body is determined. The mask may be a binary image consisting of 0 and 1. When the mask is covered, the region of the internal place of the human body can be extracted as the region of interest, the mask of the region of interest which is manufactured in advance is multiplied by the magnetic field image or the electric field image to obtain the image of the region of interest, the image value in the region of interest is kept unchanged, and the image value outside the region of interest is 0.
In one embodiment, zeroing the respective field strengths of the critical locations of the human body interior and air may be accomplished by:
in the electric field image, according to the human tissue medium parameter and the air medium parameter, determining a first critical position of the interior of the human body and the air, and setting the electric field intensity of the first critical position to zero. In the magnetic field image, a second critical position of the interior of the human body and the air is determined according to the human tissue medium parameter and the air medium parameter, and the magnetic field intensity of the second critical position is set to be zero. In the simulated image, there are two media, human tissue and air, respectively. Wherein the field intensity of the position of the human tissue medium is greater than 0, and the field intensity of the position of the air medium is 0. Thus, the critical position of human tissue and air can be determined by distinguishing human tissue medium from air medium depending on whether the field strength is 0 (or greater than 0). Specifically, the field strength at the critical location can be zeroed out using a bwperim function in Matlab software.
In one embodiment, as shown in fig. 8, a training method of a deep learning model is provided, and the method is applied to the terminal in fig. 1 for illustration, and includes the following steps:
Step S301, a simulation image is obtained, wherein the simulation image is obtained by conducting electromagnetic simulation on a human body model to an electromagnetic transmitting coil model, the simulation image comprises an electric field image and a magnetic field image, and the electromagnetic transmitting coil model comprises N channels, wherein N is more than or equal to 2;
step S302, identifying the internal field of the human body in the simulation image, and setting the corresponding field intensity of the critical positions of the internal and air of the human body to zero;
step S303, dividing the simulation image according to N channels to obtain a simulation image corresponding to each channel;
step S304, extracting the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel to obtain a training set;
step S305, training the deep learning model according to the training set until the deep learning model meets the convergence condition.
In one embodiment, the training set includes simulated magnetic field data and simulated electric field data, wherein the simulated magnetic field data includes simulated magnetic field amplitude and simulated magnetic field phase, the simulated electric field data includes simulated electric field amplitude and simulated electric field phase, and training the deep learning model according to the training set can be achieved by:
inputting the simulated magnetic field amplitude and the simulated electric field amplitude of the mth channel into a deep learning model to obtain the amplitude mapping relation between the magnetic field and the electric field in the mth channel; and inputting the simulated magnetic field phase and the simulated electric field phase of the mth channel into a deep learning model to obtain a phase mapping relation between the magnetic field phase and the electric field phase in the mth channel.
In this embodiment, it is assumed that the radio frequency transmitting coils of the magnetic resonance system have 8 channels, that is, ch1 to Ch8, and the simulation magnetic field and the simulation electric field of the 8 channels can be obtained through electromagnetic simulation. In the training of the deep learning model, the embodiment distinguishes channels and distinguishes amplitude and phase in each channel for training. And distinguishing channel training, namely training the channel Ch1 based on the simulated magnetic field and the simulated electric field to obtain the mapping relation between the magnetic field and the electric field in the channel Ch1, and training the channel Ch2 based on the simulated magnetic field and the simulated electric field to obtain the mapping relation between the magnetic field and the electric field in the channel Ch 2. The amplitude-phase training is distinguished, namely, the channel Ch1 obtains the amplitude mapping relation between the magnetic field and the electric field in the channel Ch1 based on the training of the simulated magnetic field amplitude and the simulated electric field amplitude, the channel Ch1 obtains the phase mapping relation between the magnetic field and the electric field in the channel Ch1 based on the training of the simulated magnetic field phase and the simulated electric field phase, the channel Ch2 obtains the amplitude mapping relation between the magnetic field and the electric field in the channel Ch2 based on the training of the simulated magnetic field amplitude and the simulated electric field amplitude, and the channel Ch2 obtains the phase mapping relation between the magnetic field and the electric field in the channel Ch2 based on the training of the simulated magnetic field phase and the simulated electric field phase.
