Method, device and equipment for improving spatial resolution of magnetic particle imaging

Abstract

The application relates to the technical field of magnetic particle imaging, and discloses a method, device and equipment for improving the spatial resolution of magnetic particle imaging, the method comprising the following steps: acquiring SM component data, and converting the SM component data from a complex number field to an RGB field to obtain SM sample data; adding white noise to the SM sample data to obtain SM noise-added sample data, and the SM sample data and the corresponding SM noise-added sample data forming a matching data pair; training a preset SM denoising model by using the matching data pair to obtain an SM denoising model; denoising actual SM data by using the SM denoising model, and reconstructing a magnetic particle image by using the denoised actual SM data. The SM denoising model is trained, the SM component is denoised, the noise information of the reconstructed magnetic particle image is reduced, the spatial resolution of the magnetic particle imaging is improved, a high-resolution imaging result is obtained, and the doctor is better assisted in clinical image interpretation and diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of magnetic particle imaging, and particularly relates to a method and device for improving spatial resolution of magnetic particle imaging and equipment. BACKGROUND

[0002] Magnetic particle imaging (MPI) is a new type of molecular imaging technology, and its working principle is to excite magnetic nanoparticles (MNPs) to generate a magnetization signal by using a magnetic field and to receive the signal through a receiving coil, and then to reconstruct the MNP concentration distribution map from the signal. MPI has the advantages of no imaging depth limit, high sensitivity, no background signal and no ionizing radiation, and has shown great advantages in applications such as cell tracking, angiography and drug delivery.

[0003] High spatial resolution means that the imaging image can provide clear and accurate details, textures and small size visual feature information, thereby better assisting doctors in clinical image interpretation. However, due to the noise and harmonic interference of the MPI system, the calibrated system matrix (SM) contains noise information, resulting in problems of noise in the imaging space and reduced resolution. SUMMARY

[0004] Therefore, the present application provides a method, device and equipment for improving spatial resolution of magnetic particle imaging to solve the problem of low imaging spatial resolution of the existing magnetic particle imaging system.

[0005] In a first aspect, the present application provides a method for improving spatial resolution of magnetic particle imaging, and the method comprises the following steps:

[0006] obtaining SM component data and converting the SM component data from a complex number field to an RGB field to obtain SM sample data;

[0007] adding white noise to the SM sample data to obtain SM noise sample data, and the SM sample data and the corresponding SM noise sample data constitute a matching data pair;

[0008] training a preset SM denoising model using the matching data pair to obtain an SM denoising model;

[0009] denoising actual SM data using the SM denoising model to obtain denoised actual SM data, and reconstructing a magnetic particle image using the denoised actual SM data.

[0010] The method for improving the spatial resolution of magnetic particle imaging provided by the application realizes SM component denoising by training an SM denoising model, reduces the noise information of the reconstructed magnetic particle image, and further improves the spatial resolution of MPI, so that a high-resolution imaging result is obtained, which better assists doctors in clinical image interpretation and diagnosis.

[0011] In an optional implementation, the SM component data is obtained, including:

[0012] A lower threshold of the SM component is set according to the effectiveness of the SM component;

[0013] The SM component calibrated by the magnetic particle imaging is compared with the lower threshold, and the SM component that is not less than the lower threshold is taken as the SM component data.

[0014] The method for improving the spatial resolution of magnetic particle imaging provided by the application sets the lower threshold of the SM component, screens the effective SM component, avoids further operation on invalid data, saves the computing resource, and improves the processing efficiency of the SM component data.

[0015] In an optional implementation, the SM component data is converted from a complex number field to an RGB field to obtain SM sample data, including:

[0016] The SM component data is converted into an HSV vector by using an HSV color model;

[0017] The HSV vector is converted into an RGB vector by using a preset HSV-RGB conversion standard;

[0018] The SM sample data in the RGB field is obtained by using amplitude linear scaling conversion.

