Method of determining characteristic attributes of a predetermined target, system therefor, intracranial hemorrhage detection apparatus, and computer storage medium

Abstract

Embodiments of the present application relate to the field of measurement technology for diagnostic purposes, and particularly to a method for determining a characteristic attribute of a predetermined target, a system thereof, an intracranial hemorrhage detection device, and a computer storage medium. The method of the embodiments of the present application receives a microwave signal of a predetermined target, filters out noise of the microwave signal, then performs normalization processing thereon, extracts characteristic information of the predetermined target in the normalized microwave signal, reconstructs an image of the predetermined target according to the characteristic information of the predetermined target, determines the characteristic attribute of the predetermined target by combining the image with a category, uses category semantics to guide accurate positioning of a target region in the image, uses spatial information of the image to supplement spatial quantitative information such as spatial position, shape, and energy intensity that is lacking in category discrimination, and the two complement each other to help improve the detection sensitivity of the tiny target of the hemorrhagic region, and help improve the integrity and accuracy of the determined target region.

Description

Technical Field

[0001] Embodiments of this application relate to the field of measurement technology for diagnostic purposes, specifically to a method and system for determining characteristic attributes of a predetermined target, an intracranial hemorrhage detection device, and a computer storage medium. Background Technology

[0002] The statements herein are provided merely as background information in connection with this application and do not necessarily constitute prior art.

[0003] Intracranial hemorrhage is a common acute and critical illness in clinical practice. During the diagnosis process, it is necessary to clarify the condition and provide treatment as soon as possible to avoid delaying the "golden treatment time." Therefore, it is necessary to design reliable intracranial hemorrhage detection methods to accurately assess the patient's condition.

[0004] Currently, methods for detecting intracranial hemorrhage typically involve first acquiring microwave signals from the patient's head, and then using these signals to determine the extent of the hemorrhage. However, existing methods for detecting intracranial hemorrhage still have many limitations. Summary of the Invention

[0005] A brief overview of this application is provided below to offer a basic understanding of certain aspects thereof. It should be understood that this overview is not an exhaustive summary of the application. It is not intended to identify key or essential parts of the application, nor is it intended to limit its scope. Its purpose is merely to present certain concepts in a simplified form as a prelude to the more detailed description that follows.

[0006] In a first aspect, embodiments of this application provide a method for determining the characteristic attributes of a predetermined target, comprising the following steps: S10: receiving a microwave signal of the predetermined target; S20: filtering out noise from the microwave signal and normalizing the noise-filtered microwave signal; S30: extracting characteristic information of the predetermined target from the microwave signal obtained in step S20; S40: reconstructing an image of the predetermined target based on the characteristic information of the predetermined target; S50: determining the category of the microwave signal of the predetermined target based on the characteristic information of the predetermined target; S60: determining the characteristic attributes of the target region of the predetermined target based on the image and the category.

[0007] The method provided in the embodiments of this application receives microwave signals of a predetermined target and filters out noise from the microwave signals, then normalizes them, and then extracts the feature information of the predetermined target from the normalized microwave signals. Based on the feature information of the predetermined target, an image of the predetermined target is reconstructed. In this way, the feature information of the predetermined target can be determined through the reconstructed image, thereby determining the category of the microwave signal of the predetermined target. The reconstructed image can provide spatial quantitative information of the predetermined target, such as spatial location, morphology, and energy intensity, but lacks pathological semantics, and lesions are easily confused with background clutter. The category can provide pathological semantic information on whether the predetermined target is bleeding, but it is only a qualitative judgment and lacks spatial and quantitative description. By jointly determining the feature attributes of the predetermined target by combining the image and the category, the category semantics are used to guide the precise positioning of the target area in the image, and the spatial information of the image is used to supplement the spatial quantitative information such as spatial location, morphology, and energy intensity that are lacking in the category discrimination. The two complement and corroborate each other, which helps to improve the detection sensitivity of the bleeding area, a small target, and at the same time helps to improve the completeness and accuracy of the determined target area.

[0008] Secondly, embodiments of this application also provide a system for determining the characteristic attributes of a predetermined target, comprising: a microwave signal receiving module configured to receive microwave signals of the predetermined target; a microwave signal processing module configured to filter out noise from the microwave signals and normalize the noise-filtered microwave signals; a feature information extraction module configured to extract feature information of the predetermined target from the microwave signals obtained by the microwave signal processing module; an image reconstruction module configured to reconstruct an image of the predetermined target based on the feature information of the predetermined target; a category determination module configured to determine the category of the microwave signals of the predetermined target based on the feature information of the predetermined target; and a feature attribute determination module configured to determine the feature attributes of the target area of ​​the predetermined target based on the image and the category.

[0009] The system provided in the embodiments of this application receives microwave signals from a predetermined target and filters out noise from the microwave signals. Then, it normalizes the microwave signals and extracts the feature information of the predetermined target from the normalized microwave signals. Based on the feature information of the predetermined target, an image of the predetermined target is reconstructed. In this way, the feature information of the predetermined target can be determined through the reconstructed image, thereby determining the category of the microwave signal of the predetermined target. The reconstructed image can provide spatial quantitative information of the predetermined target, such as spatial location, morphology, and energy intensity, but it lacks pathological semantics, and lesions are easily confused with background clutter. The category can provide pathological semantic information on whether the predetermined target is bleeding, but it is only a qualitative judgment and lacks spatial and quantitative description. By jointly determining the feature attributes of the predetermined target by combining the image and the category, the category semantics are used to guide the precise positioning of the target area in the image, and the spatial information of the image is used to supplement the spatial quantitative information such as spatial location, morphology, and energy intensity that are lacking in the category discrimination. The two complement and corroborate each other, which helps to improve the detection sensitivity of the bleeding area, a small target, and at the same time helps to improve the completeness and accuracy of the determined target area.

[0010] Thirdly, embodiments of this application also provide an intracranial hemorrhage detection device, comprising: a receiving portion forming a receiving space for receiving a patient's brain; a base, the receiving portion disposed on the base; a flexible headgear disposed on the receiving space, configured to be used to fill a liquid dielectric and deformable under pressure; the base is further configured to discharge the leaked liquid when the liquid dielectric leaks; a signal transmission and acquisition component for transmitting microwave signals to and receiving microwave signals from the flexible headgear; and a processor configured to be communicatively connected to the signal transmission and acquisition component, for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to execute the method of determining the characteristic attributes of a predetermined target according to embodiments of this application.

[0011] The intracranial hemorrhage detection device provided in this application, due to the flexible headgear, can adapt to different patient cranial shapes by filling the flexible headgear with a dielectric material and applying pressure to it to allow it to expand and conform to the patient's cranium. Relying on the flexible deformation and fluid adaptive properties of the headgear, it closely conforms to the patient's cranial contour, achieving a gapless fit and wrapping, thereby improving the transmission efficiency of microwave signals and thus improving the quality of transmitted or received microwave signals. By setting the base to allow leakage of liquid dielectric material, the leaked liquid can be discharged outwards, thus preventing mold growth or damage to the signal transmission and acquisition components. The device uses a processor to receive the microwave signals from the cranium, filter out noise, normalize the signals, and extract the cranial feature information from the normalized microwave signals. Based on this feature information, an image of the cranium is reconstructed. This reconstructed image allows for the determination of cranial feature information and thus cranial feature attributes. Therefore, based on the image and category, the sensitivity and quantification of cranial feature attribute prediction are more comprehensive and accurate.

[0012] Fourthly, embodiments of this application also provide an intracranial hemorrhage detection device, comprising: a receiving portion forming a receiving space for receiving a patient's brain; a flexible headgear disposed in the receiving space, configured to be used for filling with a liquid dielectric and deformable under pressure; a signal transmission and acquisition component for transmitting microwave signals to and receiving microwave signals from the flexible headgear; a dielectric filling and discharging pressurization component configured to fill the flexible headgear with a dielectric or discharge the dielectric from the flexible headgear, and to pressurize the flexible headgear so that it can expand to fit the patient's brain; and a processor configured to be communicatively connected to the signal transmission and acquisition component, for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to execute the method of determining the characteristic attributes of a predetermined target according to embodiments of this application.

[0013] Fifthly, embodiments of this application also provide a computer storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the method for determining the characteristic attributes of a predetermined target as described in embodiments of this application.

[0014] These and other advantages of this application will become more apparent from the following detailed description of preferred embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0015] To further illustrate the above and other advantages and features of this application, the specific embodiments of this application will be described in more detail below with reference to the accompanying drawings. The drawings, together with the following detailed description, are included in and form a part of this specification. Elements having the same function and structure are indicated by the same reference numerals. It should be understood that these drawings only depict typical examples of this application and should not be considered as limiting the scope of this application.

[0016] Figure 1 A schematic diagram is shown of the spatial feature vector corresponding to an image determined by the method according to an embodiment of this application; Figure 2 A schematic diagram showing the distribution of the cosine similarity of semantic vectors and spatial feature vectors determined by the method according to an embodiment of this application at various spatial locations in the imaging region is illustrated. Figure 3 A schematic diagram of the target region of an image determined by the method according to an embodiment of this application is shown; Figure 4 An image of a predetermined target is shown based on the microwave signal before noise removal; Figure 5 An image of a predetermined target obtained by filtering out noise from a microwave signal based on the method provided in the embodiments of this application is shown. Figure 6 A schematic diagram showing microwave signals corresponding to targets at different depths determined by the method according to embodiments of this application is illustrated. Figure 7 A distribution map of dielectric differences among different tissues in the frequency domain feature space, determined by the method according to an embodiment of this application, is shown. Figure 8 A comparative schematic diagram is shown showing the spatial focusing response of a microwave signal subjected to beamforming processing according to the method of this application, and the direct superposition response of a microwave signal without beamforming processing. Figure 9 A schematic diagram of the original time-domain signal of interference clutter in the matching medium between the unstripped antenna array and the predetermined target is shown. Figure 10 This diagram illustrates the target time-domain signal after removing interference clutter from the matching medium between the antenna array and the predetermined target. Figure 11 A schematic diagram of the imaging results obtained without removing interference clutter from the matching medium in the time domain is shown, as well as a magnified schematic diagram of the target region in the imaging results; Figure 12 A schematic diagram of the imaging results obtained after removing interference clutter from the matching medium in the time domain is shown, as well as a magnified schematic diagram of the target region in the imaging results; Figure 13A schematic diagram of the imaging results obtained without utilizing the standardized grid space is shown; Figure 14 A schematic diagram of the imaging results obtained using a standardized grid space is shown; Figure 15 A schematic diagram showing the result of independently encoding the feature information of a predetermined target according to an embodiment of this application is provided. Figure 16 A schematic diagram showing the result of single encoding of feature information of a predetermined target according to an embodiment of this application is provided. Figure 17 This diagram illustrates the result of adversarial training of the encoding process according to the method provided in the embodiments of this application; Figure 18 A schematic diagram of clutter region features and bleeding region features obtained from the encoding process after adversarial training is shown; Figure 19 A schematic diagram of an intracranial hemorrhage detection device according to an embodiment of this application is shown, wherein the housing and media container are omitted. Figure 20 This shows from another perspective Figure 19 The intracranial hemorrhage detection device shown; Figure 21 It shows Figure 19 The diagram shows a schematic of an intracranial hemorrhage detection device, with some components such as the processor omitted. Figure 22 An exploded view of an intracranial hemorrhage detection device according to an embodiment of this application is shown; Figure 23 A schematic diagram of the structure of the receiving portion of the intracranial hemorrhage detection device according to an embodiment of this application is shown; Figure 24 This shows from another perspective Figure 23 A schematic diagram of the structure of the receiving part is shown.

[0017] It should be noted that the accompanying drawings are not necessarily drawn to scale, but are shown only in a schematic manner without affecting the reader's understanding.