The training results of the training method of this embodiment are compared with those of the conventional training method.
Regarding the discrimination channel training:
fig. 9 provides a schematic diagram of training results of a conventional training method, in which simulation images of the channel Ch1 in the training set are used for training, and electric field distributions of the channels Ch1 to Ch8 of the subject are tested respectively. Wherein 9-1 represents the true electric field z-component; 9-2 represents a learning electric field z-component; 9-3 represents the distribution result of the error ratio of the three components of the electric field x, y and z, the abscissa represents the ratio of the learning value to the true value, the ordinate represents the frequency of occurrence of the error ratio of the three components of x, y and z, and the more the ratio is 1, the more accurate the prediction is illustrated; 9-4 represents the safety factor statistical result of the SAR, the abscissa represents the maximum predicted value of the SAR, the ordinate represents the maximum true value of the SAR, the SAR is calculated by the learning electric field and the true electric field respectively, and the maximum value of the ratio obtained by dividing the learning electric field and the true electric field is the safety factor (Safty factor) of the SAR. From the aspects of error ratio distribution and safety factors, the distribution of z-direction components of the learning electric field is greatly different from a true value, and the error ratio distribution of three components of the electric field x, y and z is far away from 1.
Fig. 10 provides a schematic diagram of training results of the training method of the present embodiment, in which simulation images of channels Ch1 to Ch8 in the training set are used for performing channel-division training, and corresponding channel tests are performed. Wherein 10-1 represents the true electric field z-component; 10-2 represents a learning electric field z-component; 10-3 represents the error ratio distribution result of the three components of the electric field x, y and z, the abscissa represents the ratio of the learning value to the true value, and the ordinate represents the number of pixels; 10-4 represents the safety factor statistical result of the SAR, the abscissa represents the maximum predicted value of the SAR, and the ordinate represents the maximum true value of the SAR. From the aspects of error ratio distribution and safety factors, the prediction result after the multichannel training is more accurate.
Regarding the differential amplitude phase training:
fig. 11 provides a second training result diagram of a conventional training method, in which the magnetic field and the electric field in the training set are both of the same magnitude and the phase is not trained, thus resulting in a large difference in the estimated SAR distribution. Where 11-1 represents the true electric field z-component phase angle and 11-2 represents the learned electric field z-component phase angle. It can be seen that the tested field phase is much different from the true value because only the magnitudes of the magnetic and electric fields are used for training.
Fig. 12 provides a second schematic diagram of training results of the training method of the present embodiment, where the amplitude and phase of the magnetic field and the electric field are trained. Wherein 12-1 represents the true electric field z-component amplitude; 12-2 represents the learning electric field z-component amplitude; 12-3 represents the true electric field z-component phase angle; 12-4 represents a learning electric field z-component phase angle; 12-5 represents the error ratio distribution result of the three components of the electric field x, y and z, the abscissa represents the ratio of the learning value to the true value, and the ordinate represents the number of pixels; 12-6 represents the safety factor statistical result of the SAR, the abscissa represents the maximum predicted value of the SAR, and the ordinate represents the maximum true value of the SAR. It can be seen that both the final error distribution and the safety factor are greatly improved.
In one embodiment, after training to obtain the deep learning model, the method further comprises: obtaining a second local specific absorption rate, wherein the second local specific absorption rate is obtained by conducting electromagnetic simulation on a human body model lead-in frequency transmitting coil model corresponding to the object to be detected; obtaining a safety factor according to the ratio of the first local specific absorption rate maximum value to the second local specific absorption rate maximum value in each channel; evaluating the prediction accuracy of the deep learning model according to the safety factor; and adjusting the weight of the deep learning model according to the prediction accuracy of the deep learning model.