[0019] The method for improving the spatial resolution of magnetic particle imaging provided by the application converts the HSV vector into the RGB vector, which is conducive to accurate representation and processing of color data, improves the accuracy and efficiency of color processing, and is conducive to magnetic particle image reconstruction by using the SM component data.

[0020] In an optional implementation, the preset SM denoising model includes a feature extraction unit, a multi-scale information mining unit and a multi-scale information fusion unit, wherein,

[0021] The feature extraction unit is configured to extract feature information in the noisy SM component data;

[0022] The multi-scale information mining unit is configured to extract useful feature information of each preset scale from the feature information according to a plurality of preset scales;

[0023] The multi-scale information fusion unit is configured to fuse the useful feature information of each preset scale to obtain useful information in the noisy SM component.

[0024] The method for improving the spatial resolution of magnetic particle imaging provided by this invention obtains a deep learning model by training a preset SM denoising model, and uses the deep learning model to denoise the SM components to support the acquisition of high spatial resolution imaging results, providing clearer and more accurate details, textures and small-size visual information for the imaging results.

[0025] In one optional implementation, a preset SM denoising model is trained using matched data to obtain the SM denoising model, including:

[0026] The matched data pairs are divided into training set, test set, and validation set according to a preset ratio;

[0027] The pre-defined SM denoising model is trained using the training set and validation set, and the trained model is evaluated using the test set to obtain the SM denoising model.

[0028] The method for improving the spatial resolution of magnetic particle imaging provided by this invention obtains a SM denoising model through deep learning model training, performs denoising processing on noisy SM data, and then uses the denoised SM data to reconstruct the image, thereby improving the spatial resolution of MPI and reducing the hardware design and task complexity of improving spatial resolution solely through hardware improvements.

[0029] In one optional implementation, a preset SM denoising model is trained using a training set and a validation set, and the trained model is validated using a test set to obtain the SM denoising model, including:

[0030] Randomly initialize the parameters of the preset SM denoising model;

[0031] The SM-denoised sample data in the training set and validation set are used as the model input, and the corresponding SM sample data are used as the label. They are input into the preset SM denoising model, and the parameters of the preset SM denoising model are updated to obtain the SM denoising training model.

[0032] The performance of the SM denoising training model is evaluated using a test set to determine whether it meets the preset performance requirements. Models that meet the preset performance requirements are used as SM denoising models.

[0033] The method for improving the spatial resolution of magnetic particle imaging provided by this invention uses a deep learning model for SM denoising, which can automatically extract global and local information of SM components and learn the ability to remove noise from training data pairs. This information is then directly applied to the acquired SM components to improve the spatial resolution of MPI.

[0034] Secondly, the present invention provides an apparatus for improving the spatial resolution of magnetic particle imaging, the apparatus comprising:

[0035] The sample acquisition module is used to acquire SM component data and convert the SM component data from the complex domain to the RGB domain to obtain SM sample data;

[0036] The data noise-adding module is used to add white noise to the SM sample data to obtain SM noise-adding sample data. The SM sample data and the corresponding SM noise-adding sample data form a matching data pair.

[0037] The model training module is used to train the preset SM denoising model using matching data to obtain the SM denoising model.

[0038] The SM denoising module is used to denoise the actual SM data using the SM denoising model, obtain the denoised actual SM data, and reconstruct the magnetic particle image using the denoised actual SM data.

[0039] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0040] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof.

[0041] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0042] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0043] Figure 1 This is a schematic flowchart of a method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention;

[0044] Figure 2 This is a flowchart illustrating another method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention;

[0045] Figure 3This is a schematic diagram of the SM denoising model in the method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram of the SM denoising model training process in the method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention;

[0047] Figure 5 This is a schematic diagram illustrating the input and output of the SM denoising model using a test set as input in a method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention.

[0048] Figure 6 This is a comparative diagram of image reconstruction results obtained by imaging using the denoised SM components in the method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention.

