Microwave imaging detection method and device based on fusion algorithm, electronic equipment and medium

Abstract

The invention relates to the technical field of brain imaging, in particular to a microwave imaging detection method and device based on a fusion algorithm, electronic equipment and a medium, and the method comprises the steps: S1, collecting actual microwave scattering electric field distribution; s2, inputting the acquired actual microwave scattering electric field distribution into a Born approximation algorithm model, and performing inversion to obtain a first brain foreign body initial imaging result; s3, inputting the first brain foreign body initial imaging result into a deep learning neural network model to obtain first microwave scattering electric field distribution; s4, inputting the first microwave scattering electric field distribution into the Born approximation algorithm model again to continue iterative calculation to obtain a second brain foreign body initial imaging result, and inputting the second brain foreign body initial imaging result into the deep learning neural network model again to obtain second microwave scattering electric field distribution; s5, the microwave scattering electric field distribution is continuously iteratively updated in the step S4 until the difference value between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is smaller than a preset threshold value, and at the moment, the brain foreign body initial imaging result corresponding to the updated microwave scattering electric field distribution is a high-resolution brain foreign body imaging result; and S6, further judging whether the brain foreign body exists or not according to the high-resolution brain foreign body imaging result. According to the method, the microwave signal features are automatically extracted by using the deep learning model, and fine physical modeling is performed through the Born iteration method, so that the accuracy and reliability of detection are remarkably improved.

Description

Technical Field

[0001] The present invention relates to brain imaging technology, in particular to a microwave imaging detection method, device, electronic device and medium based on a fusion algorithm, mainly used for the precise detection of cerebral hemorrhage, cerebral infarction, brain tumors and other intracranial foreign bodies. Background Art

[0002] Existing brain imaging technologies mainly rely on CT (Computed Tomography) and MRI (Magnetic Resonance Imaging). Although these methods can provide high-resolution images, they have problems such as expensive equipment, large volume, long imaging time and CT radiation, which limit their application in the emergency scene. Microwave imaging has become a research hotspot due to its non-radiation, low cost and portability. However, traditional microwave imaging methods rely on complex scattering matrix calculations and use the Born approximation for linearization processing, with high computational complexity and limited imaging accuracy for foreign bodies.

[0003] The Born iterative method is a numerical method for solving electromagnetic scattering problems. By iteratively calculating the scattered field, it gradually approaches the true scattered field distribution. However, the Born iterative method may have problems of poor convergence and large computational amount when dealing with targets with high contrast and complex structures. In recent years, deep learning has achieved remarkable results in the fields of image processing and signal analysis. Combining deep learning with the Born iterative method is expected to overcome their respective limitations and achieve efficient and precise detection of brain foreign bodies. Summary of the Invention

[0004] In view of the above situation, embodiments of the present application propose a microwave imaging detection method, device and medium based on a fusion algorithm. This method uses a deep learning neural network model to automatically extract microwave signal features and performs fine physical modeling through the Born iterative method, significantly improving the accuracy and reliability of detection.

[0005] In a first aspect, the present application discloses a microwave imaging detection method based on a fusion algorithm, the method comprising: S1 Collect the actual microwave scattered electric field distribution; S2 Construct a Born approximation iterative algorithm model. First, input the collected actual microwave scattered electric field distribution into the Born approximation algorithm model to inversely obtain the initial imaging result of the first brain foreign body; S3 Establish a pre-trained deep learning neural network model, and then input the initial imaging result of the first brain foreign body into the deep learning neural network model to obtain the first microwave scattered electric field distribution; S4 Input the first microwave scattered electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain the initial imaging result of the second brain foreign body, and input the initial imaging result of the second brain foreign body into the deep learning neural network model again to obtain the second microwave scattered electric field distribution; Step S5 continuously iterates and updates the microwave scattering electric field distribution until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and then stops. At this time, the initial imaging result of the brain foreign object corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; S6 further determines whether there is a brain foreign object according to the high-resolution brain foreign object imaging result.

[0006] In one embodiment, the deep learning neural network model has an input layer, a hidden layer, and an output layer. The hidden layer has a dynamic convolution kernel group. The dynamic convolution kernel group has multiple dynamic convolutions and a gating mechanism. The dynamic convolution algorithm relies on the dynamic convolution kernel to implement. The dynamic convolution algorithm adjusts the scale and direction of the dynamic convolution kernel in real time through the gating mechanism.