The safety factor calculation formula is as follows:
where m=1, 2, …, N, SF stands for safety factor, s stands for specific absorption rate data size,representing the first local specific absorption rate maximum of the object to be measured in the channel m,/>Representing a second local specific absorption rate maximum of the object under test at channel m. The closer the obtained safety factor is to 1, the closer the predicted value is to the true value, and whether the weight of the deep learning model is adjusted can be determined through the difference between the safety factor and 1, so that the prediction accuracy of the deep learning model is further improved.
In one embodiment, as shown in fig. 13, a local specific absorption rate estimation method is provided, and the method is applied to the terminal in fig. 1 for illustration, and includes the following steps:
step S401, acquiring real magnetic field data of the scanned object to be detected.
The magnetic resonance system is adopted to scan the object to be detected, human tissue can interact with the radio frequency electromagnetic field generated by the radio frequency transmitting coil, the energy loaded by the radio frequency electromagnetic field is absorbed by hydrogen protons in the human body, part of the energy is transmitted in the form of radio frequency signals, and the signals are received by the receiving coil to generate signals, so that real magnetic field data are obtained.
Step S402, inputting the real magnetic field data into a trained deep learning model, predicting to obtain a radio frequency electric field of the object to be detected, and obtaining a first local specific absorption rate of the object to be detected according to the radio frequency electric field.
And finally obtaining the local specific absorption rate of the object to be measured through the weighted superposition of the radio frequency electric fields of all the channels. The first local specific absorption rate of the object to be measured is obtained according to the radio frequency electric field, and the method can be realized as follows:
acquiring the conductivity and density of an object to be measured; and determining a first local specific absorption rate of the object to be measured according to the conductivity, the density and the radio frequency electric field of the object to be measured, wherein the first local specific absorption rate is directly proportional to the conductivity, the first local specific absorption rate is inversely proportional to the density, and the first local specific absorption rate is directly proportional to the square of the radio frequency electric field. The first local specific absorption rate calculation formula is as follows:
wherein SAR represents the first local specific absorption rate of the object to be measured, sigma represents the conductivity of the object to be measured, ρ represents the density of the object to be measured, and E represents the radio frequency electric field. It can be seen that the accuracy of the prediction of the radio frequency electric field is a key factor affecting the accuracy of the local specific absorption rate.
After a typical human model library is constructed, the model library is imported into simulation software one by one to acquire the B field and the E field of each channel. And a part of the training set is used as a training set to be imported into a machine learning program to train a learning model. Another part is used as a test set to test the accuracy of the learning model: the B field of the subject is led into a learning model to obtain E field distribution,
The deep learning model of the embodiment can be obtained through training by the method: acquiring a training set, wherein the training set is obtained by processing the simulation image preprocessing method in any embodiment; and training the deep learning model according to the training set until the deep learning model meets the convergence condition.
By combining the embodiment, the simulation image is preprocessed to construct a complete and accurate training set, so that the relation between the training set and the deep learning model is more intimate, the trained deep learning model is more accurate, the problem of low accuracy of the local specific absorption rate predicted value of the object to be detected is solved, and the accuracy of the local specific absorption rate predicted value of the object to be detected is improved.
There is also provided in this embodiment an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.
In addition, in combination with the method provided in the above embodiment, a storage medium may also be provided in the present embodiment to realize. The storage medium has a computer program stored thereon; which when executed by a processor implements any of the methods of the above embodiments.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure thereof may be as shown in fig. 14. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in fig. 14 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements are applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to be limiting. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure in accordance with the embodiments provided herein.
It is to be understood that the drawings are merely illustrative of some embodiments of the present application and that it is possible for those skilled in the art to adapt the present application to other similar situations without the need for inventive work. In addition, it should be appreciated that while the development effort might be complex and lengthy, it will nevertheless be a routine undertaking of design, fabrication, or manufacture for those of ordinary skill having the benefit of this disclosure, and further having the benefit of this disclosure.
The term "embodiment" in this disclosure means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive. It will be clear or implicitly understood by those of ordinary skill in the art that the embodiments described in the present application can be combined with other embodiments without conflict.