[0049] Figure 7 This is a structural block diagram of a device for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention;

[0050] Figure 8 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] This invention provides a method for improving the spatial resolution of magnetic particle imaging. By removing noise in the SM, the noise information of the reconstructed image is reduced, thereby improving the spatial resolution of MPI.

[0053] According to an embodiment of the present invention, a method embodiment for improving the spatial resolution of magnetic particle imaging is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0054] This embodiment provides a method for improving the spatial resolution of magnetic particle imaging, which can be used in the aforementioned computer system. Figure 1 This is a flowchart of a method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0055] Step S101: Obtain SM component data and convert the SM component data from the complex domain to the RGB domain to obtain SM sample data.

[0056] Specifically, in the image reconstruction process based on the system matrix, the imaging field of view first needs to be discretized. A system matrix is ​​constructed by scanning small samples point by point. The linear mapping relationship between the system matrix and the measured object is then used to inversely solve for the spatial distribution of magnetic particles. However, since the encoding method of magnetic field free lines cannot decode the position of magnetic particles along the direction of the magnetic field free lines, in practice, it is necessary to measure the magnetic field free lines at multiple angles to achieve the position decoding of magnetic particles. Accordingly, in the system matrix reconstruction method, system matrices measured at multiple angles need to be stitched together before use. Therefore, the system matrix directly affects the imaging effect of the reconstructed image. For multiple system matrices, the component data of each system matrix are obtained, and the SM component data is transformed from the complex domain to the RGB domain. The SM components in the RGB domain are used as SM sample data for training the preset SM denoising model.

[0057] Step S102: Add white noise to the SM sample data to obtain SM noisy sample data. The SM sample data and the corresponding SM noisy sample data form a matching data pair.

[0058] Specifically, in this embodiment of the invention, the `norm` function of the `scipy` package is used to add white noise to the SM sample data, resulting in noisy SM sample data. The SM sample data and the corresponding noisy SM sample data form a matched data pair. The process of adding white noise to the sample data is a mature technology and will not be described in detail here.

[0059] Step S103: Use the matching data to train the preset SM denoising model to obtain the SM denoising model.

[0060] Specifically, in this embodiment of the invention, the `train_test_split` function in scipy is used to divide all SM components into a training set and a test set, wherein the proportions of the training set and the test set are as follows: and Then randomly select from the training set The training and test sets are used as validation sets. The model is trained using the training and test sets, and evaluated using the validation set. Finally, the optimal SM denoising model that meets the preset requirements is obtained.

[0061] Step S104: Use the SM denoising model to denoise the actual SM data to obtain the denoised actual SM data, and use the denoised actual SM data to reconstruct the magnetic particle image.

[0062] Specifically, in the actual magnetic particle imaging space, actual SM data is acquired and denoised using an SM denoising model to obtain denoised actual SM data. The magnetic particle image is then reconstructed using the denoised actual SM data. This process is a mature technical solution and will not be elaborated here. Since the SM data has been denoised in advance, the magnetic particle image reconstructed using the denoised SM data has a higher resolution.

[0063] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment achieves denoising of the SM component by training a SM denoising model, thereby reducing noise information in the reconstructed magnetic particle image and improving the spatial resolution of MPI, resulting in high-resolution imaging results, which can better assist doctors in clinical image interpretation and diagnosis.

[0064] This embodiment provides a method for improving the spatial resolution of magnetic particle imaging, which can be used in the aforementioned computer system. Figure 2 This is a flowchart of a method for improving the spatial resolution of magnetic particle imaging according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0065] Step S201: Obtain SM component data and convert the SM component data from the complex domain to the RGB domain to obtain SM sample data.

[0066] Specifically, step S201 includes:

[0067] Step S2011: Set the lower threshold of the SM component based on the validity of the SM component.

[0068] Specifically, 'a' is set as the lower threshold of valid SM components. That is, SM components greater than or equal to 'a' can be used and retained; SM components less than 'a' are unusable and filtered out. The value of 'a' is set according to the actual situation, and the specific setting standard is not limited here.

[0069] Step S2012: Compare the SM component calibrated by magnetic particle imaging with the lower threshold limit, and take the SM component that is not less than the lower threshold limit as the SM component data.

[0070] Specifically, from the calibrated SM, the signal-to-interference-plus-noise ratio (SNR) is selected. SM components SM components that are valid can be used, while other SM components smaller than a are not usable and are filtered out.

[0071] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment filters out effective SM components by setting a lower threshold for SM components, avoiding further processing of invalid data, saving computational resources, and improving the processing efficiency of SM component data.

[0072] Step S2013: Using the HSV color model, convert the SM component data into an HSV vector.

[0073] Specifically, the HSV (Hue, Saturation, Value) color model is used to represent phase. ,Right now It should be noted that each value of the SM component is in the complex form of a + bj, which can be converted into a vector. To represent, and then convert each value of the SM component into ( The format is ), where the 0th bit represents the phase angle, the 1st bit is 1, and the 2nd bit is the absolute value of the variable modulus.

[0074] Step S2014: Convert the HSV vector into an RGB vector using the preset HSV-RGB conversion standard.

[0075] Specifically, using a preset HSV-RGB conversion standard, HSV vectors are converted into RGB vectors. The specific conversion process involves mature existing technologies, which will not be elaborated here.

[0076] Step S2015: Obtain SM sample data in the RGB domain by using amplitude linear scaling transformation.

[0077] Specifically, the SM sample data refers to the SM components that are ultimately converted to the RGB domain, represented as follows: .

[0078] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment converts HSV vectors into RGB vectors, which is beneficial for accurately representing and processing color data, improving the accuracy and efficiency of color processing, and facilitating the reconstruction of magnetic particle images using SM component data.

[0079] Step S202: Add white noise to the SM sample data to obtain SM noisy sample data. The SM sample data and the corresponding SM noisy sample data form a matched data pair. For details, please refer to [link to details]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0080] Step S203: Use the matching data to train the preset SM denoising model to obtain the SM denoising model.

[0081] In some alternative implementations, a pre-defined SM denoising model based on deep learning, such asFigure 3 As shown, it includes:

[0082] The feature extraction unit is used to extract feature information from noisy SM component data.

[0083] Specifically, using a CNN-based backbone, the noisy SM component input to the preset SM denoising model is processed through convolution, pooling, and activation layers to obtain the feature map of the noisy SM. The channel attention mechanism is used to calculate the attention weight of each feature map in the high and low frequency feature channels, and multiply it with the corresponding feature map. The results are used as high and low frequency information, that is, feature information, which contains useful information and noise information.

[0084] The multi-scale information mining unit is used to extract useful feature information of each preset scale from feature information according to multiple preset scales.

[0085] Specifically, useful feature information at different scales, such as overall and detailed information, is extracted from the feature map of the noisy SM by using receptive fields of different sizes: large receptive fields extract overall information from the noisy SM, while small receptive fields focus on abstracting detailed information from the noisy SM. Each information extraction part includes convolutional layers, pooling layers, and LeakyReLU activation layers. The convolutional layers further abstract the feature map, while the other two layers fuse information from different receptive fields and produce non-linear outputs. The output of each layer serves as the input to the next layer. Each information extraction part also includes a transition layer to process the feature map to the same size.

[0086] The multi-scale information fusion unit is used to fuse useful feature information from each preset scale to obtain useful information in the noisy SM component.

[0087] Specifically, after acquiring multiple global and local information of the SM, the useful feature information of each preset scale extracted under different receptive fields is fused globally and locally, and then useful information is selected from the noisy SM components and noise information is removed.

[0088] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment obtains a deep learning model by training a preset SM denoising model, and uses the deep learning model to denoise the SM components to support the acquisition of high spatial resolution imaging results, providing clearer and more accurate details, textures and small-size visual information for the imaging results.

[0089] Specifically, step S203 includes:

[0090] Step S2031: Divide the matched data pairs into training set, test set and validation set according to a preset ratio.

[0091] Specifically, the `train_test_split` function in SciPy is used to divide all matched data pairs into training and test sets, with the training and test sets comprising a certain percentage. and Then randomly select from the training set As a validation set.

[0092] Step S2032: Train the preset SM denoising model using the training set and validation set, and evaluate the trained model using the test set to obtain the SM denoising model.

[0093] Specifically, the network input is SM-added noisy sample data with added noise in the training set, and the SM sample data without added noise is used as labels to supervise the training of the preset SM denoising model. The optimization algorithm used for model supervision training is stochastic gradient descent (SGD), and the loss function is MSE. This is just an example, but not a limitation.

[0094] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment obtains a SM denoising model through deep learning model training, performs denoising processing on noisy SM data, and then uses the denoised SM data to reconstruct the image, thereby improving the spatial resolution of MPI and reducing the hardware design and task complexity of improving spatial resolution solely through hardware improvements.

[0095] In some alternative implementations, such as Figure 4 As shown, the model training process in step S2032 includes:

[0096] Step a1: Randomly initialize the parameters of the preset SM denoising model.

[0097] Specifically, load the preset SM denoising model that needs to be trained and randomly initialize its parameters.

[0098] Step a2: Use the SM-denoised sample data from the training set and validation set as model input, and the corresponding SM sample data as labels, and input them together into the preset SM denoising model. Update the parameters of the preset SM denoising model to obtain the SM denoising training model.

[0099] Specifically, before each training session, the SMs in the training set are preprocessed by random 90° rotation and random reversal, and the preprocessed data is input into the preset SM denoising model for forward propagation to obtain the processing result.

[0100] The mean squared error was selected as the loss function for the processing results, and the training error was backpropagated. The Adam optimizer was then selected to adjust the model parameters.

[0101] If the preset number of training sessions is not reached, this step needs to be repeated until the preset number of training sessions is reached and the training ends.

[0102] The model trained within the training iterations is validated using a validation set. The model that meets the evaluation criteria is used as the prediction model, i.e., the SM denoising training model. The evaluation criteria can be nRMSE or signal-noise ratio (SNR), which are just examples and are not limited to this.

[0103] Step a3: Use the test set to evaluate whether the performance of the SM denoising training model meets the preset performance requirements, and use the model that meets the preset performance requirements as the SM denoising model.

[0104] Specifically, the preset performance requirement could be the preset image resolution obtained by image reconstruction using the SM denoising training model, or whether the reconstructed image resolution is improved. This is just an example and not a limitation. The image resolution can be visualized using the curve of the half-value weight width. The SM denoising model is tested using noisy SM sample data from the test set as input, and unnoisy SM sample data as labels. The denoised SM components corresponding to the input are obtained. The input and output results are as follows: Figure 5 As shown, where Figure 5 (a) is the SM-added noisy sample data. Figure 5 (b) The denoised SM component obtained by the SM denoising model from the SM-denoised sample data, and compared with... Figure 5 (a) and Figure 5 (b) It can be seen that, Figure 5 (b) The noise is reduced.

[0105] Imaging was performed using the SM components before and after denoising, and analysis was conducted through full width at half maximum (FWHM) weighted width visualization to obtain the following results: Figure 6 The comparison results shown include, Figure 6 (a) The rightmost image shows the image reconstruction result using the SM components before denoising, where the blue solid line represents the pixel values ​​of each point in the 162nd row, corresponding to the blue broken line in the leftmost half-height weighted width visualization; the middle image shows the image reconstruction result using the SM components after denoising, where the red solid line represents the pixel values ​​of each point in the 162nd row, corresponding to the red broken line in the leftmost half-height weighted width visualization. Figure 6(b) The rightmost image shows the image reconstruction result using the SM components before denoising, where the blue solid line represents the pixel values ​​of each point in the 248th row, corresponding to the blue broken line in the leftmost half-height weighted width visualization; the middle image shows the image reconstruction result using the SM components after denoising, where the red solid line represents the pixel values ​​of each point in the 248th row, corresponding to the red broken line in the leftmost half-height weighted width visualization.

[0106] according to Figure 6 As can be seen from the half-height weighted width visualization in (a), at the arrow, the highest point of the red line is higher than that of the blue line. According to... Figure 6 As shown in the half-height weighted width visualization in (b), at the arrow, the lowest point of the red line is lower than that of the blue line. This indicates that the two branches of the imaging target are separated more finely, with higher pixel values ​​appearing higher and lower values ​​appearing lower, thus the two branches are separated better. If the lines in the half-height weighted width visualization become straight lines, it means that nothing can be separated.

[0107] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment uses a deep learning model for SM denoising, which can automatically extract global and local information of SM components and learn noise removal capabilities from training data pairs. This information is then directly applied to the acquired SM components to improve the spatial resolution of MPI.

[0108] Step S204: Denoise the actual SM data using the SM denoising model to obtain denoised actual SM data, and reconstruct the magnetic particle image using the denoised actual SM data. For details, please refer to [link to details]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0109] The method for improving the spatial resolution of magnetic particle imaging provided in this embodiment achieves denoising of the SM component by training a SM denoising model, thereby reducing noise information in the reconstructed magnetic particle image and improving the spatial resolution of MPI, resulting in high-resolution imaging results, which can better assist doctors in clinical image interpretation and diagnosis.

[0110] In a specific embodiment, for example, using SM 6 from the OpenMPI dataset for experimentation, the method for improving the spatial resolution of magnetic particle imaging provided by this invention is used to construct sample pairs for training the model by adding noise to SMs with high SNR, and then denoising SMs with low SNR. Comparing image reconstructions of SMs before and after denoising reveals that the resolution of the reconstructed image is improved after denoising the low SNR SMs. This improves the signal-to-noise ratio of SMs without modifying system hardware settings, thereby improving imaging resolution.

[0111] This embodiment also provides a device for improving the spatial resolution of magnetic particle imaging. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0112] This embodiment provides a device for improving the spatial resolution of magnetic particle imaging, such as... Figure 7 As shown, it includes:

[0113] The sample acquisition module 701 is used to acquire SM component data and convert the SM component data from the complex domain to the RGB domain to obtain SM sample data.

[0114] The data noise-adding module 702 is used to add white noise to the SM sample data to obtain SM noise-adding sample data. The SM sample data and the corresponding SM noise-adding sample data form a matching data pair.

[0115] The model training module 703 is used to train the preset SM denoising model using the matched data to obtain the SM denoising model.

[0116] The SM denoising module 704 is used to denoise the actual SM data using the SM denoising model to obtain the denoised actual SM data, and to reconstruct the magnetic particle image using the denoised actual SM data.

[0117] In some alternative implementations, the sample acquisition module 701 includes:

[0118] The threshold setting unit is used to set the lower threshold of the SM component based on the validity of the SM component.

[0119] The effective data filtering unit is used to compare the SM components calibrated by magnetic particle imaging with the lower threshold limit, and to use the SM components that are not less than the lower threshold limit as SM component data.

[0120] The HSV vector conversion unit is used to convert SM component data into HSV vectors using the HSV color model.

[0121] The RGB vector conversion unit is used to convert HSV vectors into RGB vectors using a preset HSV-RGB conversion standard.

[0122] The RGB domain data determination unit is used to obtain SM sample data in the RGB domain by using amplitude linear scaling transformation.

[0123] In some alternative implementations, the model training module 703 includes:

[0124] The sample data partitioning unit is used to divide the matched data pairs into training set, test set, and validation set according to a preset ratio.

[0125] The model training and evaluation unit is used to train the preset SM denoising model using the training set and validation set, and to evaluate the trained model using the test set to obtain the SM denoising model.

[0126] In some optional implementations, the model training evaluation unit includes:

[0127] The parameter initialization subunit is used to randomly initialize the parameters of the preset SM denoising model.