[0018] Explanation of reference numerals in the attached figures: 10. Antenna; 151. Wire connector; 191. Vector network analyzer; 1910. Analyzer heat dissipation hole; 1911. Side view; 192. Port extension; 193. Signal transmission line; 1931. First horizontal extension; 1932. Bent connection section; 20. Receiving part; 21. Peripheral part; 211. Plane; 2110. Opening; 212. Peripheral mounting hole; 22. End; 23. Antenna mounting part; 230. Mounting groove; 231. Extension part; 232. Connector; 2320. Clearance hole; 24. Head cover mounting part; 240. Groove; 241. Guide groove; 242. Inner liner positioning part; 250. Connecting plate; 251. First connecting post; 252. Second connecting post; 253. Reinforcing rib; 26. Neck support part; 27. Housing connector; 271. Housing clearance groove; 28. Liquid inlet / outlet positioning part; 30. Flexible head cover; 301. Head cover body; 31. Liquid inlet / outlet connector; 32. Vent connector; 33. First head cover mounting component; 331. Guide component; 332. Head cover mounting hole; 34. Second head cover mounting component; 40. Dielectric charging / discharging and pressurizing assembly; 41. Dielectric container; 42. Drive unit; 43. Inlet / outlet pipeline; 44. Exhaust pipeline; 46. Pressure measuring device; 47. Valve; 48. Pressure measuring pipeline; 50. Lining; 51. Lining positioning and mating part; 52. Lining clearance groove; 61. Base; 611. Bottom cover; 6110. First drain hole; 612. Second drain hole; 613. Heat dissipation hole; 621. First mounting component; 622. Second mounting component; 623. Third mounting component; 624. Auxiliary mounting component; 6240. Mounting plate; 6241. Mounting feet; 625. Fourth mounting component; 626. Reinforcing component; 641. Shell heat dissipation hole; 642. Main shell; 6421. Main shell heat dissipation hole; 643. Front shell; 644. Rear cover; 645. Pillow; 646. Handle; 65. Processor; 652. USB splitter; 66. Fan; 67. Operating component; 68. Communication interface. Detailed Implementation

[0019] Exemplary embodiments of this application will be described below with reference to the accompanying drawings. For clarity and brevity, not all features of actual implementations are described in the specification. However, it should be understood that many implementation-specific decisions must be made in the development of any such actual embodiment to achieve the developer's specific goals, such as complying with constraints related to the system and business, and these constraints may vary depending on the implementation. Furthermore, it should be understood that while development work can be very complex and time-consuming, such development work is merely a routine task for those skilled in the art who benefit from the content of this application.

[0020] It should also be noted that, in order to avoid obscuring this application with unnecessary details, only the equipment structure and / or processing steps closely related to the solution according to this application are shown in the accompanying drawings, while other details that are not closely related to this application are omitted.

[0021] It should be noted that, unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning as understood by a person with ordinary skills in the field to which this application pertains.

[0022] In the description of the embodiments of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0023] Currently, intracranial hemorrhage detection methods typically begin by acquiring microwave signals from the patient's head, and then use these signals to determine the extent of intracranial hemorrhage. However, the inventors of this application have discovered that existing intracranial hemorrhage detection methods, when using microwave signals to detect intracranial hemorrhage, require determining the hemorrhage status of the brain based on the acquired cranial microwave signals. However, existing methods for determining the hemorrhage status of the brain based on microwave signals produce results with poor accuracy and reliability, making it difficult to accurately, comprehensively, and quantitatively reflect the true extent of intracranial hemorrhage.

[0024] To address the aforementioned technical problems, embodiments of this application provide a method for determining the characteristic attributes of a predetermined target, the method comprising at least the following steps S10 to S60.

[0025] In some embodiments, S10: receiving microwave signals of a predetermined target; S20: filtering out noise from the microwave signals and normalizing the noise-filtered microwave signals; S30: extracting feature information of the predetermined target from the microwave signals obtained in step S20; S40: reconstructing an image of the predetermined target based on the feature information of the predetermined target; S50: determining the category of the microwave signals of the predetermined target based on the feature information of the predetermined target; S60: determining the feature attributes of the bleeding area of ​​the predetermined target based on the image and the category.

[0026] The method provided in the embodiments of this application receives microwave signals of a predetermined target and filters out noise from the microwave signals, then normalizes them, and then extracts the feature information of the predetermined target from the normalized microwave signals. Based on the feature information of the predetermined target, an image of the predetermined target is reconstructed. In this way, the feature information of the predetermined target can be determined through the reconstructed image, thereby determining the category of the microwave signal of the predetermined target. The reconstructed image can provide spatial quantitative information of the predetermined target, such as spatial location, morphology, and energy intensity, but lacks pathological semantics, and lesions are easily confused with background clutter. The category can provide pathological semantic information on whether the predetermined target is bleeding, but it is only a qualitative judgment and lacks spatial and quantitative description. By jointly determining the feature attributes of the predetermined target by combining the image and the category, the category semantics are used to guide the precise positioning of the target area in the image, and the spatial information of the image is used to supplement the spatial quantitative information such as spatial location, morphology, and energy intensity that are lacking in the category discrimination. The two complement and corroborate each other, which helps to improve the detection sensitivity of the bleeding area, a small target, and at the same time helps to improve the completeness and accuracy of the determined target area.

[0027] In some embodiments, the predetermined target may be, for example, the brain, and the target area may be, for example, the bleeding area of ​​the brain.

[0028] In some embodiments, in step S10, the microwave signal carries feature information of a predetermined target.

[0029] In some embodiments, in step S30, the feature information of the predetermined target may be, for example, a multi-dimensional intermediate feature representation that characterizes the electromagnetic scattering properties of the predetermined target tissue extracted from the preprocessed microwave signal.

[0030] In some embodiments, in step S50, the category of the microwave signal of the predetermined target can be, for example, a qualitative judgment result based on feature information regarding whether the predetermined target is bleeding (and the nature of the bleeding). For example, the category can be a semantic label / category marker used to indicate health or bleeding.

[0031] In some embodiments, in step S60, the characteristic attributes of the target area of ​​the predetermined target can be, for example, a complete and quantifiable attribute description representing "bleeding location, bleeding morphology, and bleeding energy intensity".

[0032] In some embodiments, step S60 specifically includes the following steps: S61: determining the semantic vector of the category according to the category; S62: determining the target region features of the image according to the semantic vector determined in step S61 and the image; S63: determining the feature attributes of the target region of the predetermined target according to the semantic vector and the target region features.

[0033] The method provided in the embodiments of this application, compared with the method of determining the feature attributes of the target region of a predetermined target based solely on image or text categories, uses semantic vectors and target region features of images to determine the feature attributes of the target region of a predetermined target. This is beneficial to achieving a balance between spatial accuracy and semantic interpretability, thereby improving the reliability of the determined feature attributes of the target region of the predetermined target.

[0034] In some embodiments, step S61 specifically includes the following steps: S611: Determine the text label of the category based on the category; S612: Determine the semantic vector of fixed dimension based on the text label; S613: Align the semantic vector of fixed dimension in feature space to determine the semantic vector of the category.

[0035] The method provided in the embodiments of this application aligns the semantic vectors of a fixed dimension with the feature space, so that the semantic vectors and the spatial feature vectors of the image are in the same space, which is beneficial for providing conditions for subsequent determination of feature similarity distribution.

[0036] In some embodiments, in step S612, text labels can be input into a fixed encoder to determine a fixed-dimensional semantic vector.

[0037] In some embodiments, Figure 1 A schematic diagram is shown showing the spatial feature vector corresponding to an image determined by the method according to an embodiment of this application. Figure 2 The diagram illustrates the distribution of cosine similarity between semantic vectors and spatial feature vectors determined by the method according to embodiments of this application at various spatial locations within the imaging region. Figure 3 A schematic diagram of the target region of an image determined by the method according to an embodiment of this application is shown. Figure 1 In the diagram, the vertical axis on the right represents the magnitude of the spatial feature vector. For example... Figures 1 to 3 As shown, step S62 specifically includes the following steps: S621: Determine the spatial feature vector corresponding to the image based on the image; S622: Determine the cosine similarity between the semantic vector and the spatial feature vector based on the semantic vector and the image; S623: Determine the feature similarity distribution between the semantic vector and the spatial feature vector based on the cosine similarity; S624: Determine the matching degree between the semantic vector and the spatial feature vector based on the feature similarity distribution; S625: Determine the target region and spatial location of the image based on the matching degree; S626: Determine the target region features of the image based on the target region and spatial location.

[0038] The method provided in the embodiments of this application, by determining the cosine similarity between semantic vectors and spatial feature vectors, can more clearly express the high-dimensional features of microwave signals, thereby facilitating the accurate determination of the feature attributes and spatial location of the target area of ​​a predetermined target.

[0039] In some embodiments, when the predetermined target is the brain, the target region of the image represents the hemorrhage area in the brain image, and the spatial location represents the spatial location of the hemorrhage area in the image.

[0040] In some embodiments, such as Figures 1 to 3 As shown, when determining the target region, regions with a matching degree value greater than or equal to the matching degree threshold can be identified as target regions. It is understood that those skilled in the art are aware of the possible values ​​for the matching degree threshold, and this application does not impose any restrictions on this.

[0041] In some embodiments, step S63 specifically includes the following steps: S631: Based on the semantic vector and the target region features, the semantic vector and the target region features are concatenated along the channel dimension to determine the joint feature vector of the predetermined target; S632: Based on the joint feature vector, the spatial location attribute, morphological attribute and energy intensity attribute of the predetermined target are determined; S633: Based on the semantic vector and the attributes determined in step S632, the feature attributes of the target region of the predetermined target are determined.

[0042] The method provided in the embodiments of this application can determine the joint feature vector of a predetermined target by concatenating the semantic vector with the target region features along the channel dimension. This can simultaneously fuse category semantic information and image spatial information, thereby facilitating a more accurate determination of the feature attributes of the target region of the predetermined target.

[0043] In some embodiments, the spatial location attribute of the predetermined target may be, for example, the spatial location of the bleeding area of ​​the predetermined target, that is, the location of the bleeding area in the predetermined target.

[0044] In some embodiments, the morphological attributes of the predetermined target may be, for example, the shape, size, extent, etc. of the bleeding area of ​​the predetermined target.

[0045] In some embodiments, the energy intensity attribute of the predetermined target can be, for example, the intensity of the scattered energy of the bleeding area of ​​the predetermined target.

[0046] In some embodiments, step S20 specifically includes the following steps: S21: acquiring brain microwave signals of a preset number of healthy samples; S22: determining the noise-filtered microwave signal based on the microwave signals obtained in step S10 and the brain microwave signals of the healthy samples obtained in step S21; S23: determining the normalized microwave signal based on the noise-filtered microwave signal.

[0047] The method provided in the embodiments of this application acquires the brain microwave signal of a healthy sample and determines the noise-filtered microwave signal based on the microwave signal obtained in step S10 and the brain microwave signal of the healthy sample. This helps to filter out electromagnetic reflection clutter generated by layered structures such as the skull and scalp, thereby improving the signal-to-noise ratio of the microwave signal. By normalizing the noise-filtered microwave signal, it is beneficial to eliminate amplitude shifts caused by non-target factors and map the microwave scattering parameters to a uniform numerical range, thereby improving the generalization ability of the prediction.

[0048] In some embodiments, the signal transmission and acquisition components in the intracranial hemorrhage detection device can be used to pre-acquire brain microwave signals from a predetermined number of healthy samples.

[0049] In some embodiments, step S22 specifically includes the following steps: S221: Based on the brain microwave signal of the healthy sample obtained in step S21, for each transceiver channel of the transceiver antenna array, determine the time-domain mean of the brain microwave signal of the healthy sample, and determine the baseline mean signal corresponding to the transceiver channel based on the time-domain mean; S222: Based on the brain microwave signal of the healthy sample and the baseline mean signal, determine the centered healthy sample matrix; perform singular value decomposition on the centered healthy sample matrix, and determine the eigenvectors corresponding to the first N principal components, and determine the background clutter subspace corresponding to the transceiver channel based on the eigenvectors, where N is the preset number of principal components; S223: Based on the microwave signal obtained in step S10, convert the microwave signal to a time-domain signal, and determine the centered target microwave signal based on the baseline mean signal; S224: Based on the centered target microwave signal and the background clutter subspace, determine the estimated clutter component in the target microwave signal; based on the centered target microwave signal and the estimated clutter component, determine the microwave signal after noise filtering.

[0050] The method provided in the embodiments of this application, by performing singular value decomposition on the brain microwave signal of a healthy sample and determining the eigenvectors corresponding to the first N principal components, can determine the background clutter subspace that reflects the electromagnetic scattering patterns of static structures such as the skull and tissue boundaries. Based on the centered target microwave signal and the background clutter subspace, the estimated clutter components in the target microwave signal are determined, which is beneficial for accurately filtering out the dominant background clutter and avoiding misfiltering due to the overlap of background clutter and effective signal in time delay and frequency band. By establishing a dedicated background clutter subspace model for each transceiver channel, it helps to adapt to the differences in asymmetric scattering paths caused by different antenna spatial positions.

[0051] In some embodiments, different transceiver channels correspond to different antenna spatial locations, resulting in different electromagnetic wave propagation paths. The distance between the transmitting and receiving antennas, the thickness of the skull / tissue it passes through, and the anatomical structures it traverses are all asymmetrical. Therefore, by determining the background clutter subspace corresponding to each transceiver channel, the background clutter of each channel has its own distinct time delay, amplitude, and spatial-frequency structure, i.e., asymmetric scattering path differences. If all transceiver channels share a unified background clutter model, it is impossible to accurately match the different clutter structures of each channel, leading to a decrease in filtering accuracy. By establishing a dedicated baseline and background clutter subspace model for each transceiver channel, the clutter specific to that channel can be accurately matched and subtracted, thereby achieving precise filtering of clutter across the entire array.

[0052] In some embodiments, Figure 4 An image of a predetermined target is shown based on the microwave signal before noise filtering. Figure 5 An image of a predetermined target obtained from a microwave signal after noise removal based on the method provided in the embodiments of this application is shown, such as... Figure 4 As shown, due to interference and masking from background clutter, the characteristics of the microwave signal 80 of the hemorrhage target in the hemorrhage area before noise filtering are severely obscured, making effective identification impossible; while as Figure 5 As shown, after filtering out noise based on the method provided in this application embodiment, the clutter at the hemorrhage area is effectively suppressed, making the microwave signal 80 of the blood cell target in the hemorrhage area stand out and clearly present a high-brightness energy convergence point. This shows that the method in this application embodiment can significantly improve the signal-to-noise ratio of the microwave signal and achieve accurate localization and visualization of the hemorrhage location in the brain.

[0053] In some embodiments, the value of N can be in the range of 1-10, such as N being 1, 2, 3, 4, 5, 10, etc., so that a balance can be achieved between "sufficient clutter filtering" and "effective signal retention".

[0054] In some embodiments, if the value of N is too small (e.g., N=1), only the principal component with the strongest energy can be removed, and clutter such as skull / scalp cannot be fully characterized and filtered out, resulting in a lot of residual clutter and limited improvement in signal-to-noise ratio.

[0055] In some embodiments, if the value of N is too large (e.g., N>10), the background clutter subspace will exceed the boundary and include components that are actually valid signals, causing the valid bleeding signal to be removed along with the clutter during projection removal, i.e., the "false filtering" described in the previous embodiments, which will damage the lesion signal and reduce sensitivity.

[0056] In some embodiments, when N is in the range of 2-5, it can cover several dominant principal components corresponding to the main static structures such as the skull, scalp, and matching medium, and fully filter out the dominant background clutter, without intruding into the effective signal subspace, thereby achieving the best balance between "filtering out clutter" and "preserving the lesion signal".

[0057] In some embodiments, the specific value of N can be determined by the singular value energy ratio (cumulative energy threshold method), that is, selecting the minimum N value required so that the sum of the energies of the first N singular values ​​accounts for a preset ratio (e.g., 85%–95%) of the total energy, thereby adaptively covering the dominant clutter component without intruding into the signal subspace.

[0058] In some embodiments, in step S224, the target microwave signal can be projected onto the background clutter subspace to determine the estimated clutter components in the target microwave signal.

[0059] In some embodiments, step S23 specifically includes the following steps: S231: Determine the amplitude of the microwave signal based on the microwave signal after noise removal, and determine the maximum absolute value of the amplitude in the frequency dimension; S232: Determine the normalization denominator based on the maximum absolute value and a preset constant; S233: Determine the normalized microwave signal based on the microwave signal after noise removal and the normalization denominator.

[0060] The method provided in the embodiments of this application determines the maximum absolute value of the amplitude in the frequency dimension based on the amplitude of the microwave signal after noise filtering. Then, it determines the normalized denominator based on the maximum absolute value and a preset constant to determine the normalized microwave signal. This helps to strictly map the microwave signal to a preset amplitude range, unify the dimensions of each transmitting and receiving channel, and thus help to avoid strong channel signals drowning out weak channel signals. Determining the normalized microwave signal based on the microwave signal after noise filtering and the normalized denominator can preserve the relative amplitude ratio between the frequency points of the microwave signal, and thus can preserve the zero-point characteristics and phase structure of the microwave signal.

[0061] In some embodiments, zero-point features represent frequency points with zero amplitude, which remain zero after being multiplied by any scaling factor. Therefore, the location and features of zero points (amplitude dips / nulls) in microwave signals are not changed. These zero points often correspond to specific scattering cancellation and can serve as important pathological discrimination features.

[0062] In some embodiments, the phase structure indicates that the normalization only scales the amplitude in real terms (by a positive real denominator) and does not involve phase rotation of the complex signal. Therefore, the phase information at each frequency point is preserved intact, and the phase structure of the scattered signal is complete.

[0063] In such an embodiment, a single global maximum value is taken as the normalization denominator in the frequency dimension, and then the entire signal uses the same single scalar denominator for proportional scaling. Proportional scaling naturally keeps the ratio between elements unchanged, so the relative amplitude ratio of each frequency point, zero-point characteristics and phase structure can all be preserved.

[0064] In some embodiments, in step S232, the preset constant can be, for example, a very small positive number introduced to prevent the normalized denominator from being zero, such as... - Preferably, the preset constant can be This prevents numerical overflow caused by a zero denominator while avoiding interference with normal normalization results.

[0065] In some embodiments, step S30 specifically includes the following steps: S31: Determine the time-domain characteristics of the microwave signal based on the microwave signal obtained in step S20; S32: Determine the frequency-domain characteristics of the microwave signal based on the microwave signal obtained in step S20; S33: Determine the spatial-domain characteristics of the microwave signal based on the microwave signal obtained in step S20; S34: Determine the feature information of the predetermined target based on the time-domain characteristics, frequency-domain characteristics, and spatial-domain characteristics.

[0066] The method provided in the embodiments of this application can express the scattering characteristics of microwave signals in cranial tissue from multiple dimensions by extracting the time-domain features, frequency-domain features and spatial-domain features of microwave signals respectively, thereby more comprehensively reflecting the feature information of the target area of ​​the predetermined target.

[0067] In some embodiments, step S31 specifically includes the following steps: S311: Determine the echo delay of the microwave signal based on the microwave signal obtained in step S20; S312: Determine the pulse peak amplitude of the microwave signal based on the microwave signal obtained in step S20; S313: Determine the waveform symmetry of the microwave signal based on the microwave signal obtained in step S20; S314: Determine the time-domain characteristics of the microwave signal based on the echo delay, pulse peak amplitude, and waveform symmetry.

[0068] The method provided in the embodiments of this application can reflect the scattering response law of a predetermined target in the time dimension by determining the echo delay, pulse peak amplitude and waveform symmetry of the microwave signal, thereby improving the ability to express the position and intensity of the target area in the predetermined target.

[0069] Figure 6 A schematic diagram illustrating microwave signals corresponding to targets at different depths determined by the method according to embodiments of this application is shown, such as... Figure 6As shown in the figure, based on the three echo curves at different depths, it can be seen that as the depth increases, the time of the echo peak appears later. Specifically, at a depth of 10 mm, the corresponding echo peak appears at 8 ns; at a depth of 25 mm, it appears at 16 ns; and at a depth of 45 mm, it appears at 27 ns. There is a clear monotonic correspondence between depth and the time of the corresponding echo peak. In this embodiment, since the echo delay τ is proportional to the round-trip propagation distance of the electromagnetic wave in brain tissue, the depth location of the target area (hemorrhage area) in the brain can be deduced from the echo delay. Therefore, by extracting the echo delay as a temporal feature, the ability to express the spatial location of the hemorrhage area in the brain can be improved, and the delay information can be directly converted into lesion depth.

[0070] In some embodiments, step S32 specifically includes the following steps: S321: Perform a short-time Fourier transform on the microwave signal obtained in step S20 to determine the time-frequency distribution of the microwave signal; S322: Determine the power spectral density distribution of the microwave signal based on the time-frequency distribution; S323: Determine the frequency domain energy concentration of the microwave signal based on the time-frequency distribution; S324: Determine the frequency domain characteristics of the microwave signal based on the power spectral density distribution and the frequency domain energy concentration.

[0071] The method provided in the embodiments of this application performs a short-time Fourier transform on the microwave signal and determines the power spectral density distribution and frequency domain energy concentration of the microwave signal based on the time-frequency distribution of the microwave signal. This can reflect the scattering energy distribution law of the predetermined target in the frequency dimension, thereby improving the ability to express the differences in dielectric properties of the predetermined target.

[0072] Figure 7 The diagram illustrates the dielectric difference distribution of different tissues in the frequency domain feature space determined by the method according to embodiments of this application, such as... Figure 7 As shown, this feature space is composed of the characteristic frequencies corresponding to the power spectral density distribution determined by S322 and the frequency domain energy concentration determined by S323, which respectively show the sample distribution of hemorrhagic tissue, normal tissue, ischemic tissue, and edematous tissue. It can be seen that the samples of the four types of tissues are clustered together with clear boundaries between clusters, and arranged sequentially along a monotonic trend. Hemorrhagic tissue is located in the region with the lowest characteristic frequency and the highest frequency domain energy concentration, followed by edematous tissue and ischemic tissue, and normal tissue is located in the region with the highest characteristic frequency and the lowest frequency domain energy concentration.

[0073] The order of the four types of tissues is completely consistent with the order of their dielectric properties (dielectric constant and conductivity) from high to low. Tissues with higher dielectric properties have their scattered echo energy concentrated at lower frequencies, resulting in lower characteristic frequencies and higher frequency domain energy concentration. Therefore, the method provided in this application, by determining the power spectral density distribution and frequency domain energy concentration, can effectively express the differences in the dielectric properties of a predetermined target as separable positional differences in the frequency domain feature space. This significantly improves the ability to express the differences in the dielectric properties of the predetermined target, providing a clearly physical basis for subsequently determining the category of microwave signals. Dielectric property differences can be distinguished.

[0074] In some embodiments, step S33 specifically includes the following steps: S331: Determine the spatial distribution of the microwave signal in each receiving channel based on the microwave signal obtained in step S20; S332: Perform beamforming processing on the microwave signal based on the spatial distribution to determine the spatial focusing response of the microwave signal; S333: Determine the scattering intensity of the microwave signal in each spatial region of the predetermined target based on the spatial focusing response; S334: Determine the spatial characteristics of the microwave signal based on the spatial focusing response and the scattering intensity.

[0075] The method provided in the embodiments of this application performs beamforming processing on microwave signals based on the spatial distribution of microwave signals in each receiving channel and determines their scattering intensity in each spatial region of a predetermined target. This can reflect the scattering distribution law of the predetermined target in the spatial dimension, thereby improving the ability to locate the spatial position of the predetermined target in the target area.

[0076] In some embodiments, the spatial focusing response refers to the response value at each spatial location within the imaging area, obtained by applying corresponding time delay compensation to the microwave signals of each channel based on the antenna spatial location corresponding to each receiving channel and then coherently superimposing them; the response values ​​at all spatial locations together constitute the spatial focusing response.

[0077] In some embodiments, Figure 8 The diagram illustrates a comparison of the spatial focusing response of a microwave signal subjected to beamforming processing according to an embodiment of this application, and the direct superposition response of a microwave signal without beamforming processing. Figure 8 The horizontal axis represents the profile along the target region, and the vertical axis represents the scattered energy response intensity at each location in the imaging space, mapped to the 0-1 interval after normalization. For example... Figure 8As shown, the curve representing the spatial focusing response of the microwave signal after beamforming fully illustrates the distribution pattern of scattered energy along the cross-section of the target region, i.e., in the spatial dimension. However, the curve representing the direct superposition response of the microwave signal without beamforming cannot effectively reflect the distribution pattern of scattered energy in the spatial dimension due to energy dispersion and waveform roughness.

[0078] In some embodiments, step S40 specifically includes the following steps: S41: Based on the feature information obtained in step S30, remove interference clutter from the matching medium between the antenna array and the predetermined target to obtain a target time-domain feature signal containing only brain tissue; S42: Based on the spatial geometry of the antenna array and the average dielectric properties of the predetermined target, determine a multi-dimensional spatial grid coordinate matrix representing the spatial geometry and the average dielectric properties of the predetermined target; S43: Based on the target time-domain feature signal obtained in step S41 and the multi-dimensional spatial grid coordinate matrix determined in step S42, determine the transmission distance and delay time of each pair of transceiver antennas at each grid point; S44: Based on the transmission distance and delay time determined in step S43, perform multi-frequency multi-antenna confocal energy superposition and attenuation weighted compensation on the target time-domain feature signal obtained in step S41 at each grid point to obtain an image of the predetermined target.

[0079] The method provided in the embodiments of this application firstly strips the interface reflection clutter of the matching medium by time-domain truncation. Then, based on the target time-domain characteristic signal and the multi-dimensional spatial grid coordinate matrix, the transmission distance and delay time of each pair of transceiver antennas at each grid point are determined. This establishes a mapping relationship between the preset grid space and the antenna propagation path, thereby enabling rapid confocal superposition and geometric attenuation compensation of multi-frequency and multi-antenna signals. This significantly improves the image reconstruction speed in the actual detection process while greatly enhancing the spatial resolution of the target image, providing a geometric reference for subsequently determining the feature attributes of the target area of ​​the predetermined target.

[0080] In some embodiments, step S41 specifically includes the following steps: S411: Determine the thickness of the matching medium between the antenna transmitting and receiving microwave signals and the scalp of the predetermined target based on the scattering signal from the reflection port of the vector network analyzer, and obtain the dielectric constant of the matching medium; S412: Determine the transmission time of the electromagnetic wave through the matching medium based on the thickness and dielectric constant determined in step S411; S413: Convert the feature information obtained in step S30 from the frequency domain to the time domain using inverse Fourier transform to obtain a time-domain microwave signal; S414: Determine the parameter configuration of a preset window function based on the transmission time determined in step S412 to determine the target window function; S415: Combine the target window function determined in step S414 with the time-domain microwave signal determined in step S413 to determine a target time-domain feature signal that includes only brain tissue.

[0081] Since the matching medium is the region through which electromagnetic waves first pass in the time domain, the method provided in the embodiments of this application can greatly filter out and completely remove the influence of the matching medium on microwave signals in the time domain by determining the transmission time of electromagnetic waves through the matching medium. This makes the position of the target area displayed in the final generated image clearer and reduces the clutter in the pixels of the image.

[0082] In some embodiments, Figure 9 This diagram illustrates the original time-domain signal of interference clutter in the matching medium between the unstripped antenna array and the intended target. Figure 10 This diagram illustrates the target time-domain signal after removing interference clutter from the matching medium between the antenna array and the predetermined target. Figure 9 and Figure 10 As shown, the original time-domain signal includes interface reflection clutter 91 of the matching medium and microwave signals 92 of the cranial tissue, while the target time-domain feature signal only contains microwave signals of the cranial tissue. Compared to the microwave signals 92 of the cranial tissue in the original time-domain signal, the microwave signals 93 of the cranial tissue represented by the target time-domain feature signal after removing the interference clutter of the matching medium are clearer.

[0083] In some embodiments, Figure 11 A schematic diagram of the imaging result obtained without removing interference clutter from the matching medium in the time domain is shown, along with a magnified schematic diagram of the target region in the imaging result. Figure 12 A schematic diagram of the imaging result obtained after removing interference clutter from the matched medium in the time domain is shown, along with a magnified schematic diagram of the target region in the imaging result, as shown below. Figure 11 and Figure 12 As shown, the location of the target area is more clearly displayed in the imaging results obtained after removing interference clutter from the matching medium between the antenna array and the predetermined target, and there are fewer clutter points in the pixels of the image.

[0084] In some embodiments, step S42 specifically includes the following steps: S421: obtaining the initial values ​​of the radius and angle of the antenna array and the average dielectric constant of electromagnetic waves propagating in the brain; S422: determining the spatial coordinate position of each antenna based on the initial values ​​of the radius and angle obtained in step S421; S423: determining the grid points of the imaging area based on the spatial coordinate positions determined in step S422, and generating a multi-dimensional spatial grid coordinate matrix containing spatial coordinates and dielectric properties by combining the average dielectric constant.

[0085] The method provided in the embodiments of this application determines a standardized grid space by obtaining the initial values ​​of the radius and angle of the antenna array and the average dielectric constant of electromagnetic waves propagating in the brain. This provides a precise geometric reference for the subsequent confocal energy superposition process, effectively preventing imaging distortion caused by the stretching and deformation of the wearable antenna radome with the antenna array on its exterior. This helps to accurately lock the spatial anatomical location. At the same time, the grid points of the imaging area are determined according to the spatial coordinate position, and a multi-dimensional spatial grid coordinate matrix containing spatial coordinates and dielectric properties is generated by combining the average dielectric constant. This can avoid deviations in wave velocity estimation caused by dielectric properties, ensuring the accuracy of subsequent time delay alignment and preventing defocusing or offset in imaging.

[0086] Figure 13 A schematic diagram of the imaging results obtained without utilizing the normalized grid space is shown. Figure 14 A schematic diagram of the imaging results obtained using a standardized grid space is shown, such as... Figure 13 and Figure 14 As shown, establishing a standardized grid space based on the physical topology parameters of the antenna array can provide an accurate geometric reference for the spatial mapping of scattered energy. This can effectively prevent the antenna transmitting and receiving positions from being inconsistent due to the stretching and deformation of the wearable antenna radome. The accurate antenna geometric position ensures the authenticity of the imaging results.

[0087] In some embodiments, step S43 specifically includes the following steps: S431: For the feature information obtained in step S30, traverse each frequency sampling point in the full frequency domain to determine the spatial wavenumber corresponding to each frequency point; S432: Based on the average dielectric constant obtained in step S421 and the spatial wavenumber determined in step S431, traverse each pair of transmitting and receiving antennas and use the multi-dimensional spatial grid coordinate matrix determined in step S423 to determine the transmission distance of the microwave signal propagating in each pair of transmitting and receiving antennas to each grid point; and based on the transmission distance and the microwave propagation speed determined by the average dielectric constant, determine the delay time of the microwave signal to each grid point.

[0088] The method provided in the embodiments of this application determines the propagation path of each pair of transmitting and receiving antennas in the grid space, and uses a multi-dimensional spatial grid coordinate matrix to determine the transmission distance of the microwave signal propagating in each pair of transmitting and receiving antennas to each grid point. This is beneficial for accurately capturing the phase change of microwaves in the predetermined target, thereby effectively correcting the signal delay error caused by multi-layer complex reflections. Furthermore, by using a multi-dimensional spatial grid coordinate matrix, the path relationship between each grid coordinate and the antenna can be determined in advance, which helps to improve the imaging speed in actual imaging.

[0089] In some embodiments, in step S431, the full frequency domain refers to the entire frequency range of the intracranial hemorrhage detection device used to acquire microwave signals. In such embodiments, the full frequency domain in this application embodiment may be, for example, 0.5 GHz - 2.5 GHz.

[0090] In some embodiments, step S44 specifically includes the following steps: S441: Based on the transmission distance, delay time determined in step S432 and the average dielectric constant obtained in step S421, perform time delay alignment on the target time-domain feature signal obtained in step S415 to determine the initial signal strength of each pair of transceiver antennas at each grid point; S442: Based on the transmission distance determined in step S432, determine the attenuation compensation weighting value of the initial signal strength at each grid point, and determine the signal strength map based on the initial signal strength and the weighting value; S443: Map the signal strength map to the spatial grid corresponding to the multi-dimensional spatial grid coordinate matrix to determine the image of the predetermined target.

[0091] The method provided in the embodiments of this application performs time delay alignment on the target time-domain feature signal based on the transmission distance, delay time, and average dielectric constant. Then, based on the transmission distance, it determines the attenuation compensation weighting value of the initial signal strength at each grid point, and determines the signal strength map based on the initial signal strength and the weighting value. This can utilize the scattering information in microwave signals of different frequencies, thereby improving the accuracy of the imaging results. By implementing gain weighting on the energy contribution of different antenna channels through the transmission distance, it can overcome the energy attenuation defect of microwaves propagating to deeper layers in lossy brain media, thereby improving the spatial resolution of the image.

[0092] In some embodiments, it can be understood that in step S441, the target time-domain feature signal used for time delay alignment is a time-domain pulse obtained by inverse Fourier transform of the frequency-domain scattering parameters. Therefore, the target time-domain feature signal used for time delay alignment includes scattering information of all probe frequencies.

[0093] In some embodiments, step S50 specifically includes the following steps: S51: determining the clutter region features and bleed region features of the predetermined target based on the feature information of the predetermined target; S52: determining the bleed region features of the target region through adversarial learning based on the clutter region features and the bleed region features; S53: determining the category of the microwave signal of the predetermined target based on the bleed region features.

[0094] The method provided in the embodiments of this application extracts clutter region features and bleed region features from the feature information, which can separate the two types of features in the coding space; by applying decoupling constraints to the clutter region features and bleed region features through adversarial learning, the residual clutter components in the bleed region features can be suppressed, thereby improving the accuracy of microwave signal category determination.

[0095] In some embodiments, step S51 specifically includes the following steps: S511: Encoding the feature information of the predetermined target towards clutter information to determine the clutter region features in the feature information; S512: Encoding the feature information of the predetermined target towards bleeding information to determine the bleeding region features in the feature information.

[0096] Figure 15 The diagram illustrates the result of independently encoding the feature information of a predetermined target according to an embodiment of this application. Figure 16 This diagram illustrates the result of single encoding of feature information of a predetermined target according to an embodiment of this application, such as... Figures 15 to 16 As shown, the method provided in the embodiments of this application, by independently encoding the feature information of the predetermined target towards clutter information and bleed information, can extract background clutter scattering and bleed region scattering from the feature information of the predetermined target, and make the clutter region features characterizing background clutter scattering and the bleed region features characterizing bleed region scattering clustered in the encoding space, thereby obtaining clutter region features and bleed region features. This helps to avoid the clutter signal and bleed signal from overlapping in a single encoding, thereby providing a basis for the subsequent feature decoupling process.

[0097] In some embodiments, step S52 specifically includes the following steps: S521: Determine the feature correlation measure between clutter region features and bleeding region features; S522: Determine the adversarial loss function based on the feature correlation measure, wherein the adversarial loss function is used to constrain the bleeding region features and clutter region features to be independent of each other in the feature space; S523: Perform adversarial training on the encoding process based on the adversarial loss function, so that the clutter region features and bleeding region features are orthogonal to each other; S524: Determine the decoupled bleeding region features based on the encoding process after adversarial training.

[0098] The method provided in the embodiments of this application determines an adversarial loss function based on the feature correlation measure between clutter region features and bleed region features, and performs adversarial training on the encoding process. This can impose orthogonality constraints on clutter region features and bleed region features, so that the bleed region features in the encoded output do not contain the background clutter scattering components represented by the clutter region features. By performing adversarial training on the encoding process, each part of the encoding process can focus on modeling its corresponding feature space in the adversarial game, thereby helping to improve the ability of the decoupled bleed region features to distinguish the bleed state.

[0099] In some embodiments, Figure 17 The diagram illustrates the result of adversarial training of the encoding process according to the method provided in the embodiments of this application, such as... Figure 17As shown, with the increase of adversarial training rounds, the correlation between clutter region features and bleeding region features gradually decreases, while the accuracy of category judgment results gradually increases.

[0100] In some embodiments, Figure 18 The diagram illustrates the clutter region features and hemorrhage region features obtained from the encoding process after adversarial training, as shown below. Figure 18 As shown, after adversarial training of the encoding process, mutually orthogonal clutter region features and bleeding region features can be obtained.

[0101] In some embodiments, step S53 specifically includes the following steps: S531: Determine the logical value of the bleeding category by linear transformation based on the characteristics of the decoupled bleeding area; S532: Determine the probability of bleeding based on the logical value; S533: Determine the category of the microwave signal based on the probability of bleeding.

[0102] The method provided in the embodiments of this application determines the logical value of the bleeding category by performing a linear transformation on the decoupled bleeding region features, and then determines the probability of bleeding based on the logical value. This can transform the bleeding region features into a quantifiable basis for category determination. Since the bleeding region features have been freed from clutter interference through adversarial learning, the bleeding probability of determining the bleeding category can more accurately reflect the true bleeding state of the predetermined target.

[0103] Embodiments of this application also provide a system for determining the feature attributes of a predetermined target, comprising: a microwave signal receiving module configured to receive microwave signals of the predetermined target; a microwave signal processing module configured to filter out noise from the microwave signals and normalize the noise-filtered microwave signals; a feature information extraction module configured to extract feature information of the predetermined target from the microwave signals obtained by the microwave signal processing module; an image reconstruction module configured to reconstruct an image of the predetermined target based on the feature information of the predetermined target; a category determination module configured to determine the category of the microwave signals of the predetermined target based on the feature information of the predetermined target; and a feature attribute determination module configured to determine the feature attributes of the target area of ​​the predetermined target based on the image and the category.

[0104] The system provided in the embodiments of this application receives microwave signals from a predetermined target and filters out noise from the microwave signals. Then, it normalizes the microwave signals and extracts the feature information of the predetermined target from the normalized microwave signals. Based on the feature information of the predetermined target, an image of the predetermined target is reconstructed. In this way, the feature information of the predetermined target can be determined through the reconstructed image, thereby determining the category of the microwave signal of the predetermined target. The reconstructed image can provide spatial quantitative information of the predetermined target, such as spatial location, morphology, and energy intensity, but it lacks pathological semantics, and lesions are easily confused with background clutter. The category can provide pathological semantic information on whether the predetermined target is bleeding, but it is only a qualitative judgment and lacks spatial and quantitative description. By jointly determining the feature attributes of the predetermined target by combining the image and the category, the category semantics are used to guide the precise positioning of the target area in the image, and the spatial information of the image is used to supplement the spatial quantitative information such as spatial location, morphology, and energy intensity that are lacking in the category discrimination. The two complement and corroborate each other, which helps to improve the detection sensitivity of the bleeding area, a small target, and at the same time helps to improve the completeness and accuracy of the determined target area.

[0105] In some embodiments, the feature attribute determination module includes: a semantic vector determination submodule, configured to determine the semantic vector of a category based on the category; a target region feature determination submodule, configured to determine the target region features of an image based on the semantic vector determined by the semantic vector determination module and the image; and a feature attribute determination submodule, configured to determine the feature attributes of a target region of a predetermined target based on the semantic vector and the target region features.

[0106] The system provided in the embodiments of this application, compared with the method of determining the feature attributes of the target region of a predetermined target based solely on image or text categories, uses semantic vectors and target region features of images to determine the feature attributes of the target region of a predetermined target. This is beneficial for achieving a balance between spatial accuracy and semantic interpretability, thereby improving the reliability of the determined feature attributes of the target region of the predetermined target.

[0107] In some embodiments, the semantic vector determination submodule includes: a text label determination unit configured to determine a text label for a category based on the category; a fixed-dimensional semantic vector determination unit configured to determine a fixed-dimensional semantic vector based on the text label; and a category semantic vector determination unit configured to perform feature space alignment on the fixed-dimensional semantic vector to determine the category semantic vector.

[0108] The system provided in the embodiments of this application aligns the semantic vectors of a fixed dimension with the feature space, so that the semantic vectors and the spatial feature vectors of the image are in the same space, which is beneficial for providing conditions for subsequent determination of feature similarity distribution.

[0109] In some embodiments, the target region feature determination submodule includes: a spatial feature vector determination unit configured to determine the spatial feature vector corresponding to the image based on the image; a cosine similarity determination unit configured to determine the cosine similarity between the semantic vector and the spatial feature vector based on the semantic vector and the image; a feature similarity distribution determination unit configured to determine the feature similarity distribution between the semantic vector and the spatial feature vector based on the cosine similarity; a matching degree determination unit configured to determine the matching degree between the semantic vector and the spatial feature vector based on the feature similarity distribution; a target region and spatial location determination unit configured to determine the target region and spatial location of the image based on the matching degree; and a target region feature determination unit configured to determine the target region features of the image based on the target region and spatial location.

[0110] The system provided in the embodiments of this application can more clearly express the high-dimensional features of microwave signals by determining the cosine similarity between semantic vectors and spatial feature vectors, thereby facilitating the accurate determination of the feature attributes and spatial location of the target area of ​​a predetermined target.

[0111] In some embodiments, the feature attribute determination submodule includes the following steps: a joint feature vector determination unit, configured to concatenate the semantic vector and target region features according to the channel dimension to determine the joint feature vector of the predetermined target; an attribute determination unit, configured to determine the spatial location attribute, morphological attribute and energy intensity attribute of the predetermined target according to the joint feature vector; and a feature attribute determination unit, configured to determine the feature attributes of the target region of the predetermined target according to the semantic vector and the attributes determined by the attribute determination unit.

[0112] The system provided in the embodiments of this application can determine the joint feature vector of a predetermined target by concatenating semantic vectors with target region features along the channel dimension. This allows for the simultaneous fusion of category semantic information and image spatial information, thereby facilitating a more accurate determination of the feature attributes of the target region of the predetermined target.

[0113] In some embodiments, the microwave signal processing module includes: a brain microwave signal acquisition submodule configured to acquire brain microwave signals of a preset number of healthy samples; a noise filtering submodule configured to determine a noise-filtered microwave signal based on the microwave signal obtained by the microwave signal receiving module and the brain microwave signals of the healthy samples obtained by the brain microwave signal acquisition unit; and a normalization processing submodule configured to determine a normalized microwave signal based on the noise-filtered microwave signal.

[0114] The system provided in the embodiments of this application acquires the brain microwave signal of a healthy sample and determines the noise-filtered microwave signal based on the acquired microwave signal and the brain microwave signal of the healthy sample. This helps to filter out electromagnetic reflection clutter generated by layered structures such as the skull and scalp, thereby improving the signal-to-noise ratio of the microwave signal. By normalizing the noise-filtered microwave signal, it is beneficial to eliminate amplitude shifts caused by non-target factors and map the microwave scattering parameters to a uniform numerical range, thereby improving the generalization ability of the prediction.

[0115] In some embodiments, the noise filtering submodule includes: a baseline mean signal determination unit, configured to determine the time-domain mean of the brain microwave signal of the healthy sample obtained by the brain microwave signal acquisition submodule, for each transceiver channel of the transceiver antenna array, and determine the baseline mean signal corresponding to the transceiver channel based on the time-domain mean; and a background clutter subspace determination unit, configured to determine a centered healthy sample matrix based on the brain microwave signal of the healthy sample and the baseline mean signal; perform singular value decomposition on the centered healthy sample matrix, and determine the first N principal components. The corresponding feature vectors are used to determine the background clutter subspace corresponding to the transceiver channel, where N is the preset number of principal components; the target microwave signal determination unit is configured to convert the microwave signal obtained by the microwave signal receiving module into a time domain signal, and determine the centered target microwave signal based on the baseline mean signal; the noise filtering unit is configured to determine the estimated clutter component in the target microwave signal based on the centered target microwave signal and the background clutter subspace; and determine the noise-filtered microwave signal based on the centered target microwave signal and the estimated clutter component.

[0116] The system provided in the embodiments of this application can determine the background clutter subspace that reflects the electromagnetic scattering patterns of static structures such as skulls and tissue boundaries by performing singular value decomposition on the brain microwave signal of a healthy sample and determining the eigenvectors corresponding to the first N principal components. Based on the centered target microwave signal and the background clutter subspace, the estimated clutter components in the target microwave signal are determined, which helps to accurately filter out the dominant background clutter and avoid misfiltering caused by the overlap of background clutter and effective signal in time delay and frequency band. By establishing a dedicated background clutter subspace model for each transceiver channel, it helps to adapt to the differences in asymmetric scattering paths caused by different antenna spatial positions.

[0117] In some embodiments, the normalization processing submodule includes: a maximum absolute value determination unit, configured to determine the amplitude of the microwave signal based on the noise-filtered microwave signal, and determine the maximum absolute value of the amplitude in the frequency dimension; a normalization denominator determination unit, configured to determine the normalization denominator based on the maximum absolute value and a preset constant; and a normalized microwave signal determination unit, configured to determine the normalized microwave signal based on the noise-filtered microwave signal and the normalization denominator.

[0118] The system provided in the embodiments of this application determines the maximum absolute value of the amplitude in the frequency dimension based on the amplitude of the microwave signal after noise filtering. Then, it determines the normalized denominator based on the maximum absolute value and a preset constant to determine the normalized microwave signal. This helps to strictly map the microwave signal to a preset amplitude range, unify the dimensions of each transmitting and receiving channel, and thus help to avoid strong channel signals drowning out weak channel signals. Determining the normalized microwave signal based on the microwave signal after noise filtering and the normalized denominator can preserve the relative amplitude ratio between the frequency points of the microwave signal, and thus can preserve the zero-point characteristics and phase structure of the microwave signal.

[0119] The applicant of this application also discovered that in the related technology, when using microwaves to image the patient's brain, there is distortion, which cannot accurately reflect the intracranial hemorrhage, thus affecting the timely and accurate judgment of the patient's condition in practical applications.

[0120] Based on this, embodiments of this application also provide an intracranial hemorrhage detection device, which improves the effect of using microwaves to image the patient's brain.

[0121] Figure 19 A schematic diagram of an intracranial hemorrhage detection device according to an embodiment of this application is shown, wherein the housing and media container are omitted. Figure 20 This shows from another perspective Figure 19 The intracranial hemorrhage detection device shown is, for example Figure 19 and Figure 20 As shown, the intracranial hemorrhage detection device of this application embodiment may include: a receiving part 20, a base 61, a flexible headgear 30, a signal transmission and acquisition component, and a processor. The receiving part 20 is disposed on the base 61; the receiving part 20 forms a receiving space for receiving a patient's cranium; the flexible headgear 30 is disposed in the receiving space, configured to be able to fill a liquid dielectric and deformable after pressure; the signal transmission and acquisition component is used to transmit microwave signals to and receive microwave signals from the flexible headgear 30, wherein the base 61 is configured to discharge the leaked liquid when the liquid dielectric leaks; the processor is configured to be communicatively connected to the signal transmission and acquisition component, used to receive microwave signals transmitted from the signal transmission and acquisition component, and is also configured to execute the method of the embodiment of this application.

[0122] The intracranial hemorrhage detection device provided in the embodiments of this application, due to the flexible headgear 30, can adapt to different patient cranial shapes by filling the flexible headgear 30 with a dielectric material and applying pressure to the flexible headgear 30 to allow it to expand and conform to the patient's cranium. Relying on the flexible deformation and fluid adaptive properties of the flexible headgear 30, it closely conforms to the patient's cranial contour, achieving a gapless fit, thereby improving the transmission efficiency of microwave signals and thus improving the quality of transmitted or received microwave signals. By setting the base 61 to prevent leakage of the liquid dielectric, it can... The leaked liquid is drained outwards to prevent mold growth or damage to the signal transmission and acquisition components caused by the leaked liquid dielectric. The processor receives the microwave signal from the brain and filters out the noise in the microwave signal. Then, it is normalized and the brain feature information of the normalized microwave signal is extracted. Based on the brain feature information, the brain image is reconstructed. In this way, the brain feature information can be determined through the reconstructed image, and then the brain feature attributes can be determined. Therefore, based on the image and category, the sensitivity and quantification of the brain feature attribute prediction are more comprehensive and accurate.

[0123] In some embodiments, the flexible headgear 30 may be the wearable antenna cover as described in the preceding embodiments.

[0124] In some embodiments, the antenna 10 may be disposed on the outside of the flexible headgear 30 in the receiving portion 20, and the antenna 10 may be configured to receive microwave signals from the port extension member 192 and transmit microwave signals to the flexible headgear 30; or receive microwave signals transmitted by another antenna 10 from the flexible headgear 30 and transmit them to the port extension member 192.

[0125] In some embodiments, a first drain hole 6110 is provided on the base 61 below the receiving portion 20 to drain leaked liquid. Since the leaked liquid dielectric can be drained out through the first drain hole 6110, mold growth or damage to the signal transmission and acquisition components caused by the leaked liquid dielectric is avoided.

[0126] In some embodiments, such as Figure 19 and Figure 20As shown, the signal transmission and acquisition assembly may include: multiple antennas 10, a vector network analyzer 191, and a port extender 192; the vector network analyzer 191 is configured to transmit microwave signals to the port extender 192, and receive microwave signals from the port extender 192, and analyze the received microwave signals; the port extender 192 is configured to extend the port of the vector network analyzer 191, receive microwave signals transmitted by the vector network analyzer 191, transmit microwave signals to one antenna 10, and receive corresponding microwave signals from another antenna 10 and transmit them to the vector network analyzer 191; the antenna 10 is disposed on the outer side of the flexible headgear 30 in the receiving portion 20, and is configured to receive microwave signals from the port extender 192 and transmit microwave signals to the flexible headgear 30; or receive microwave signals transmitted by another antenna 10 from the flexible headgear 30 and transmit them to the port extender 192. In the embodiments of this application, by setting the port extension 192, the vector network analyzer 191 can be connected to multiple pairs of antennas 10 to transmit and receive microwave signals, thereby increasing the number of antennas 10 and improving the accuracy of detection.

[0127] In some embodiments, the accommodating space has an opening on one side to facilitate the entry of the human brain.

[0128] In some embodiments, the first drain hole 6110 is also used to dissipate heat for the vector network analyzer 191 and the port expansion member 192. Dissipating heat for the vector network analyzer 191 and the port expansion member 192 through the first drain hole 6110 helps to prevent the vector network analyzer 191 and the port expansion member 192 from getting too hot.

[0129] In some embodiments, each antenna 10 is disposed in the receiving portion 20, thereby enabling the antenna 10 to transmit microwave signals to and receive microwave signals from the flexible headgear 30. The vector network analyzer 191 and the port extension 192 are disposed on one side of the receiving portion 20 on the base 61, and there is a gap between them and the base 61 in the height direction.

[0130] By placing the vector network analyzer 191 and port extension 192 on one side of the receiving portion 20, rather than at the bottom of the receiving portion 20, the height of the receiving portion 20 is prevented from becoming too high. This avoids discomfort caused by the patient's head being too high after entering the flexible headgear 30. Furthermore, the presence of a height gap between the vector network analyzer 191 and port extension 192 and the base 61 prevents them from being immersed in the liquid dielectric for extended periods, thus avoiding damage should a liquid dielectric leak occur.

[0131] In some embodiments, the port extender 192 includes a port extender body, two first connectors connected to the port extender body, and a plurality of second connectors. The two first connectors are used to connect to the vector network analyzer 191, and are respectively used to receive microwave signals transmitted by the vector network analyzer 191 and to transmit microwave signals to the vector network analyzer 191. Each second connector is used to connect to a corresponding antenna 10 via a signal transmission line 193 to transmit microwave signals from the vector network analyzer 191 to the antenna 10 or to receive corresponding microwave signals from the antenna 10. The two first connectors can be electrically connected to any two different second connectors. In such an embodiment, by providing the port extender 192, the ports of the vector network analyzer 191 are effectively extended, thereby enabling communication connections with more antennas 10.

[0132] In some embodiments, the two main surfaces (i.e. the two surfaces with the largest area) of the port extension body are respectively positioned facing the end 22 and the vector network analyzer 191; the second connector and the first connector are disposed on the side of the port extension body, which is conducive to saving space.

[0133] In some embodiments, such as Figure 23 and Figure 24 As shown, the receiving portion 20 includes a peripheral portion 21 and an end portion 22, which are connected to form a receiving space. The antenna 10 is disposed on the peripheral portion 21, and the port extension body is disposed facing the end portion 22 of the receiving portion 20. The vector network analyzer 191 is disposed on the side of the port extension body opposite to the end portion 22. Since the antenna 10 is disposed on the receiving portion 20 and the port extension body is disposed facing the end portion 22 of the receiving portion 20, it is beneficial to arrange the signal transmission line 193 for connecting the second connector of the antenna 10 and the port extension 192. By disposing the vector network analyzer 191 on the side of the port extension body opposite to the end portion 22, so that the vector network analyzer 191 faces the port extension body, it is beneficial to arrange the signal transmission line for connecting the first connector of the vector network analyzer 191 and the port extension 192.

[0134] In some embodiments, the antenna 10 is arranged along the circumferential direction of the periphery 21, thereby enabling more comprehensive acquisition of microwave signals.

[0135] In some embodiments, the number of antennas 10 can be 16.

[0136] In some embodiments, the signal transmission and acquisition assembly may further include multiple signal transmission lines 193 for connecting each antenna 10 to a second connector of the port extension 192. For example... Figure 24As shown, the signal transmission line 193 may include a first horizontal extension 1931 and a bent connecting section 1932 connected to the first horizontal extension 1931. The first horizontal extension 1931 extends along the periphery 21 of the receiving portion 20 and is used for detachable connection with the antenna 10; the bent connecting section 1932 is used for detachable connection with the second connector of the port extension 192. In this embodiment, since the first horizontal extensions 1931 of each signal transmission line 193 are arranged in parallel, and each bent connecting section 1932 is detachably connected to the second connector of the port extension 192, it is beneficial to simplify wiring and also to ensure that adjacent signal transmission lines 193 are spaced apart and not pressed together, thus avoiding interference with microwave signal transmission.

[0137] In some embodiments, the vector network analyzer 191 is arranged parallel to the port expansion body; the two opposite sides 1911 of the vector network analyzer 191 located in the vertical plane are parallel to the sides of the port expansion body located in the vertical plane, and the two opposite sides 1911 of the vector network analyzer 191 are provided with analyzer heat dissipation holes 1910 for heat dissipation. Since the two opposite sides 1911 of the vector network analyzer 191 located in the vertical plane do not face the port expansion body and the receiving portion 20, airflow is facilitated, which in turn is beneficial for heat dissipation.

[0138] In some embodiments, a fan is provided inside the vector network analyzer 191 to force air to flow through the analyzer's heat dissipation holes 1910 to dissipate heat from the vector network analyzer 191.

[0139] In some embodiments, the flexible headgear 30, when filled with liquid dielectric, is not circular to match the shape of the patient's skull. Specifically, when the flexible headgear 30 is installed in the receiving portion 20, the length of the flexible headgear 30 in the height direction is greater than its length in the horizontal direction.

[0140] In some embodiments, such as Figure 22As shown, the flexible headgear 30 may include a headgear body 301 and a headgear mounting fitting part connected to the headgear body 301. The headgear body 301 is configured to be filled with a liquid dielectric and deformable under pressure. The liquid dielectric is configured to match the impedance of the antenna 10 and the human brain tissue. The headgear mounting fitting part is configured to cooperate with the receiving part 20 to install the headgear body 301 in the receiving part 20. In the embodiments of this application, by configuring the headgear body 301 and the headgear mounting fitting part connected to the headgear body 301, the headgear body 301 can be filled with a liquid dielectric that facilitates impedance matching and deformed under inward pressure, allowing the patient's brain to enter the headgear body 301 and conform to the patient's brain after deformation, thereby improving the transmission efficiency of microwave signals. Furthermore, by configuring the headgear mounting fitting part to cooperate with the receiving part 20, the headgear body 301 can be installed in the receiving part 20, preventing movement of the headgear body 301 and affecting the transmission efficiency of microwave signals.

[0141] In some embodiments, such as Figure 23 and Figure 24 As shown, the receiving portion 20 may further include: a plurality of antenna mounting portions 23 and a head cover mounting portion 24, the head cover mounting portion 24 being used to mount the flexible head cover 30 in the receiving space; each antenna mounting portion 23 being used to mount an antenna 10 facing the flexible head cover 30 on the periphery 21.

[0142] The embodiments of this application, by providing an antenna mounting part 23 and a headgear mounting part 24, enable the flexible headgear 30 to be mounted on the receiving part 20 and the antenna 10 to be mounted facing the flexible headgear 30. This allows the antenna 10 to transmit microwave signals to the flexible headgear 30 and receive microwave signals from the flexible headgear 30, which is beneficial for adapting the flexible headgear 30 to the shape of different patients' skulls and improving the transmission efficiency of microwave signals.

[0143] In some embodiments, the antenna mounting portion 23 is disposed on the periphery 21.

[0144] In some embodiments, the headgear mounting mating part may include a first headgear mounting member 33, and a headgear mounting part 24 connected to an end 22 on the inner side of the periphery 21; the headgear mounting part 24 is configured to mate with the first headgear mounting member 33 and the headgear body 301 to prevent the headgear body 301 from moving.

[0145] In some embodiments, the headgear mounting portion 24, the peripheral portion 21, and the end portion 22 together form a groove 240. The first headgear mounting member 33 can be inserted into the groove 240 formed by the receiving portion 20, and the headgear body 301 can cover the groove 240. By enabling the first headgear mounting member 33 to be inserted into the groove 240 formed by the receiving portion 20, the first headgear mounting member 33 is prevented from detaching from the receiving portion 20. At the same time, since the headgear body 301 can cover the groove 240, the aesthetics are improved.

[0146] In some embodiments, the headgear mounting portion 24 further forms a guide groove 241, and the headgear mounting mating portion may further include a guide member 331 disposed on the first headgear mounting member 33. The guide member 331 is configured to slide in the guide groove 241 of the receiving portion 20 so that the first headgear mounting member 33 can be embedded in the groove 240. Since both the headgear body 301 and the first headgear mounting member 33 are made of flexible materials, it is difficult to directly embed the first headgear mounting member 33 into the groove 240. By providing the guide member 331 to cooperate with the guide groove 241, and by pushing the guide member 331 to slide in the guide groove 241, the first headgear mounting member 33 can be embedded into the groove 240.

[0147] In some embodiments, the headgear body 301 may abut against the headgear mounting portion 24 to prevent the headgear body 301 from moving toward the end portion 22, thereby avoiding interference with the detection of intracranial hemorrhage.

[0148] In some embodiments, the guide 331 protrudes from the first headgear mounting 33 to facilitate the application of force.

[0149] In some embodiments, the first headgear mounting member 33 further forms a headgear mounting hole 332, and the periphery 21 is provided with a periphery mounting hole 212. A positioning member passes through the headgear mounting hole 332 of the first headgear mounting member 33 and the periphery mounting hole 212 of the periphery 21, thereby mounting the first headgear mounting member 33 onto the periphery 21. In this embodiment, mounting the first headgear mounting member 33 onto the periphery 21 using a positioning member ensures the fixation of the flexible headgear 30 and facilitates the replacement of the flexible headgear 30.

[0150] In some embodiments, the positioning element may be a positioning pin.

[0151] In some embodiments, the peripheral mounting hole 212 corresponds to the guide groove 241. By configuring the guide member 331 to slide in the guide groove 241 of the receiving portion 20, the peripheral mounting hole 212 can be aligned with the head cover mounting hole 332 formed by the first head cover mounting member 33 of the flexible head cover 30, thereby facilitating the replacement of the flexible head cover 30.

[0152] In some embodiments, the number of headgear mounting holes 332 and guide members 331 are the same, and the position of each headgear mounting hole 332 is spaced apart from the corresponding guide member 331 along the extending direction of the first headgear mounting member 33; when the guide member 331 slides to the bottom in the guide groove 241, the headgear mounting hole 332 can be aligned with the peripheral mounting hole 212 of the receiving part 20, thereby simplifying the assembly of the first headgear mounting member 33 and the receiving part 20.

[0153] In some embodiments, the guide 331 is located at the end of the first headgear mounting member 33 away from the headgear body 301, and the guide groove 241 and the recess 240 have the same depth. When the guide 331 slides to the bottom in the guide groove 241, the first headgear mounting member 33 also reaches the bottom of the recess 240, so as to facilitate the assembly of the first headgear mounting member 33 with the receiving part 20.

[0154] In some embodiments, such as Figure 24 As shown, the receiving part 20 may further include: a housing connector 27, which is detachably connected to the front shell 643 of the intracranial hemorrhage detection device to press the flexible headgear 30 between the front shell 643 and the housing connector 27 to prevent the flexible headgear 30 from rotating and falling off.

[0155] In some embodiments, the headgear mounting mating part further includes a second headgear mounting member 34, which is connected to the headgear body 301 on the side away from the first headgear mounting member 33. When the front shell 643 of the intracranial hemorrhage detection device is detachably connected to the shell connector 27 of the receiving part 20, the second headgear mounting member 34 can be pressed to prevent the flexible headgear 30 from rotating and falling off.

[0156] In some embodiments, the flexible headgear 30 may further include: a liquid inlet / outlet connector 31, in fluid communication with the headgear body 301, for supplying liquid dielectric to and from the headgear body 301. The liquid inlet / outlet connector 31 is connected to the bottom of the headgear body 301. In such an embodiment, liquid dielectric can be supplied to and discharged from the headgear body 301 through the liquid inlet / outlet connector 31.

[0157] In some embodiments, the outer surface of the headgear body 301 is flush with the outer surface of the first headgear mounting member 33; the thickness of the first headgear mounting member 33 is less than the thickness of the headgear body 301; the inlet / outlet connector 31 is connected to the side of the bottom of the headgear body 301 and extends inside the first headgear mounting member 33 toward the side away from the second headgear mounting member 34. In such an embodiment, it is beneficial for the headgear body 301 and the first headgear mounting member 33 to fit snugly against the inner wall of the receiving portion 20, while the inlet / outlet connector 31 can be extended outward without occupying the space at the bottom of the receiving portion 20 for mounting the antenna 10.

[0158] In some embodiments, the flexible headgear 30 may further include an exhaust connector 32, disposed on the top of the headgear body 301, for venting air outward when liquid is filled into the headgear body 301 using the liquid inlet / outlet connector 31, and for allowing liquid to flow out from the exhaust connector 32 when the headgear body 301 is filled with liquid. By providing the exhaust connector 32 on the top of the headgear body 301, liquid can be smoothly filled into the headgear body 301 through the liquid inlet / outlet connector 31 by venting air outward, and the headgear body 301 can be filled with liquid, thereby improving the transmission efficiency of microwave signals.

[0159] In some embodiments, a housing relief groove 271 is formed on the top of the housing connector 27 to allow the exhaust connector 32 of the flexible headgear 30 to pass through. The exhaust connector 32 extends beyond the receiving portion 20 through the housing relief groove 271 and then extends along the periphery 21 toward the end portion 22. The housing relief groove 271 can limit the exhaust connector 32 to prevent it from moving or shifting, ensuring that the exhaust connector 32 is in the highest position, thereby preventing air residue in the headgear body 301.

[0160] In some embodiments, the detection device may further include: a liner 50 disposed radially inside the headgear mounting portion 24, the headgear mounting portion 24 being further configured to prevent the liner 50 from detaching. The liner 50 improves the aesthetics of the interior of the receiving portion 20.

[0161] In some embodiments, the detection device may further include a dielectric filling and discharging pressurization assembly 40, which is configured to fill the flexible headgear 30 with dielectric or discharge the dielectric from the flexible headgear 30, and to pressurize the flexible headgear 30 so that it can expand and fit against the patient's brain. In the embodiments of this application, by providing the flexible headgear 30, after the patient's brain enters the receiving portion 20, the dielectric filling and discharging pressurization assembly 40 can fill the flexible headgear 30 with dielectric and pressurize it so that it can expand and fit against the patient's brain. This allows it to adapt to different patient brain shapes. Relying on the flexible deformation and fluid adaptive characteristics of the flexible headgear 30, it closely fits the contour of the patient's brain, achieving a gapless fit and wrapping, thereby improving the transmission efficiency of microwave signals and thus improving the quality of microwave signals transmitted or received by the signal transmission and acquisition assembly.

[0162] In some embodiments, the dielectric filling and pressurizing assembly 40 may include: a dielectric container 41, an inlet / outlet liquid line 43, an outlet liquid line 44, and a drive member 42; the dielectric container 41 is used to contain liquid dielectric; the drive member 42 is used to drive the liquid dielectric to flow between the dielectric container 41 and the flexible headgear 30 via the inlet / outlet liquid line 43, and to pressurize the flexible headgear 30 through the inlet / outlet liquid line 43 after the flexible headgear 30 is filled with liquid dielectric, so that the flexible headgear 30 expands and fits against the skull; when the drive member 42 drives the liquid dielectric to flow into the flexible headgear 30, the outlet liquid line 44 is used to discharge gas from the flexible headgear 30; when the drive member 42 pressurizes the flexible headgear 30, the outlet liquid line 44 is disconnected. In such an embodiment, liquid can be smoothly introduced and pressurized into the flexible headgear 30.

[0163] In some embodiments, the drive element 42 is a pump.

[0164] In some embodiments, the inlet / outlet lines 43 connect the inlet / outlet connectors 31 of the flexible head sleeve 30 to the medium container 41, and the vent line 44 connects the vent connector 32 of the flexible head sleeve 30 to the medium container 41. The dielectric charging / discharging pressurization assembly 40 may further include a liquid detection element, a pressure measuring element 46, and a valve 47. The liquid detection element detects liquid in the vent line 44 to determine the filling amount of the flexible head sleeve 30; the pressure measuring element 46 measures the pressure within the flexible head sleeve 30; the valve 47 is disposed in the vent line 44 and is configured to disconnect the vent line 44 when the liquid detection element detects liquid in the vent line 44, thereby pressurizing the flexible head sleeve 30 using the drive element 42. In such an embodiment, since the vent connector 32 is located at the top, when the flexible head sleeve 30 is filled with dielectric, the dielectric enters the vent line 44, and when the liquid detection element detects the presence of liquid in the vent line 44, it indicates that the flexible head sleeve 30 is filled with dielectric. At this point, after disconnecting the exhaust pipe 44 via valve 47, pressure can be applied to the flexible headgear 30 using the drive unit 42, causing the flexible headgear 30 to expand and fit against the brain. The pressure measuring unit 46 is used to measure the pressure inside the flexible headgear 30, so that the drive unit 42 can be stopped from continuing to apply pressure based on the pressure value measured by the pressure measuring unit 46, ensuring that the pressure on the patient's brain is within an appropriate range.

[0165] In some embodiments, the liquid detection element may be disposed in the exhaust pipe 44.

[0166] In some embodiments, the dielectric charging and discharging pressurization assembly 40 may further include: a pressure measuring line 48 in fluid communication with the exhaust line 44 upstream of the valve 47; and a pressure measuring element 46 disposed in the pressure measuring line 48. By providing the pressure measuring line 48, the pressure inside the flexible head sleeve 30 can be measured when the exhaust line 44 is disconnected from the medium container 41.

[0167] In some embodiments, the antenna mounting portion 23 forms a plurality of mounting slots 230, which communicate with the receiving space; the antenna 10 is mounted in each mounting slot 230. Since the mounting slots 230 communicate with the receiving space, the antenna 10 can directly face the flexible headgear 30, so that after the flexible headgear 30 is pressurized and expanded, it can simultaneously fit with the patient's skull and the antenna 10, thereby improving the transmission efficiency of microwave signals.

[0168] In some embodiments, such as Figure 23 As shown, the surface of the peripheral portion 21 facing the flexible head cover 30 includes a plurality of planes 211 connected circumferentially. Each plane 211 forms an opening 2110 communicating with the mounting groove 230. The radiating port surface of the antenna 10 is flush with the plane 211 containing the opening 2110. By making the plane 211 flush with the radiating port surface of the corresponding antenna 10, it is beneficial to reduce microwave signal loss and improve the microwave signal reception quality.

[0169] In some embodiments, the antenna mounting portion 23 includes an extension portion 231 extending outward from the periphery of the opening 2110 and a connector 232 connected to the extension portion 231. The antenna 10 is connected to the connector 232, and the connector 232 forms a clearance hole 2320 for the wire connector 151 of the antenna 10 to pass through. Since the extension portion 231 extends outward from the periphery of the opening 2110, the antenna 10 can be mounted in the mounting groove 230 on the inner side of the periphery 21 without obstructing the radiating port surface of the antenna 10, thereby affecting the transmission and reception of microwave signals. At the same time, by forming a clearance hole 2320 in the connector 232, the wire connector 151 of the antenna 10 can extend to the outside through the clearance hole 2320, thereby realizing the transmission of microwave signals.

[0170] In some embodiments, such as Figure 23 As shown, the headgear mounting part 24 forms an inner lining positioning member 242, such as Figure 22 As shown, the inner liner 50 forms an inner liner positioning mating part 51 that mates with the inner liner positioning member 242, so that the inner liner 50 is positioned by the mating of the inner liner positioning member 242 and the inner liner positioning mating part 51, and the inner liner 50 is prevented from detaching.

[0171] In some embodiments, the inner liner positioning and fitting portion 51 is a guide groove extending along the extension direction of the peripheral portion 21; the inner liner positioning member 242 is a protrusion extending along the extension direction of the peripheral portion 21, and the guide groove and the protrusion slide into each other to prevent the inner liner 50 from dislodging. This arrangement helps to simplify the assembly of the inner liner 50.

[0172] In some embodiments, the liner 50 forms a liner clearance groove 52 to make way for the guide 331, thereby facilitating the engagement of the guide 331 with the guide groove 241, and also enabling the flexible headgear 30 to be disassembled and assembled without removing the liner 50.

[0173] In some embodiments, the liner 50 may be made of the same material as the flexible headgear 30 to improve aesthetics. In some embodiments, both the liner 50 and the flexible headgear 30 may be made of silicone.

[0174] In some embodiments, the detection device may further include: a mounting assembly, wherein the receiving portion 20, the vector network analyzer 191, and the port extension 192 are mounted on the base 61 via the mounting assembly.

[0175] In some embodiments, such as Figure 19 and Figure 20As shown, the mounting assembly may include: a first mounting member 621 disposed on a base 61, and a vector network analyzer 191 and a port extension member 192 mounted opposite each other on both sides of the first mounting member 621. A receiving portion 20 is connected to the first mounting member 621 on the side facing the port extension member 192. In this embodiment, since the receiving portion 20 is connected to the first mounting member 621 on the side facing the port extension member 192, and the vector network analyzer 191 and the port extension member 192 are mounted opposite each other on both sides of the first mounting member 621, the receiving portion 20, the port extension member 192, and the vector network analyzer 191 can be arranged sequentially at intervals, thereby facilitating wiring.

[0176] In some embodiments, the first mounting member 621 is a metal plate extending in a direction perpendicular to the base 61.

[0177] In some embodiments, the mounting assembly may further include: a second mounting member 622 and a third mounting member 623, which are disposed on the base 61; the receiving portion 20 is connected to the third mounting member 623 on the side opposite to the port extension 192, and is also connected to the second mounting member 622 on the side facing the port extension 192, thereby improving the stability of the receiving portion 20 and allowing for a more reasonable height setting of the receiving portion 20. In some embodiments, the receiving portion 20 may further include a neck support portion 26, which is connected to the peripheral portion 21 on the side away from the end 22. The neck support portion 26 provides support for the patient's neck and is connected to the third mounting member 623 above the third mounting member 623, thereby providing stable support for the patient's neck by the third mounting member 623.

[0178] In some embodiments, such as Figure 24 As shown, the receiving portion 20 includes a mounting fitting portion disposed at the end 22 for engaging with the first mounting member 621 and the second mounting member 622, thereby connecting the receiving portion 20 with the first mounting member 621 and the second mounting member 622.

[0179] In some embodiments, the mounting mating portion may include a plurality of connecting plates 250 extending from the end 22 in a direction away from the periphery 21, a plurality of first connecting posts 251 disposed on the connecting plates 250, and a plurality of second connecting posts 252. The first connecting posts 251 extend in a direction away from the periphery 21, and the connecting plates 250 are connected to the first mounting member 621 by the engagement of fasteners passing through the first mounting member 621 and the first connecting posts 251. The second connecting posts 252 are perpendicular to the extending direction of the first connecting posts 251, and the connecting plates 250 are connected to the second mounting member 622 by the engagement of fasteners passing through the second mounting member 622 and the second connecting posts 252. Since the antenna 10 is mounted in the receiving portion 20 and the flexible headgear 30 is mounted in the receiving portion 20, the stability requirements for the receiving portion 20 are higher. The above-described connection method can improve the stability of the receiving portion 20.

[0180] In some embodiments, there are two second mounting members 622, which are disposed opposite to each other on both sides of the connecting plate 250 to improve the stability of the receiving portion 20.

[0181] In some embodiments, the receiving portion 20 further includes a plurality of reinforcing ribs 253, which are used to increase the strength of the connecting plate 250.

[0182] In some embodiments, such as Figure 20 As shown, the mounting assembly also includes an auxiliary mounting component 624, which is connected to the first mounting component 621 and is used to clamp the vector network analyzer 191 between the auxiliary mounting component 624 and the first mounting component 621; thereby mounting the vector network analyzer 191 onto the first mounting component 621.

[0183] In some embodiments, the auxiliary mounting member 624 may include a mounting plate 6240 disposed parallel to the first mounting member 621 and mounting legs 6241 for connecting the mounting plate 6240 to the first mounting member 621. The mounting plate 6240 has a flange bent toward the first mounting member 621 for clamping and limiting the vector network analyzer 191, thereby eliminating the need to provide mounting holes or mounting members in the vector network analyzer 191 itself for connection with the first mounting member 621.

[0184] In some embodiments, such as Figure 19 and Figure 20 As shown, the mounting assembly may further include a fourth mounting member 625, which is connected to the top of the first mounting member 621 and is used to mount the valve 47 and the pressure measuring element 46. Since the height of the receiving part 20 is higher than that of the first mounting member 621, mounting the valve 47 and the pressure measuring element 46 on the fourth mounting member 625 not only makes reasonable use of space, but also facilitates the connection of the valve 47 and the pressure measuring element 46 to the exhaust pipe 44.

[0185] In some embodiments, the intracranial hemorrhage detection device may include: a processor 65, and a vector network analyzer 191 communicatively connected to the processor 65 for sending microwave signals to the processor 65 for processing.

[0186] In some embodiments, the processor 65 is configured to: analyze the microwave signal of the vector network analyzer 191; filter out clutter from the microwave signal to obtain a microwave signal containing only brain information; and determine brain imaging based on the clutter-filtered microwave signal. In the embodiments of this application, because the clutter from the microwave signal is filtered out, there are fewer clutter points in the pixels of the brain imaging, enabling a clearer display of the lesion location.

[0187] In some embodiments, multiple antennas 10 of the signal transmission and acquisition component are arranged in the receiving portion 20 along the same circumference; an antenna 10 for transmitting microwave signals and an antenna 10 for receiving microwave signals are collectively referred to as a pair of transmitting and receiving antennas.

[0188] In some embodiments, the processor is specifically configured to determine brain imaging by: determining the coordinate position of each antenna 10 and the average dielectric constant of the microwave signal propagating in the brain; determining grid points in the imaging region and generating a grid coordinate matrix; for the microwave signal, traversing each frequency point to determine the wavenumber corresponding to each frequency point; based on the average dielectric constant and the wavenumber, determining the transmission distance of the microwave signal propagating in each pair of transmit and receive antennas to each grid point by traversing each pair of transmit and receive antennas; determining the signal strength of each grid point based on the microwave signal and the transmission distance; and determining brain imaging based on the signal strength. In such embodiments, microwave signal information at different frequencies can be fully utilized, thereby improving the accuracy of the imaging results.

[0189] In some embodiments, the processor is specifically configured to determine brain imaging based on signal strength by: weighting and summing the signal strength of each grid point according to the distance between each pair of transmit and receive antennas to determine a signal strength map; and determining brain imaging based on the signal strength map. Embodiments of this application effectively improve imaging resolution by weighting the contribution of different antennas to signal strength based on the distance between each pair of transmit and receive antennas.

[0190] In some embodiments, the processor 65 is mounted on the side of the auxiliary mounting bracket 624 opposite to the vector network analyzer 191, which also removes the processor 65 from the base 61 to avoid contamination by leaked dielectrics. At the same time, it also allows the processor 65, the vector network analyzer 191, and the port expansion bracket 192 to be set close together, which is conducive to wiring and makes the layout of the entire intracranial hemorrhage detection device more compact.

[0191] In some embodiments, the inlet / outlet connector 31 is quickly connected to the inlet / outlet pipeline 43, the vent connector 32 is quickly connected to the vent pipeline 44, and the inlet / outlet pipeline 43 and the vent connector 32 are quickly connected to the medium container 41 respectively.

[0192] In some embodiments, the intracranial hemorrhage detection device may include a housing that mates with the housing 20 and the base 61 to enclose the housing 20, the antenna 10, the vector network analyzer 191, the port extension 192, and the processor 65, thereby improving aesthetics and providing protection for the components.

[0193] In some embodiments, the housing forms two sets of housing heat dissipation holes 641 arranged opposite to each other, facing the heat dissipation holes 1910 on both sides of the vector network analyzer 191, for airflow, thereby facilitating heat dissipation of the vector network analyzer 191.

[0194] In some embodiments, the intracranial hemorrhage detection device may include: a plurality of fans 66 disposed inside the housing and facing the housing heat dissipation hole 641, for driving ambient air into the housing, into the vector network analyzer 191 to dissipate heat from the vector network analyzer 191, and then leaving from the other side of the housing heat dissipation hole 641.