The user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party. The embodiment of the application relates to the acquisition, storage, use, processing and the like of data, which all meet the relevant regulations of national laws and regulations.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the patent claims. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.
Claims (10)
1. A simulated image preprocessing method, characterized by comprising:
the method comprises the steps of obtaining a simulation image, wherein the simulation image is obtained by conducting a human body model to an electromagnetic transmission coil model, the simulation image comprises an electric field image and a magnetic field image, and the radio frequency transmission coil model comprises N channels, wherein N is more than or equal to 2;
identifying the internal field of the human body in the simulation image, and setting the corresponding field intensity of the critical positions of the internal part of the human body and the air to zero;
dividing the simulation image according to the N channels to obtain a simulation image corresponding to each channel;
extracting the amplitude and the phase of the internal field of the human body in the simulation image corresponding to each channel to obtain a training set;
The training set is used for training a deep learning model, and the deep learning model is used for predicting a radio frequency electric field.
2. The simulated image preprocessing method of claim 1, wherein said identifying a human body internal field in said simulated image comprises:
in the electric field image, determining the area of the first human body internal field according to the human body tissue medium parameter and the air medium parameter;
covering a first mask in the electric field image, wherein the area where the first human body internal field is located is not covered by the first mask; the method comprises the steps of,
in the magnetic field image, determining the region of the second human body internal place according to the human body tissue medium parameters and the air medium parameters;
and covering a second mask in the magnetic field image, wherein the second human body internal place is not covered by the second mask in a region.
3. The simulated image preprocessing method of claim 1, wherein said zeroing the respective field strengths of the critical locations of the human body interior and air comprises:
in the electric field image, determining a first critical position of the interior of the human body and air according to a human tissue medium parameter and an air medium parameter, and setting the electric field intensity of the first critical position to zero; and
In the magnetic field image, a second critical position of the interior of the human body and air is determined according to the human body tissue medium parameter and the air medium parameter, and the magnetic field intensity of the second critical position is set to zero.
4. The simulated image preprocessing method of claim 1, wherein said radio frequency transmit coil model comprises a multi-channel array transmit coil model, a full-volume transmit coil model or a partial-volume transmit coil model; and/or, the magnetic field intensity of the radio frequency transmitting coil model is not lower than 3T.
5. A local specific absorption rate estimation method, comprising:
acquiring real magnetic field data of an object to be detected after scanning;
inputting the real magnetic field data into a trained deep learning model, predicting to obtain a radio frequency electric field of the object to be detected, and obtaining a first local specific absorption rate of the object to be detected according to the radio frequency electric field; wherein training the deep learning model comprises:
obtaining a training set, wherein the training set is obtained by processing the simulation image preprocessing method according to any one of the claims 1 to 4;
and training the deep learning model according to the training set until the deep learning model meets convergence conditions.
6. The method of estimating local specific absorption rate according to claim 5, wherein obtaining the first local specific absorption rate of the object to be measured according to the radio frequency electric field includes:
acquiring the conductivity and density of the object to be measured;
and determining a first local specific absorption rate of the object to be measured according to the conductivity of the object to be measured, the density and the radio frequency electric field, wherein the first local specific absorption rate is directly proportional to the conductivity, the first local specific absorption rate is inversely proportional to the density, and the first local specific absorption rate is directly proportional to the square of the radio frequency electric field.
7. The local specific absorption rate estimation method according to claim 5, wherein the training set includes simulated magnetic field data and simulated electric field data, wherein the simulated magnetic field data includes simulated magnetic field amplitude and simulated magnetic field phase, and wherein the simulated electric field data includes simulated electric field amplitude and simulated electric field phase, and wherein training the deep learning model according to the training set includes:
inputting the simulated magnetic field amplitude and the simulated electric field amplitude of the mth channel into the deep learning model to obtain an amplitude mapping relation between the magnetic field and the electric field in the mth channel;
And inputting the simulated magnetic field phase and the simulated electric field phase of the mth channel into the deep learning model to obtain a phase mapping relation between the magnetic field phase and the electric field phase in the mth channel.