[0007] In one embodiment, the hidden layer of the deep learning neural network model also has a multi-branch feature adaptive fusion module. The multi-branch feature adaptive fusion module uses different convolutions to extract the features of the same feature map. The multi-branch feature adaptive fusion module includes a parallel shallow layer with 3×3 small kernels and a parallel deep layer with 5×5 large kernels. The parallel shallow layer focuses on local high-frequency scattering details, and the parallel deep layer captures the global interference pattern.

[0008] In one embodiment, the Born approximation iterative algorithm model is established based on the Born approximation algorithm formula. The Born approximation algorithm formula: (Formula 1) is the microwave scattering electric field distribution of the (n - 1)th iteration; : Green's function, which describes the interaction between the scattered field and the incident field; : Polarization response function of the scattering object, which describes the response of the scattering object medium to electromagnetic waves; : Incident field, representing the electromagnetic field generated by an external source; k : Wave number, representing the propagation characteristics of electromagnetic waves; V: Integration domain, representing the region range considered in the calculation process; Based on the scattering formula of the Born approximation, the polarization response function of the scattering object is obtained by the 0th iteration inversion , as the initial imaging result of the brain foreign object, and each microwave scattering electric field has its corresponding polarization response function .

[0009] In one embodiment, inputting the high-resolution brain foreign object imaging result into the Softmax classification layer to determine whether there is a brain foreign object further includes: The Softmax classification layer converts the high-resolution brain foreign object imaging result into a probability distribution; The output of the Softmax layer includes two categories: the presence of a foreign object and the absence of a foreign object.

[0010] In one embodiment, the high-resolution brain foreign object imaging result is visualized using a heat map.

[0011] In one embodiment, it further includes step S11 of performing data preprocessing on the actual microwave scattering electric field distribution, and the preprocessing method includes normalization and noise suppression.

[0012] The second aspect of the present application also discloses a brain foreign object detection device based on a fusion algorithm, and the device includes: A data acquisition module for acquiring the actual microwave scattering electric field distribution; A Born approximation iterative algorithm model calculation module that first inputs the acquired actual microwave scattering electric field distribution into the Born approximation algorithm model to inversely obtain a first initial brain foreign object imaging result; A deep learning neural network model calculation module that then inputs the first initial brain foreign object imaging result into the deep learning neural network model to obtain a first microwave scattering electric field distribution; A fusion calculation module that inputs the first microwave scattering electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain a second initial brain foreign object imaging result, and then inputs the second initial brain foreign object imaging result into the deep learning neural network model to obtain a second microwave scattering electric field distribution; A condition judgment module that iterates continuously until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and stops. At this time, the initial imaging result corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; A result acquisition module that determines whether there is a brain foreign object based on the high-resolution brain foreign object imaging result. The third aspect of the present application also discloses an electronic device, including: A processor; and A memory arranged to store computer-executable instructions, and the executable instructions, when executed, cause the processor to execute the above detection method.

[0013] The fourth aspect of the present application also discloses a computer-readable storage medium, and the computer-readable storage medium stores one or more programs. When the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device is caused to execute the above detection method.

[0014] The above-mentioned at least one technical solution adopted in the embodiments of the present application can achieve the following beneficial effects: In the present application, by combining an advanced convolutional neural network (CNN) and the Born iterative algorithm, high-precision imaging and foreign object detection of brain microwave signals are realized. This method uses a deep learning neural network model to automatically extract the features of microwave signals and performs fine physical modeling through the Born iterative algorithm, significantly improving the accuracy and reliability of detection. This method has the advantages of non-invasiveness, low cost, and strong real-time performance, and can be widely applied to first aid sites, bedside detection in hospitals, and remote medical monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings: Figure 1 A logic diagram of a brain foreign object detection method according to a first embodiment of the present application is shown.

[0016] Figure 2 A schematic diagram of a brain foreign object detection device according to a second embodiment of the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0018] To solve the defects of the prior art, Figure 1 As shown, a microwave imaging detection method based on a fusion algorithm is provided in the first embodiment of the present application, including: S1: Collect the actual microwave scattering electric field distribution.

[0019] The actual microwave scattering electric field distribution includes a normal microwave scattering electric field distribution and a microwave scattering electric field distribution suspected of having a foreign object.

[0020] A plurality of microwave antennas are evenly distributed outside the head of the patient to transmit microwave signals and receive the reflected signals. N microwave antennas (a1, a2,..., aN) are evenly distributed outside the head of the patient. The microwave antennas can be dipole antennas. These dipole antennas have the same structure, and each antenna can receive and transmit microwave signals. When one dipole antenna transmits microwave signals, all antennas receive and calculate the corresponding scattering S parameters, that is, the actual microwave scattering electric field distribution.

[0021] In an embodiment of the present application, 16 dipole antennas are provided. Antennas a1, a2, …, a16 are evenly distributed on the circumference outside the brain imaging area of the brain. In order to reduce the reflection of microwave signals when entering the brain, a matching medium is filled between the antennas and the brain. There is a foreign object at a certain position in the brain. The relative permittivity and conductivity of the blood clot are 58 and 2.3 S / m respectively; the relative permittivity and conductivity of the brain are 42 and 0.9 S / m respectively, and the relative permittivity and conductivity of the matching medium are 40 and 0.9 S / m respectively. Preferably, the dipole antenna adopted has S(1,1) less than -10 dB in the frequency band of 0.58 GHz - 0.66 GHz, and S(1,1) refers to the input reflection coefficient (characterizing the microwave scattering electric field distribution), that is, the input return loss.

[0022] The measurement parameters include the position vectors a1, a2, …, aN of the measurement antennas, the frequency f of the electromagnetic wave emitted during measurement, the magnetic permeability μ of the brain, and the relative permittivity ε of the brain b and the conductivity σ of the brain b ; the S parameter obtained when a brain foreign object is suspected to exist is denoted as S tot (m,n), and the S parameter obtained when measuring a normal brain is denoted as S inc (m,n).

[0023] Step S1 further includes step S11 of performing data preprocessing on the microwave scattering electric field distribution, and the preprocessing methods include normalization and noise suppression.

[0024] S2. Construct a Born approximation iterative algorithm model. First, input the collected actual microwave scattering electric field distribution into the Born approximation algorithm model to inversely obtain the initial imaging result of the first brain foreign object; S3. Establish a pre-trained deep learning neural network model, and then input the initial imaging result of the first brain foreign object into the deep learning neural network model to obtain the first microwave scattering electric field distribution; The initial imaging result of the brain foreign object in the present application is essentially the dielectric constant distribution of the imaging area obtained by iteratively solving the scattering field through step S1. Input this initial imaging result into the deep learning neural network model to obtain the microwave scattering electric field distribution corresponding to the initial imaging result.

[0025] The present application helps to improve the imaging accuracy of strong scatterers based on the pre-trained deep learning neural network model on the basis of the initial imaging result.

[0026] The deep learning neural network model of the present application is established based on the convolutional neural network model (CNNs); this deep learning neural network model has an input layer, a hidden layer, and an output layer, and the hidden layer has a dynamic convolution kernel group and a multi-branch feature adaptive fusion module.

[0027] A group of dynamic convolution kernels. A group of dynamic convolution kernels has multiple dynamic convolutions, while a conventional convolution has only one static convolution kernel and is independent of the input samples. For dynamic convolution, it linearly weights multiple convolution kernels, and the weighting values are related to the input, which makes dynamic convolution have input dependence.

[0028] The dynamic convolution algorithm depends on the implementation of dynamic convolution kernels. The dynamic convolution algorithm (ODConv) dynamically adjusts the shape and size of the convolution kernel according to the characteristics of the input data during the convolution process to adapt to different input data. The dynamic convolution algorithm adjusts the scale and direction of the convolution kernel in real time through a gating mechanism (Gating Network) so that it can adaptively match the optimal feature extraction mode for different dielectric shapes (such as sharp edges and curved surface structures).

[0029] The gating mechanism (Gating Network) is a learnable deformation module introduced by ODConv, which dynamically adjusts the shape and size of the convolution kernel according to the characteristics of the input data, thereby improving the performance of the convolutional neural network.

[0030] The input of the multi-branch feature adaptive fusion module comes from the output of the group of dynamic convolution kernels. The multi-branch feature adaptive fusion module adopts multi-scale feature fusion. This process involves fusing deep features and shallow features. The adaptive fusion module has parallel shallow (3×3 small kernels) and parallel deep (5×5 large kernels) convolution paths. The parallel shallow layer (referred to as the shallow branch) focuses on local high-frequency scattering details (such as corner diffraction), and the parallel deep layer (referred to as the deep branch) captures global interference patterns (such as multiple reflections at the dielectric interface). The parallel multi-branch network uses different convolutions to extract the features of the same feature map, and then uses concatenation to fuse the extracted features to obtain a multi-scale fusion feature map.

[0031] Iterative-driven parameter update: The network adopts a two-stage parameter update strategy during the Born iteration process - freezing the backbone network weights in the initial stage to stabilize feature extraction, and opening some layers to participate in end-to-end optimization in the later stage of iteration to achieve the co-evolution of the feature model and the physical inversion process.

[0032] The training process of the deep learning neural network model in this application is the same as that of the prior art, and both include data input, forward propagation, calculation of loss, backpropagation, update of weights, and output of results.

[0033] S4, input the first microwave scattering electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain the initial imaging result of the second brain foreign object, and then input the initial imaging result of the second brain foreign object into the deep learning neural network model to obtain the second microwave scattering electric field distribution; S5. In the above step S4, the microwave scattering electric field distribution is continuously iteratively updated until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and then stops. At this time, the initial imaging result of the brain foreign object corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; S4 - S5 is the fusion step of the deep learning neural network model and the Born approximation iterative algorithm model. The output of the deep learning neural network is used as the input of the Born approximation iterative algorithm model, and the output of the Born approximation iterative algorithm model is used as the input of the deep neural network, and it is continuously iterated. Repeating this way until, during the iteration process, when the change amount of the electromagnetic scattering field in two consecutive iterations is less than a certain preset threshold, the iteration process stops. Specifically, the convergence condition requires that the change difference of the scattering field between two consecutive iterations be small enough, meaning that the model has fully learned the accurate scattering field distribution. The setting of this threshold is determined according to actual needs, and the most appropriate value can be determined through experiments. The judgment of the convergence condition is crucial because it determines whether the algorithm stops iterating after obtaining a stable result, avoiding unnecessary computational burden.

[0034] The Born approximation algorithm model mentioned in this application is established based on the Born approximation algorithm formula. The Born approximation assumes that the scattering effect of the scatterer is weak and the incident field does not change significantly inside the scatterer. Based on this assumption, a Born approximation iterative algorithm model is established. The Born approximation algorithm formula is: (Formula 1) When the number of iterations n is greater than or equal to 1, Formula 1 is the formula for the scattering electric field intensity distribution of the nth iteration based on the Born approximation.

[0035] When the number of iterations n = 0, Formula 1 is the initial scattering electric field intensity distribution formula of the Born approximation.

[0036] The initial scattering electric field intensity distribution formula of the Born approximation is: (Formula 2) Explanation of formula parameters: is the actually collected microwave scattering electric field distribution; n is the number of iterations, is the microwave scattering electric field distribution of the (n - 1)th iteration; : Green's function, which describes the interaction between the scattering field and the incident field.

[0037] : Polarization response function of the scattering object, which describes the response of the scattering object medium to electromagnetic waves.

[0038] : Incident field, representing the electromagnetic field generated by an external source.

[0039] k : Wave number, representing the propagation characteristics of electromagnetic waves.

[0040] V: Integration domain, representing the range of the region considered in the calculation process.

[0041] In Equation 2, the above , , k、 V, are known quantities, is the value to be solved by the scattering formula of the Born approximation.

[0042] Based on the scattering formula of the Born approximation, the polarization response function of the scattering object is inversely obtained , as the initial imaging result of brain foreign objects. Each microwave scattered electric field has its corresponding polarization response function . By continuously iteratively updating the microwave scattered electric field distribution, the imaging accuracy for strong scatterers is improved.

[0043] In the embodiment of the present application, for the above formula parameters, the scatterer in the polarization response function of the scattering object is the head of the patient. The Green's function describes the interaction between the scattered field and the incident field, where the scattered field is the electric field scattered by the head of the patient, and the incident field is the electric field incident on the head of the patient, representing the electromagnetic field generated by an external source. The external source is specifically a microwave signal source connected to the antenna, and the microwave signal source outputs a microwave signal to the antenna.

[0044] Initially, the actually measured microwave scattered electric field distribution is input into the initial scattered electric field intensity distribution formula (Equation 2) based on the Born approximation to obtain the first initial imaging result of brain foreign objects. At this time, the accuracy of the first initial imaging result of brain foreign objects is not high. The first initial imaging result of brain foreign objects with low accuracy is input into the deep learning neural network model. The deep learning neural network model outputs the updated microwave scattered electric field distribution, that is, the first microwave scattered electric field distribution, and then the updated microwave scattered electric field distribution is input into Equation 1 to inversely obtain the second initial imaging result of brain foreign objects.

[0045] The initial imaging result of brain foreign objects obtained from each iteration is input into the deep neural network CNN to further extract high-level features. And the high-level features are used for feature analysis and image reconstruction, so as to output the initial conditions for updating the next iteration, realizing the fusion of deep learning and the Born iteration method. Using the updated polarization response function as the updated imaging result, continue to iteratively calculate the scattered field.

[0046] The fusion algorithm of this application refers to the fusion of deep learning and the Born iterative method. Through multiple iterations of the Born approximation algorithm and the fusion with the deep learning neural network model, the accuracy of the initial imaging results of brain foreign objects is continuously improved.

[0047] S6. Further determine whether there is a brain foreign object based on the high-resolution imaging results of brain foreign objects.

[0048] Through multiple iterations and the fusion with the deep learning model, the high-resolution imaging results of brain foreign objects are finally output. The high-resolution imaging results of brain foreign objects are input into the Softmax classification layer to determine whether there is a brain foreign object in the image.

[0049] The Softmax classification layer uses the Softmax function to convert the output of the deep learning neural network model into a probability distribution, which can effectively handle multi-class classification problems.

[0050] In this application, the output of the Softmax layer contains two categories: the presence of a foreign object and the absence of a foreign object. By comparing the probability values of these two categories, it can be determined whether there is a brain foreign object in the image.

[0051] To further locate the position of the foreign object, a heatmap can be used for visualization, which is widely used in the field of deep learning. The heatmap represents the importance or activation degree of each region in the image through the depth of color. In the detection of brain foreign objects, the heatmap can show the degree of attention of the model to different regions, thus helping doctors or researchers visually identify the possible positions of foreign objects. This visualization method not only improves the transparency of detection but also enhances the interpretability of the model.

[0052] This application uses the Softmax classification layer to determine whether there is a brain foreign object and visualizes the position of the foreign object through a heatmap.

[0053] As Figure 2 shown, the second embodiment of this application provides a brain foreign object detection device based on a fusion algorithm. The device 200 includes: A data acquisition module 210 for acquiring the actual microwave scattering electric field distribution; A Born approximation iterative algorithm model calculation module 220. First, input the acquired actual microwave scattering electric field distribution into the Born approximation algorithm model to inversely obtain the initial imaging results of the first brain foreign object; A deep learning neural network model calculation module 230. Then, input the initial imaging results of the first brain foreign object into the deep learning neural network model to obtain the first microwave scattering electric field distribution; The fusion calculation module 240 inputs the first microwave scattering electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain the initial imaging result of the second brain foreign object, and then inputs the initial imaging result of the second brain foreign object into the deep learning neural network model to obtain the second microwave scattering electric field distribution; The condition judgment module 250 iterates continuously until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and then stops. At this time, the initial imaging result corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; The result acquisition module 260 determines whether there is a brain foreign object according to the high-resolution brain foreign object imaging result.

[0054] The third embodiment of the present application provides an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, and the executable instructions, when executed, cause the processor to execute the above-mentioned brain foreign object detection method.

[0055] The fourth embodiment of the present application discloses a computer-readable storage medium, and the computer-readable storage medium stores one or more programs. When the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device is caused to execute the above-mentioned brain foreign object detection method.

[0056] The above are only the embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A microwave imaging detection method based on a fusion algorithm, characterized in that The method includes: S1 Collect the actual microwave scattering electric field distribution; S2 Construct a Born approximation iterative algorithm model. First, input the collected actual microwave scattering electric field distribution into the Born approximation algorithm model to inversely obtain the initial imaging result of the first brain foreign object; S3 Establish a pre-trained deep learning neural network model, and then input the initial imaging result of the first brain foreign object into the deep learning neural network model to obtain the first microwave scattering electric field distribution; S4 Input the first microwave scattering electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain the initial imaging result of the second brain foreign object. Then input the initial imaging result of the second brain foreign object into the deep learning neural network model to obtain the second microwave scattering electric field distribution; S5 Continuously iterate and update the microwave scattering electric field distribution in the above step S4 until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and stop. At this time, the initial imaging result of the brain foreign object corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; S6 Further judge whether there is a brain foreign object according to the high-resolution brain foreign object imaging result.

2. The brain foreign object detection method according to claim 1, wherein, The deep learning neural network model has an input layer, a hidden layer, and an output layer. The hidden layer has a dynamic convolution kernel group. The dynamic convolution kernel group has multiple dynamic convolutions and a gating mechanism using a dynamic convolution algorithm. The dynamic convolution algorithm relies on a dynamic convolution kernel to implement, and the scale and direction of the dynamic convolution kernel are adjusted in real time through the gating mechanism.

3. The brain foreign object detection method according to claim 2, wherein The hidden layer of the deep learning neural network model also has a multi-branch feature adaptive fusion module. The multi-branch feature adaptive fusion module uses different convolution kernels to extract the features of the same feature map. The convolution kernels used by the multi-branch feature adaptive fusion module include multiple parallel shallow layers of 3×3 small kernels and multiple parallel deep layers of 5×5 large kernels. The parallel shallow layers focus on local high-frequency scattering details, and the parallel deep layers capture global interference patterns.

4. The brain foreign object detection method according to claim 1, wherein, The Born approximation iterative algorithm model is established based on the Born approximation algorithm formula. The Born approximation algorithm formula: (Formula 1) is the microwave scattering electric field distribution for the (n - 1)-th iteration; : The Green's function, which describes the interaction between the scattered field and the incident field; : Polarization response function of the scattering object, which describes the response of the scattering object to electromagnetic waves; : Incident field, representing the electromagnetic field generated by an external source; k : Wave number, representing the propagation characteristics of electromagnetic waves; V: Integration domain, representing the range of the region considered in the calculation process; Based on the scattering formula of the Born approximation, the polarization response function of the scattering object is obtained by the 0th iteration inversion , as the initial imaging result of brain foreign objects, each microwave scattering electric field has its corresponding polarization response function .

5. The brain foreign object detection method according to claim 1, characterized in that, Further including inputting the high-resolution brain foreign object imaging result into the Softmax classification layer to judge whether there is a brain foreign object: The Softmax classification layer converts the high-resolution brain foreign object imaging result into a probability distribution; The output of the Softmax layer includes two categories: presence of foreign object and absence of foreign object.

6. The brain foreign object detection method according to claim 1, characterized in that Visualize the high-resolution brain foreign object imaging result with a heat map.

7. The brain foreign object detection method according to claim 1, wherein, It also includes step S11 of performing data preprocessing on the actual microwave scattering electric field distribution. The data preprocessing method includes normalization and noise suppression.

8. A brain foreign object detection device based on a fusion algorithm, characterized in that, The device includes: A data acquisition module for collecting the actual microwave scattering electric field distribution; A Born approximation iterative algorithm model calculation module that first inputs the collected actual microwave scattering electric field distribution into the Born approximation algorithm model to inversely obtain the initial imaging result of the first brain foreign object; A deep learning neural network model calculation module that then inputs the initial imaging result of the first brain foreign object into the deep learning neural network model to obtain the first microwave scattering electric field distribution; The fusion calculation module inputs the first microwave scattering electric field distribution into the Born approximation algorithm model again for iterative calculation to obtain the initial imaging result of the second brain foreign object, and then inputs the initial imaging result of the second brain foreign object into the deep learning neural network model to obtain the second microwave scattering electric field distribution; The condition judgment module iterates continuously until the difference between the updated microwave scattering electric field distribution and the actual microwave scattering electric field distribution is less than a preset threshold and stops. At this time, the initial imaging result corresponding to the updated microwave scattering electric field distribution is the high-resolution brain foreign object imaging result; The result acquisition module determines whether there is a brain foreign object according to the high-resolution brain foreign object imaging result.

9. An electronic device, comprising: a processor; and a memory arranged to store computer-executable instructions, which when executed cause the processor to execute the method according to claims 1 to 7.

10. A computer-readable storage medium storing one or more programs, which when executed by an electronic device including a plurality of application programs, cause the electronic device to execute the method according to claims 1 to 7.

Images