[0128] The model training subunit is used to take the SM-denoised sample data from the training set and validation set as model input, and the corresponding SM sample data as labels, and input them together into the preset SM denoising model to update the parameters of the preset SM denoising model, thereby obtaining the SM denoising training model.

[0129] The model evaluation subunit is used to evaluate whether the performance of the SM denoising training model meets the preset performance requirements using the test set, and the model that meets the preset performance requirements is used as the SM denoising model.

[0130] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0131] In this embodiment, the device for improving the spatial resolution of magnetic particle imaging is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0132] This invention also provides a computer device having the above-described features. Figure 7 The device shown is designed to improve the spatial resolution of magnetic particle imaging.

[0133] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 8As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 8 Take a processor 10 as an example.

[0134] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0135] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0136] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device 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.

[0137] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0138] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0139] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0140] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0141] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method of improving spatial resolution of magnetic particle imaging, characterized by, The method comprises: obtaining SM component data, and converting the SM component data from a complex number field to an RGB field to obtain SM sample data; adding white noise to the SM sample data to obtain SM noisy sample data, and the SM sample data and the corresponding SM noisy sample data constitute a matching data pair; training a preset SM denoising model by using the matching data pair to obtain an SM denoising model; denoising actual SM data by using the SM denoising model to obtain denoised actual SM data, and reconstructing a magnetic particle image by using the denoised actual SM data.

2. The method of claim 1, wherein, The method comprises: setting a lower threshold of SM components according to the effectiveness of the SM components; comparing the calibrated SM components in magnetic particle imaging with the lower threshold, and taking the SM components not less than the lower threshold as SM component data.

3. The method according to claim 1 or 2, characterized in that, The method comprises: converting the SM component data into an HSV vector by using an HSV color model; converting the HSV vector into an RGB vector by using a preset HSV-RGB conversion standard; obtaining SM sample data in the RGB field by using linear scaling conversion of the amplitude.

4. The method of claim 1, wherein, The preset SM denoising model comprises a feature extraction unit, a multi-scale information mining unit and a multi-scale information fusion unit, wherein the feature extraction unit is configured to extract feature information in noisy SM component data; the multi-scale information mining unit is configured to extract useful feature information of each preset scale from the feature information according to a plurality of preset scales; the multi-scale information fusion unit is configured to fuse the useful feature information of each preset scale to obtain useful information in the noisy SM component.

5. The method of claim 1, wherein, The method comprises: dividing the matching data pair into a training set, a test set and a validation set according to a preset proportion; training the preset SM denoising model by using the training set and the validation set, and evaluating the trained model by using the test set to obtain an SM denoising model.

6. The method of claim 5, wherein, The method comprises: randomly initializing parameters of the preset SM denoising model; inputting SM noisy sample data in the training set and the validation set as model input, and inputting corresponding SM sample data as labels into the preset SM denoising model, updating parameters of the preset SM denoising model to obtain an SM denoising training model; evaluating whether the performance of the SM denoising training model meets a preset performance requirement by using the test set, and taking a model meeting the preset performance requirement as an SM denoising model.

7. An apparatus for improving spatial resolution of magnetic particle imaging, characterized in that, The device comprises: a sample acquisition module configured to obtain SM component data, and convert the SM component data from a complex number field to an RGB field to obtain SM sample data; a data noise adding module configured to add white noise to the SM sample data to obtain SM noisy sample data, and the SM sample data and the corresponding SM noisy sample data constitute a matching data pair; The model training module is configured to train a preset SM denoising model by using the matching data pair, and obtain an SM denoising model. The SM denoising module is configured to denoise actual SM data by using the SM denoising model, obtain denoised actual SM data, and reconstruct a magnetic particle image by using the denoised actual SM data.

8. A computer device, comprising: The method comprises the following steps: The memory and the processor are in communication connection with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the method in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to make a computer execute the method in any one of claims 1 to 6.

10. A computer program product, characterised in that, The computer instructions are used to make a computer execute the method in any one of claims 1 to 6.

Images