8. The local specific absorption rate estimation method according to claim 5, further comprising:
obtaining a second local specific absorption rate, wherein the second local specific absorption rate is obtained by electromagnetic simulation of a human body model guided-incidence frequency transmitting coil model corresponding to the object to be detected;
obtaining a safety factor according to the ratio of the first local specific absorption rate maximum value to the second local specific absorption rate maximum value in each channel;
evaluating the prediction accuracy of the deep learning model according to the safety factor;
and adjusting the weight of the deep learning model according to the prediction accuracy of the deep learning model.
9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of claims 1 to 8.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310727563.XA CN117197521A (en) | 2023-06-19 | 2023-06-19 | Simulation image preprocessing method, local specific absorption rate estimation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310727563.XA CN117197521A (en) | 2023-06-19 | 2023-06-19 | Simulation image preprocessing method, local specific absorption rate estimation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117197521A true CN117197521A (en) | 2023-12-08 |
Family
ID=89000492
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310727563.XA Pending CN117197521A (en) | 2023-06-19 | 2023-06-19 | Simulation image preprocessing method, local specific absorption rate estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117197521A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117594193A (en) * | 2024-01-17 | 2024-02-23 | 西安电子科技大学 | Transcranial direct current personalized stimulation target positioning method based on deep learning |
-
2023
- 2023-06-19 CN CN202310727563.XA patent/CN117197521A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117594193A (en) * | 2024-01-17 | 2024-02-23 | 西安电子科技大学 | Transcranial direct current personalized stimulation target positioning method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN205665697U (en) | Medical science video identification diagnostic system based on cell neural network or convolution neural network | |
| CN104603630B (en) | Magnetic resonance imaging system with the motion detection based on omniselector | |
| CN105100789A (en) | Method for evaluating video quality | |
| CN105308473B (en) | The Automatic Optimal of parallel imaging acceleration parameter | |
| CN117197521A (en) | Simulation image preprocessing method, local specific absorption rate estimation method and device | |
| CN101853504A (en) | Image quality evaluating method based on visual character and structural similarity (SSIM) | |
| CN106793971A (en) | MR imaging apparatus | |
| CN111986172A (en) | Infrared image fault detection method and device for power equipment | |
| CN105894522B (en) | A kind of more distortion objective evaluation method for quality of stereo images | |
| CN111640095B (en) | Quantification method of cerebral micro hemorrhage and computer readable storage medium | |
| CN105051564A (en) | A method for improved k-space sampling in compressed sensing MRI | |
| CN109741416A (en) | Nuclear magnetic resonance image method for reconstructing, device, computer equipment and its storage medium | |
| CN104101843B (en) | Decoupling of parallel transmission arrays in magnetic resonance imaging | |
| CN108847899A (en) | Electronic equipment radio communication Performance Test System, method, device and equipment | |
| CN102066966B (en) | For testing the electronic load simulator apparatus of RF coil | |
| CN113017598B (en) | Radio frequency energy deposition prediction and monitoring method, device, equipment and medium | |
| CN109039486A (en) | Multi-user's multiple-input, multiple-output equipment communication performance test macro, method and device | |
| CN105187139B (en) | A kind of outdoor radio signal reception strength map constructing method based on intelligent perception | |
| Singh et al. | 3-D EM modeling of medical microwave imaging scenarios with controllable accuracy | |
| CN117289097A (en) | Power equipment partial discharge detection method, model training method, device and equipment | |
| CN102369452A (en) | Noise matching in couplet antenna arrays | |
| CN104515961B (en) | MR imaging method and device | |
| CN115714732A (en) | Whole-house wireless network coverage condition detection method and equipment | |
| CN103230274B (en) | Computing method of dispersion magnetic resonance image and analysis method based on computing method | |
| WO2019179838A1 (en) | Anomaly detection using magnetic resonance fingerprinting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |