Data processing method and device, electronic equipment and storage medium

Abstract

The application discloses a data processing method and device, electronic equipment and a storage medium. The method comprises the following steps: collecting first data by using a magnetic induction tomography (MIT) device, wherein the first data comprises one-dimensional sequence data; performing conversion processing on the first data by using a Gram angle field (GAF) algorithm to obtain second data, wherein the second data comprises two-dimensional image data; generating a first image by using the second data and a first model; and generating an MIT image by using the first image and a second model. The scheme provided in the application can reduce the requirements for the precision and data volume of the MIT device, and can improve the resolution of the MIT image.

Description

Technical Field

[0001] This application relates to the field of imaging algorithms, and more particularly to a data processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] Magnetic induction tomography (MIT) is a magnetically excited, non-contact imaging technique. In MIT imaging, the target object is placed in the area to be measured. An excitation coil generates a magnetic field and applies excitation to the area. A detection coil detects the physical changes within the area, allowing for the reconstruction of an image showing the distribution of the target object's internal conductivity based on these changes.

[0003] However, when using the MIT imaging method in related technologies, the imaging results have low resolution. Summary of the Invention

[0004] To address the related technical problems, embodiments of this application provide a data processing method, apparatus, electronic device, and storage medium.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides a data processing method, including:

[0007] First data was collected using MIT equipment, and the first data contained one-dimensional sequence data.

[0008] The first data is transformed using the Gramian Angular Field (GAF) algorithm to obtain the second data, which includes two-dimensional image data.

[0009] Using the second data and the first model, a first image is generated;

[0010] Using the first image and the second model, generate the MIT image.

[0011] In the above scheme, the step of using the GAF algorithm to transform the first data to obtain the second data includes:

[0012] The first data is compared with the baseline data to obtain the comparison result;

[0013] The comparison results are transformed using the GAF algorithm to obtain the second data.

[0014] The above scheme, the method further includes:

[0015] In the absence of a target object within the measured object field, physical quantities of the measured object field are collected to obtain third data;

[0016] The third data is used as the baseline data.

[0017] In the above scheme, generating the first image using the second data and the first model includes:

[0018] The first image is generated using the second data and the residual network; wherein,

[0019] The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction result of the first model and the actual result.

[0020] In the above scheme, generating the MIT image using the first image and the second model includes:

[0021] The MIT image is generated using the first image and the adversarial network.

[0022] The method in the above scheme further includes:

[0023] The imaging region of the measured object field is divided into triangles to obtain multiple imaging sub-regions;

[0024] The conductivity of each imaging sub-region in multiple imaging sub-regions is determined to obtain the third data.

[0025] The first model is trained using the conductivity of multiple imaging sub-regions.

[0026] The method in the above scheme further includes:

[0027] Obtain first information, which represents the recall rate of samples during the training of the first model;

[0028] The first model is trained using the third data and the first information.

[0029] This application also provides a data processing apparatus, including:

[0030] A data acquisition unit is used to acquire first data using MIT equipment, the first data comprising one-dimensional sequence data;

[0031] The processing unit is configured to use the GAF algorithm to transform the first data to obtain second data, the second data including two-dimensional image data; use the second data and the first model to generate a first image; and use the first image and the second model to generate an MIT image.

[0032] This application also provides an electronic device, including: a processor and a memory for storing a computer program capable of running on the processor.

[0033] When the processor runs the computer program, it executes the steps of any of the above methods.

[0034] This application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the above methods.

[0035] The data processing method, apparatus, electronic device, and storage medium provided in this application embodiment acquire first data using a MIT device, the first data comprising one-dimensional sequence data; the first data is transformed using the GAF algorithm to obtain second data, the second data comprising two-dimensional image data; a first image is generated using the second data and a first model; and an MIT image is generated using the first image and the second model. The solution provided in this application embodiment converts the data acquired by the MIT device into image data, and then uses the image data for pre-imaging. Since it does not require direct MIT imaging using the data (i.e., physical information) acquired by the MIT device, it can reduce the requirements for the accuracy and data volume of the MIT device. Because the amount of data required to train the MIT imaging model is reduced under the same image resolution requirements, the upper limit of the MIT image resolution can be increased, thereby improving the resolution of the MIT image. Simultaneously, by using the two-dimensional image data converted from the one-dimensional sequence data acquired by the MIT device for preliminary imaging, the spatiotemporal information of the data can be preserved, thereby further improving the preliminary imaging quality and thus increasing the resolution of the MIT image. Attached Figure Description

[0036] Figure 1 This is a schematic flowchart of the data processing method in an embodiment of this application;

[0037] Figure 2 This is a schematic diagram of the structure of an MIT device as described in an embodiment of this application;

[0038] Figure 3 This is a schematic diagram of a pix2pix network architecture in an embodiment of this application;

[0039] Figure 4 This is a schematic diagram of the data processing device structure according to an embodiment of this application;

[0040] Figure 5 This is a schematic diagram of the electronic device structure according to an embodiment of this application. Detailed Implementation

[0041] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0042] MIT (Metal-Induced Tomography) technology is a novel magnetically excited non-contact imaging technique that improves upon Electrical Impedance Tomography (EIT). Both MIT and EIT aim to reconstruct the relative distribution of conductivity of a target object. However, in terms of hardware design—specifically, the physical information acquisition section—MIT uses non-contact coils to generate excitation, while EIT applies excitation through electrodes in direct contact with the target object. MIT technology offers advantages such as non-contact operation, real-time monitoring, and low cost, making it widely applicable in fields like industrial non-destructive testing and geological exploration, and also showing great promise in clinical medical imaging.

[0043] When using MIT technology for imaging, non-contact excitation coils in the MIT device generate a magnetic field to excite the area under test. Detection coils detect physical changes (i.e., physical information) within the area, such as voltage changes, and calculate the imaging result based on these changes. This method of acquiring physical information makes MIT imaging results highly sensitive to both electrical conductivity and magnetic permeability. Furthermore, the excitation signal can penetrate objects with low conductivity effectively. Therefore, when using MIT technology for imaging, it can effectively detect the conductivity distribution within the target object. For example, in medical imaging applications, the excitation signal can penetrate the skull, which has low conductivity, to detect tumors, cerebral hematomas, and cerebral edema. Moreover, compared to EIT technology, which requires direct contact, MIT technology is applicable to a wider range of scenarios, especially those unsuitable for direct contact. For instance, in medical imaging applications, when the area to be imaged has skin allergies or injuries, non-contact MIT technology is clearly more suitable.

[0044] As mentioned above, in the process of imaging using MIT technology, the physical information collected by MIT devices needs to be calculated through a series of algorithms. Among the related technologies, the imaging algorithms in MIT technology mainly include imaging algorithms based on forward problems and imaging algorithms based on inverse problems.

[0045] Among them, the imaging algorithm based on the forward problem utilizes the simulation study of the forward problem in MIT technology and derives the inverse problem through mathematical calculation. It can be understood as an algorithm that combines the theoretical knowledge of electromagnetic fields with the physical information collected by MIT equipment to perform imaging. It belongs to the traditional mathematical iterative algorithm, such as the linear back projection algorithm, the conjugate gradient method, the sensitivity matrix algorithm, the Newton-Raphson algorithm, etc.

[0046] When using a forward problem-based imaging algorithm, the magnetic field distribution inside and around the target conductor within the measured area is calculated based on known information such as the geometric structure, internal conductivity distribution, and excitation signal applied in the excitation coil. Other parameters are then derived from this magnetic field distribution. This process requires continuous correction of the forward problem in the MIT (Made in Taiwan) technique. This can be understood as comparing the derived internal magnetic field distribution with the mathematical calculation results to correct the mathematical calculations. This correction process often requires computer simulation. Since simulation introduces noise, and different physical devices exhibit varying levels of noise due to differences in manufacturing processes and other interference factors, the simulation-based algorithm cannot be directly applied to physical devices. It requires continuous adjustment and optimization based on the actual conditions of the physical device to match its characteristics. Furthermore, the mathematical derivation method suffers from nonlinearity and uncertainty, making this imaging method unsuitable for situations with too many parameters. It also easily falls into local minimization traps when seeking the optimal solution, leading to low resolution or partial distortion in the imaging results.

[0047] As for the imaging algorithm based on the inverse problem, this algorithm uses emerging artificial intelligence technology to directly study the MIT imaging problem. It can be understood as using the boundary induced magnetic field signal or induced voltage data obtained by the detection coil, as well as the excitation signal applied to the excitation coil, to reconstruct the distribution or change of the conductivity inside the target object.

[0048] Imaging algorithms based on inverse problems mainly include the following two types:

[0049] The first approach involves image optimization of the imaging results from traditional iterative methods. Specifically, images generated using traditional mathematical iterative algorithms (such as linear backprojection and Newton-Raphson algorithms) are used as pre-imaging results. These pre-imaging results are then integrated with deep learning algorithms. This means using deep learning networks (such as Fully Convolutional Networks (FCNs), D-bar dressing methods, U-Net, and Generative Adversarial Networks (GANs)) to optimize the pre-imaging results. In other words, the pre-imaging results are used as input to the deep learning network, which then outputs the final imaging result. However, when using this method, the imaging results are limited by the quality of the pre-imaging generated by traditional mathematical iterative methods. It cannot handle special scenarios where the pre-imaging effect is poor, such as a significant offset between the target object's position in the pre-imaging result and its actual position, thus affecting the final imaging effect.

[0050] The second approach involves end-to-end imaging. Specifically, physical information collected by MIT devices is directly fed into deep learning networks (e.g., autoencoders, backpropagation neural networks, radial basis function (RBF) neural networks) for imaging. While this method eliminates reliance on traditional mathematical iterative methods, it places high demands on the accuracy of the MIT devices and the amount of data required. As the resolution requirements of the MIT imaging model increase, the amount of training data needed to train the model also increases. Therefore, when high resolution is required for MIT images, a large amount of training data is needed to train the MIT imaging model. This can easily lead to the model being unable to fit the data due to excessive data requirements or the algorithm reaching its bottleneck, thus limiting further improvements in the resolution of the MIT images generated by the MIT imaging model.

[0051] Based on this, in various embodiments of this application, since it is not necessary to directly perform MIT imaging using physical information collected by the MIT device, the requirements for the accuracy and data volume of the MIT device can be reduced. Since the amount of data required to train the MIT imaging model is reduced under the same image resolution requirements, the upper limit of MIT image resolution can be increased, thereby improving the resolution of MIT images. At the same time, by using the two-dimensional image data converted from the one-dimensional sequence data collected by the MIT device for preliminary imaging, the spatiotemporal information of the data can be preserved, thereby further improving the preliminary imaging quality and thus improving the resolution of MIT images.

[0052] This application provides a data processing method applied to electronic devices, such as... Figure 1 As shown, the method includes:

[0053] Step 101: Collect first data using MIT equipment, the first data containing one-dimensional sequence data;

[0054] Step 102: Using the GAF algorithm, the first data is transformed to obtain the second data, which includes two-dimensional image data;

[0055] Step 103: Using the second data and the first model, generate the first image;

[0056] Step 104: Generate the MIT image using the first image and the second model.

[0057] In practical applications, the electronic device can be a device capable of imaging using multiple electrodes to excite signals, such as an MIT device. Accordingly, when the electronic device is an MIT device, the acquisition of the first data using the MIT device can be understood as the direct acquisition of the first data.

[0058] In step 101, the MIT device can be understood as the hardware component in MIT technology used to collect physical information; for example, it can be employed as follows: Figure 2 The illustrated MIT device has 16 electrodes and dual channels. The dual channels can be understood as using two electrodes for transmission, i.e., generating excitation, and two electrodes for reception, i.e., collecting physical information. Of course, a single-channel MIT device can also be used, i.e., using one electrode for transmission and one electrode for reception. This application embodiment does not limit this.

[0059] In practical applications, the first data can be understood as the physical quantity of the measured object field (i.e., the collected physical information) collected by the MIT device when the target object is present in the measured object field; specifically, the first data can be the voltage signal collected by the MIT device; wherein, the measured object field can be understood as the object field corresponding to the measured area, and the physical quantity of the measured object field can be understood as the physical quantity within the measured area.

[0060] In practical applications, the use of MIT equipment to collect the first data means designing a data collection scheme and collecting data for the applicable scenarios of MIT technology. This can be understood as follows: since simulation calculation methods cannot be directly applied to physical devices with noise, in application scenarios such as medical imaging, the physical quantities actually collected by MIT equipment can be used for MIT imaging.

[0061] For example, using Figure 2 The MIT device shown acquires the voltage signal of the measured object field. The 16 electrodes in the MIT device are numbered sequentially from 0 to 15. The excitation transmission rule is as follows: data is transmitted from electrode 0, and electrodes 0-15 receive the voltage signal in sequence. Then, data is transmitted from electrode 1, and the voltage signal is received by electrodes 0-15 in sequence, and so on, until all 16 electrodes have completed their transmissions. Each excitation transmission by the MIT device is recorded as one frame. Therefore, after all 16 electrodes have completed their transmissions, frames u1 to u... will be generated. 256 A total of 256 voltage data points constitute the voltage data set u, u = {u1, u2, ..., u...} 256}

[0062] In step 102, after the MIT device completes the acquisition of physical information, it needs to preprocess the acquired physical information, that is, to use the GAF algorithm to transform the first data. Specifically, it transforms the one-dimensional sequence data into two-dimensional image data. For example, the GASF (Gramian AngularSummation) package in Python is used to transform the data. Figure 2 The 256 one-dimensional sequence data collected by the MIT device are converted into an image with a size of 256×256. Since the converted data is image data, it can acquire spatial attributes. At the same time, since the physical quantities collected by the MIT device have the characteristics of sequence data, that is, arranged in chronological order, and the GAF algorithm is an algorithm that can convert time series data into images, it can perceive the sorting order of the input one-dimensional sequence data. Therefore, the converted data can also retain the temporal information of the one-dimensional sequence data. Compared with the preprocessing methods in related technologies that only include noise reduction, this embodiment of the application explores the correlation between spatiotemporal information, so that the preprocessed data (i.e., the second data) retains spatiotemporal information, thereby improving the pre-imaging effect when the second data is input into the first model for pre-imaging.

[0063] In practical applications, the physical quantities collected by the MIT device are affected by environmental noise. Comparing the physical quantities collected by the MIT device with reference data can offset some of the noise interference, thereby reducing the impact of environmental noise on the physical quantities.

[0064] Based on this, in one embodiment, the specific implementation of step 102 may include:

[0065] The first data is compared with the benchmark data to obtain the comparison results.

[0066] Accordingly, the first data is transformed using the GAF algorithm to obtain the second data, which may include:

[0067] The comparison results are transformed using the GAF algorithm to obtain the second data.

[0068] The reference data can also be called reference data or reference frame; this application does not limit the specific terminology used, as long as the function can be achieved. In practical applications, the physical quantities within the measured object field when the target object is not present can be used as the reference data.

[0069] In practical applications, the reference data can be obtained directly by the MIT device.

[0070] Based on this, in one embodiment, the method may further include:

[0071] In the absence of a target object within the measured object field, physical quantities of the measured object field are collected to obtain third data;

[0072] The third data is used as the baseline data.

[0073] In practical applications, the benchmark data can be collected before comparing the first data with the benchmark data. For example, the benchmark data can be collected when the MIT device is powered on.

[0074] Of course, the benchmark data can also be calculated based on the configuration information of the MIT device, and this application embodiment does not limit this.

[0075] In practical applications, the first data is compared with the reference data. Specifically, this may include comparing the real and imaginary parts of each frame of data in the first data with the real and imaginary parts of the corresponding frame in the reference data. The comparison results constitute a preprocessed data set, namely the second data.

[0076] For example, the comparison result for each frame can be calculated using the following formula:

[0077]

[0078] Where z represents the comparison result, a′ represents the real part of the comparison frame data (i.e., the first data), b′ represents the imaginary part of the comparison frame data, a represents the real part of the background frame data (i.e., the reference data), and b represents the imaginary part of the background frame data; [the comparison will be] utilized Figure 2 After comparing all 256 frames of data collected by the MIT device shown, a preprocessed data set Z is obtained, Z = {z1, z2, z3, ..., z...}. 256}

[0079] In step 103, generating the first image using the second data and the first model can be understood as performing pre-imaging using pre-processed data. In practical applications, a residual network (e.g., a ResNet50 network) can be used for pre-imaging, meaning the first model can include a residual network, i.e., introducing a residual network for pre-imaging. In order to make the residual network suitable for MIT reconstruction, i.e., to make the residual network output continuous values, the activation function and loss function at the end of the residual network can be replaced. Specifically, the activation function of the residual network is replaced with a Sigmoid activation function, and the loss function of the residual network is replaced with a binary cross-entropy function.

[0080] Based on this, in one embodiment, generating the first image using the second data and the first model includes:

[0081] The first image is generated using the second data and the residual network; wherein,

[0082] The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction result of the first model and the actual result.

[0083] In practical applications, the first image can also be called a pre-image or a pre-image result. This application does not limit this, as long as it can achieve its function.

[0084] In practical applications, the first model can be trained and generated based on the physical quantities of samples collected by the MIT device.

[0085] Based on this, in one embodiment, the method may further include:

[0086] The first model is trained using the sample data.

[0087] Specifically, samples with different shapes and structures can be made and placed into the test field of the MIT device. The MIT device is used to collect physical quantities of the test field of the placed sample. The collected physical quantities are then used as sample data and input into the first model for training.

[0088] In order to simulate the differences in electrical conductivity of different objects, such as the differences in conductivity caused by different lesions in the human body in medical imaging applications, the samples can be made using solutions with several different conductivity. In addition, in order to enable the samples to better reproduce the edge shapes of different objects, samples of different shapes can be made by means of 3D printing, thereby enhancing the ability of the first model to distinguish the shape of the target object (e.g., the location of the lesion in medical imaging applications).

[0089] In practical applications, during the training of the first model, the measured object field can be uniformly divided into triangles, i.e., divided into multiple triangular regions. Using the position coordinates of different triangular regions and the actual resistance values ​​of the triangular regions, the relative conductivity distribution within the measured object field is constructed, and finally the conductivity set R is obtained. R is used as the label data when training the residual network.

[0090] Based on this, in one embodiment, the method may further include:

[0091] The imaging region of the measured object field is divided into triangles to obtain multiple imaging sub-regions;

[0092] The conductivity of each imaging sub-region in multiple imaging sub-regions is determined to obtain the third data.

[0093] The first model is trained using the conductivity of multiple imaging sub-regions.

[0094] For example, 512 identical triangles are used to divide the area, and the conductivity σ of each triangular region is calculated based on the resistance information of the samples. i For i∈{1,2,...,512}, based on the conductivity of each triangular region, we obtain the conductivity set R, R={σ1,σ2,...,σ 512}

[0095] In practical applications, during the training of the first model, the penalty coefficient can be increased when the judgment of the presence of samples in the test field is incorrect. This allows the first model to focus more on the judgment of the scene where the target object exists, that is, to focus on detecting the area where the target object exists. For example, in medical imaging applications, more attention is paid to the presence of lesion locations.

[0096] Based on this, in one embodiment, the method may further include:

[0097] When the expected output of the sample indicates that the first model has made an incorrect judgment, the weight coefficient of the binary cross-entropy loss function is a first value; when the expected output of the sample indicates that the first model has made a correct judgment, the weight coefficient of the binary cross-entropy loss function is a second value, and the first value is greater than the second value.

[0098] For example, the binary cross-entropy of the first model can be calculated using the following formula:

[0099]

[0100] Where, loss WBCE This represents the binary cross-entropy of the first model, where m and n represent the number of samples in the data batch and the number of classes, respectively. ji Indicates the expected output. This represents the actual output, and w represents the weighting coefficient (i.e., the first value).

[0101] When the expected output is 1, meaning the first model has made an incorrect judgment, the loss function of the first model is:

[0102]

[0103] When the expected output is 0, meaning the first model's judgment is correct, the loss function of the first model is:

[0104]

[0105] Because the weight coefficients are increased, when the network training process makes a mistake in judging whether there is an object in the tested area, the penalty coefficient will be increased, thereby improving the imaging effect of the first model on scenes with target objects.

[0106] In practical applications, to avoid overfitting of the first model during training, which would result in high prediction accuracy for training data but poor prediction accuracy for untrained data, recall rate can be used to control the fitting direction of the first model.

[0107] Based on this, in one embodiment, the method may further include:

[0108] Obtain first information, which represents the recall rate of samples during the training of the first model;

[0109] The first model is trained using the third data and the first information.

[0110] In practical applications, since recall reflects the proportion of correctly identified positive examples out of the total number of positive examples, the recall of the training results can be compared when training the first model. Training can be stopped when the optimal value is reached, thereby preventing the first model from overfitting and thus preventing insufficient effective information from being input into the second model due to overfitting.

[0111] Here, the pre-image generated by the first model, i.e. the first image, can already reflect the basic features of the target object such as its position and area. However, it still has certain deficiencies in restoring the edge information of the target object and its anti-interference ability. Therefore, the first image can be optimized by the second model.

[0112] In step 104, in practical applications, GAN (e.g., pix2pix network) can be used to optimize the first image. The process of generating an MIT image using the first image and the second model is to use an adversarial network to optimize the imaging results of the residual network.

[0113] Based on this, in one embodiment, generating an MIT image using the first image and the second model includes:

[0114] The MIT image is generated using the first image and GAN.

[0115] The GAN is used to optimize the first image.

[0116] In practical applications, the MIT image can be understood as the final result of MIT imaging, that is, the image of the relative distribution change of the conductivity of the target object.

[0117] The GAN is used to optimize the first image, which can be understood as using GAN to optimize the edge information and anti-interference ability of the first image.

[0118] In practical applications, the second model can be trained using historical preprocessed images generated by the first model.

[0119] Based on this, in one embodiment, the method may further include:

[0120] The second model is trained using the sample data.

[0121] For example, the historical preprocessed images (i.e., sample data) generated by the first model for the samples are input. Figure 3 In the pix2pix network shown, the generator of the pix2pix network is input, and the actual sample conductivity distribution map is used as the training label to train the pix2pix network. Unlike the circular image labels pieced together from triangles used by residual networks, the training labels of GANs are high-resolution images directly generated from the sample positions within the measured object field. Therefore, the imaging results are not limited by the number of triangles dividing the measured object field, thus compensating for the shortcomings of direct imaging in residual networks during secondary imaging. That is, it optimizes the deficiencies in target object edge information restoration and anti-interference ability that exist in the pre-imaging of residual networks.

[0122] The data processing method provided in this application embodiment collects first data using a MIT device, the first data comprising one-dimensional sequence data; the first data is transformed using the GAF algorithm to obtain second data, the second data comprising two-dimensional image data; a first image is generated using the second data and a first model; and a MIT image is generated using the first image and the second model. The solution provided in this application embodiment converts the data collected by the MIT device into image data, and then uses the image data for pre-imaging. Since it does not require direct MIT imaging using the data collected by the MIT device (i.e., physical information), it can reduce the requirements for the accuracy and data volume of the MIT device. Because it reduces the amount of data required to train the MIT imaging model under the same image resolution requirements, it can increase the upper limit of MIT image resolution, thereby improving the resolution of the MIT image. Simultaneously, by using the two-dimensional image data converted from the one-dimensional sequence data collected by the MIT device for preliminary imaging, the spatiotemporal information of the data can be preserved, thereby further improving the preliminary imaging quality and thus increasing the resolution of the MIT image.

[0123] Based on the above method, this application also provides a data processing device, disposed on an electronic device, such as... Figure 4 As shown, the device includes:

[0124] Acquisition unit 401 is used to acquire first data using MIT equipment, the first data including one-dimensional sequence data;

[0125] Processing unit 402 is configured to use the GAF algorithm to transform the first data to obtain second data, the second data including two-dimensional image data; use the second data and the first model to generate a first image; and use the first image and the second model to generate an MIT image.

[0126] In one embodiment, the processing unit 402 is configured to:

[0127] The first data is compared with the baseline data to obtain the comparison result;

[0128] The comparison results are transformed using the GAF algorithm to obtain the second data.

[0129] In one embodiment, the acquisition unit 401 is further configured to:

[0130] In the absence of a target object within the measured object field, physical quantities of the measured object field are collected to obtain third data;

[0131] The third data is used as the baseline data.

[0132] In one embodiment, the processing unit 402 is used for:

[0133] The first image is generated using the second data and the residual network; wherein,

[0134] The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction result of the first model and the actual result.

[0135] In one embodiment, the processing unit 402 is configured to:

[0136] The MIT image is generated using the first image and the adversarial network; wherein,

[0137] The adversarial network is used to optimize the first image.

[0138] In one embodiment, the processing unit 402 is further configured to:

[0139] The imaging region of the measured object field is divided into triangles to obtain multiple imaging sub-regions;

[0140] The conductivity of each imaging sub-region in multiple imaging sub-regions is determined to obtain the third data.

[0141] The first model is trained using the conductivity of multiple imaging sub-regions.

[0142] In one embodiment, the processing unit 402 is further configured to:

[0143] Obtain first information, which represents the recall rate of samples during the training of the first model;

[0144] The first model is trained using the third data and the first information.

[0145] In practical applications, the acquisition unit 401 can be implemented by the communication interface in the data processing device, and the processing unit 402 can be implemented by the processor in the data processing device in combination with the communication interface.

[0146] It should be noted that the data processing apparatus provided in the above embodiments is only illustrated by the division of the above program modules. In practical applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the apparatus can be divided into different program modules to complete all or part of the processing described above. In addition, the data processing apparatus and data processing method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0147] Based on the hardware implementation of the above program modules, and in order to implement the method of the embodiments of this application, the embodiments of this application also provide an electronic device, such as... Figure 5 As shown, the electronic device 500 includes:

[0148] Communication interface 501 enables information exchange with MIT devices, such as using MIT devices to collect first data;

[0149] The processor 502, connected to the communication interface 501, can interact with the MIT device and execute the methods provided by one or more of the above-mentioned technical solutions when running a computer program;

[0150] Memory 503, on which the computer program is stored.

[0151] Specifically, the communication interface 501 is used to collect first data using the MIT device, the first data comprising one-dimensional sequence data;

[0152] The processor 502 is configured to use the GAF algorithm to transform the first data to obtain second data, the second data including two-dimensional image data; generate a first image using the second data and a first model; and generate an MIT image using the first image and the second model.

[0153] In one embodiment, the processor 502 is configured to:

[0154] The first data is compared with the baseline data to obtain the comparison result;

[0155] The comparison results are transformed using the GAF algorithm to obtain the second data.

[0156] In one embodiment, the processor 502 is further configured to:

[0157] When there is no target object in the measured object field, the physical quantities of the measured object field are collected using the communication interface 501 to obtain third data;

[0158] The third data is used as the baseline data.

[0159] In one embodiment, the processor 502 is configured to:

[0160] The first image is generated using the second data and the residual network; wherein,

[0161] The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction result of the first model and the actual result.

[0162] In one embodiment, the processor 502 is configured to:

[0163] The MIT image is generated using the first image and the adversarial network; wherein,

[0164] The adversarial network is used to optimize the first image.

[0165] In one embodiment, the processor 502 is further configured to:

[0166] The imaging region of the measured object field is divided into triangles to obtain multiple imaging sub-regions;

[0167] The conductivity of each imaging sub-region in multiple imaging sub-regions is determined to obtain the third data.

[0168] The first model is trained using the conductivity of multiple imaging sub-regions.

[0169] In one embodiment, the processor 502 is further configured to:

[0170] Obtain first information, which represents the recall rate of samples during the training of the first model;

[0171] The first model is trained using the third data and the first information.

[0172] It should be noted that the specific processing procedures of processor 502 and communication interface 501 can be understood by referring to the above method.

[0173] Of course, in practical applications, the various components in electronic device 500 are coupled together through bus system 504. It can be understood that bus system 504 is used to realize the connection and communication between these components. In addition to a data bus, bus system 504 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in... Figure 5 The general designated all buses as Bus System 504.

[0174] The memory 503 in this embodiment is used to store various types of data to support the operation of the electronic device 500. Examples of such data include any computer program used to operate on the electronic device 500.

[0175] The methods disclosed in the embodiments of this application can be applied to the processor 502, or implemented by the processor 502. The processor 502 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 502 or by instructions in the form of software. The processor 502 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 502 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in the memory 503. The processor 502 reads the information in the memory 503 and combines its hardware to complete the steps of the aforementioned method.

[0176] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0177] It is understood that the memory (memory 503) in this embodiment of the application can be volatile memory or non-volatile memory, or both. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); the magnetic surface memory can be disk storage or magnetic tape storage. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable types of memories.

[0178] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 503 storing a computer program, which can be executed by the processor 502 of the electronic device 500 to complete the steps described in the aforementioned method. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0179] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0180] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0181] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A data processing method, characterized in that, include: First data was acquired using the MIT magnetic induction tomography device, and the first data contained one-dimensional sequence data. The first data is transformed using the Gram angle field (GAF) algorithm to obtain the second data, which includes two-dimensional image data. Using the second data and the first model, a first image is generated; Using the first image and the second model, generate an MIT image; The process of using the GAF algorithm to transform the first data to obtain the second data includes: The first data is compared with the baseline data to obtain the comparison result; The comparison results are transformed using the GAF algorithm to obtain the second data; The step of generating the first image using the second data and the first model includes: The first image is generated using the second data and the residual network; wherein, The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction results of the first model and the actual results. The step of generating an MIT image using the first image and the second model includes: The MIT image is generated using the first image and the adversarial network.

2. The method according to claim 1, characterized in that, The method further includes: In the absence of a target object within the measured object field, physical quantities of the measured object field are collected to obtain third data; The third data is used as the baseline data.

3. The method according to claim 2, characterized in that, The method further includes: The imaging region of the measured object field is divided into triangles to obtain multiple imaging sub-regions; The conductivity of each imaging sub-region in multiple imaging sub-regions is determined to obtain the third data; The first model is trained using the conductivity of multiple imaging sub-regions.

4. The method according to claim 3, characterized in that, The method further includes: Obtain first information, which represents the recall rate of samples during the training of the first model; The first model is trained using the third data and the first information.

5. A data processing apparatus, characterized in that, include: A data acquisition unit is used to acquire first data using MIT equipment, the first data comprising one-dimensional sequence data; The processing unit is configured to use the GAF algorithm to transform the first data to obtain second data, the second data including two-dimensional image data; use the second data and the first model to generate a first image; and use the first image and the second model to generate an MIT image. The processing unit is further configured to compare the first data with the benchmark data to obtain a comparison result; The comparison results are transformed using the GAF algorithm to obtain the second data; The processing unit is further configured to generate the first image using the second data and the residual network; wherein... The Sigmoid activation function in the residual network is used for delinearization, and the binary cross-entropy loss function in the residual network is used to evaluate the degree of difference between the prediction results of the first model and the actual results. And the MIT image is generated using the first image and the adversarial network.

6. An electronic device, characterized in that, include: The processor and the memory used to store computer programs that can run on the processor. When the processor is used to run the computer program, it performs the steps of the method according to any one of claims 1 to 4.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images