[0195] In some embodiments, such as Figure 22 As shown, the housing may include a main housing 642, a front housing 643, and a rear cover 644. The main housing 642 is detachably connected to the base 61, the front housing 643 is detachably connected to the receiving portion 20, and the rear cover 644 is detachably connected to the main housing 642. The dielectric receiving member 41 is detachably mounted on the rear cover 644. In this embodiment, dividing the housing into a main housing 642, a front housing 643, and a rear cover 644 facilitates the assembly of the housing and also facilitates the removal of the dielectric receiving member 41 for dielectric replacement.

[0196] In some embodiments, the side of the main housing 642 opposite to the receiving portion 20 is recessed toward the receiving portion 20, thereby forming a space for receiving the media receiving member 41; the rear cover 644 is L-shaped and detachably connected to the recess of the main housing 642 for installing the media receiving member 41.

[0197] In some embodiments, the main housing 642 forms a handle 646 above the recess to facilitate the transfer of the intracranial hemorrhage detection device.

[0198] In some embodiments, the main shell 642 is further provided with a heat dissipation hole 6421 below the recess for dissipating heat from the processor 65.

[0199] In some embodiments, such as Figure 22As shown, the flexible headgear 30 may also include a second headgear mounting member 34, which is connected to the headgear body 301. The front shell 643 is detachably connected to the receiving portion 20 and is used to press the second headgear mounting member 34 of the flexible headgear 30 to prevent the flexible headgear 30 from rotating.

[0200] In some embodiments, such as Figure 19 As shown, the housing may also include a pillow portion 645, which is detachably connected to a neck support portion 26 to provide support for the patient's neck and improve aesthetics.

[0201] In some embodiments, the mounting assembly further includes a reinforcement 626 for engaging with the pillow portion 645 to provide support.

[0202] In some embodiments, the receiving portion 20 may further include an inlet / outlet positioning member 28, disposed at the bottom of the peripheral portion 21 and located between the antenna mounting portion 23 and the end portion 22, and extending to the bottom of the end portion 22; the inlet / outlet connector 31 of the flexible head cover 30 extends to the outside through the inlet / outlet positioning member 28. By providing the inlet / outlet positioning member 28 to position the inlet / outlet connector 31, displacement of the inlet / outlet connector 31 is prevented, and since the inlet / outlet positioning member 28 is located between the antenna mounting portion 23 and the end portion 22, it does not affect the installation of the antenna 10.

[0203] In some embodiments, such as Figure 22 As shown, the base 61 is provided with an opening and a bottom cover 611 for opening or closing the opening. The opening is located below the receiving part 20 and directly below the inlet / outlet positioning member 28, thereby facilitating the disassembly and assembly of the inlet / outlet pipe 43 and the inlet / outlet connector 31 by opening the bottom cover 611.

[0204] In some embodiments, the opening is located between the two second mounting members 622 and on the side of the first mounting member 621 facing the receiving portion 20, so as not to affect the overall installation strength.

[0205] In some embodiments, a first drain hole 6110 may be formed on the bottom cover 611.

[0206] In some embodiments, such as Figure 22 As shown, the base 61 is provided with a second drain hole 612 below the connection between the drive component 42 and the inlet / outlet pipe 43 and the exhaust pipe 44, for draining liquid in case of leakage.

[0207] In some embodiments, the intracranial hemorrhage detection device may further include an operating element 67 and a plurality of data communication interfaces 68, disposed on the side of the main housing 642 opposite to the receiving portion 20 and located below the rear cover 644. Since the processor 65 is also disposed on the side opposite to the receiving portion 20, placing the operating element 67 and the plurality of communication interfaces 68 at the rear of the main housing 642 below the rear cover 644 facilitates wiring. The operating element 67 may be a button or a touch screen.

[0208] In some embodiments, such as Figure 20 As shown, the intracranial hemorrhage detection device may further include: a USB splitter 652, disposed on the side of the base 61 facing away from the processor 65 and the receiving portion 20, for connecting to multiple communication interfaces 68 to expand the communication interfaces of the processor 65. Figure 22 As shown, the base 61 has a heat dissipation hole 613 located below the USB splitter 652 for heat dissipation of the USB splitter 652.

[0209] In some embodiments, the driver 42 and the USB splitter 652 are disposed on the side of the base 61 away from the receiving portion 20 to make efficient use of space.

[0210] Embodiments of this application also provide an intracranial hemorrhage detection device, comprising: a receiving portion forming a receiving space for receiving a patient's brain; a flexible headgear disposed in the receiving space, configured to be filled with a liquid dielectric and deformable under pressure; a signal transmission and acquisition component for transmitting microwave signals to and receiving microwave signals from the flexible headgear; a dielectric filling and discharging pressurization component configured to fill the flexible headgear with a dielectric or discharge the dielectric from the flexible headgear, and to pressurize the flexible headgear so that it can expand to fit the patient's brain; and a processor configured to be communicatively connected to the signal transmission and acquisition component, for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to execute the method of determining the characteristic attributes of a predetermined target according to embodiments of this application.

[0211] In some embodiments, the dielectric filling and discharging pressurization assembly includes: a dielectric container for containing liquid dielectric; inlet and outlet pipes and a drive unit for driving the liquid dielectric to flow between the dielectric container and the flexible headgear via the inlet and outlet pipes, and for pressurizing the flexible headgear through the inlet and outlet pipes after the flexible headgear is filled with liquid dielectric to cause the flexible headgear to expand and fit against the cranium; and an exhaust pipe for discharging gas from the flexible headgear when the drive unit drives the liquid dielectric to flow into the flexible headgear; and for disconnecting the exhaust pipe when the drive unit pressurizes the flexible headgear.

[0212] In some embodiments, the dielectric charging and discharging pressurization assembly further includes: a liquid detection element for detecting liquid in the exhaust pipe to determine the filling amount of the flexible headgear; a pressure measuring element for measuring the pressure inside the flexible headgear; and a valve disposed in the exhaust pipe, the valve being configured to disconnect the exhaust pipe when the liquid detection element detects liquid in the exhaust pipe, so as to pressurize the flexible headgear using a drive element.

[0213] In some embodiments, the dielectric charging and discharging pressurization assembly further includes: a pressure measuring line in fluid communication with the exhaust line upstream of the valve; and a pressure measuring element disposed in the pressure measuring line.

[0214] In some embodiments, the signal transmission and acquisition assembly includes: a plurality of antennas, a vector network analyzer, and a port extender; the vector network analyzer is configured to transmit microwave signals to the port extender, and to receive microwave signals from the port extender and analyze the received microwave signals; the port extender is configured to extend the port of the vector network analyzer, receive microwave signals transmitted by the vector network analyzer, transmit microwave signals to one antenna, and receive corresponding microwave signals from another antenna and transmit them to the vector network analyzer; the antenna is disposed on the outside of the flexible headgear in the receiving portion, and is configured to receive microwave signals from the port extender and transmit microwave signals to the flexible headgear; or, receive microwave signals transmitted by another antenna from the flexible headgear and transmit them to the port extender.

[0215] In some embodiments, the device further includes a base and a mounting assembly, wherein the housing, the vector network analyzer, and the port extension are mounted on the base via the mounting assembly.

[0216] In some embodiments, the mounting assembly includes: a first mounting member disposed on a base, a vector network analyzer and a port extension member mounted opposite each other on both sides of the first mounting member; and a receiving portion connected to the first mounting member on the side facing the port extension member.

[0217] In some embodiments, the mounting assembly further includes: a second mounting member and a third mounting member, the second mounting member and the third mounting member being disposed on the base; the side of the receiving portion opposite to the port extension member is connected to the third mounting member, and the side of the receiving portion facing the port extension member is also connected to the second mounting member.

[0218] In some embodiments, the mounting component further includes an auxiliary mounting element connected to the first mounting element for clamping the vector network analyzer between the auxiliary mounting element and the first mounting element.

[0219] In some embodiments, the mounting assembly further includes a fourth mounting member connected to the first mounting member on top of the first mounting member for mounting the valves and pressure measuring elements of the dielectric charging and discharging pressurization assembly.

[0220] Embodiments of this application also provide a computer storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the methods of the embodiments of this application.

[0221] Regarding the embodiments of this application, it should also be noted that, without conflict, the embodiments of this application and the features in the embodiments can be combined with each other to obtain new embodiments.

[0222] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. A method for determining the characteristic attributes of a predetermined target, characterized in that, It includes the following steps: S10: Receive the microwave signal from the predetermined target; S20: Filter out noise from the microwave signal and normalize the microwave signal after filtering out the noise; S30: Extract the feature information of the predetermined target from the microwave signal obtained in step S20; S40: Reconstruct an image of the predetermined target based on the feature information of the predetermined target; S50: Determine the category of the microwave signal of the predetermined target based on the characteristic information of the predetermined target; S60: Determine the characteristic attributes of the target region of the predetermined target based on the image and the category.

2. The method according to claim 1, characterized in that, Step S60 specifically includes the following steps: S61: Determine the semantic vector of the category based on the category; S62: Based on the semantic vector determined in step S61 and the image, determine the target region features of the image; S63: Determine the feature attributes of the target region of the predetermined target based on the semantic vector and the target region features.

3. The method according to claim 2, characterized in that, Step S61 specifically includes the following steps: S611: Determine the text label for the category based on the category; S612: Determine a fixed-dimensional semantic vector based on the text tags; S613: Align the semantic vectors of the fixed dimension in the feature space to determine the semantic vectors of the category.

4. The method according to claim 2, characterized in that, Step S62 specifically includes the following steps: S621: Determine the spatial feature vector corresponding to the image based on the image; S622: Determine the cosine similarity between the semantic vector and the spatial feature vector based on the semantic vector and the image; S623: Determine the feature similarity distribution between the semantic vector and the spatial feature vector based on the cosine similarity; S624: Determine the degree of matching between the semantic vector and the spatial feature vector based on the feature similarity distribution; S625: Determine the target region and spatial location of the image based on the matching degree; S626: Determine the target region features of the image based on the target region and spatial location.

5. The method according to claim 2, characterized in that, Step S63 specifically includes the following steps: S631: Based on the semantic vector and the target region features, the semantic vector and the target region features are concatenated along the channel dimension to determine the joint feature vector of the predetermined target; S632: Determine the spatial location attribute, morphological attribute, and energy intensity attribute of the predetermined target based on the joint feature vector; S633: Based on the semantic vector and the attributes determined in step S632, determine the feature attributes of the target region of the predetermined target.

6. The method according to any one of claims 1-5, characterized in that, Step S20 specifically includes the following steps: S21: Acquire brain microwave signals from a preset number of healthy samples; S22: Based on the microwave signal obtained in step S10 and the brain microwave signal of the healthy sample obtained in step S21, determine the microwave signal after noise filtering. S23: Determine the normalized microwave signal based on the microwave signal after noise filtering.

7. The method according to claim 6, characterized in that, Step S22 specifically includes the following steps: S221: Based on the brain microwave signal of the healthy sample obtained in step S21, for each transceiver channel of the transceiver antenna array, determine the time-domain mean of the brain microwave signal of the healthy sample, and determine the baseline mean signal corresponding to the transceiver channel based on the time-domain mean. S222: Based on the brain microwave signal of the healthy sample and the baseline mean signal, determine the centered healthy sample matrix; perform singular value decomposition on the centered healthy sample matrix and determine the feature vectors corresponding to the first N principal components; based on the feature vectors, determine the background clutter subspace corresponding to the transceiver channel, where N is the preset number of principal components; S223: Based on the microwave signal obtained in step S10, convert the microwave signal into a time domain signal, and determine the centered target microwave signal based on the baseline mean signal; S224: Based on the centered target microwave signal and the background clutter subspace, determine the estimated clutter component in the target microwave signal; based on the centered target microwave signal and the estimated clutter component, determine the microwave signal after noise filtering.

8. The method according to claim 6, characterized in that, Step S23 specifically includes the following steps: S231: Based on the microwave signal after noise removal, determine the amplitude of the microwave signal, and determine the maximum absolute value of the amplitude in the frequency dimension; S232: Determine the normalized denominator based on the maximum absolute value and the preset constant; S233: Determine the normalized microwave signal based on the microwave signal after noise removal and the normalized denominator.

9. A system for determining the characteristic attributes of a predetermined target, characterized in that, It includes: A microwave signal receiving module, configured to receive microwave signals from the predetermined target; A microwave signal processing module configured to filter out noise from the microwave signal and normalize the microwave signal after the noise has been filtered out; The feature information extraction module is configured to extract feature information of the predetermined target from the microwave signal obtained by the microwave signal processing module. An image reconstruction module is configured to reconstruct an image of the predetermined target based on the feature information of the predetermined target; The category determination module is configured to determine the category of the microwave signal of the predetermined target based on the feature information of the predetermined target; The feature attribute determination module is configured to determine the feature attributes of the target region of the predetermined target based on the image and the category.

10. The system according to claim 9, characterized in that, The feature attribute determination module includes: A semantic vector determination submodule is configured to determine the semantic vector of the category based on the category. The target region feature determination submodule is configured to determine the target region features of the image based on the semantic vector determined by the semantic vector determination module and the image. The feature attribute determination submodule is configured to determine the feature attributes of the target region of the predetermined target based on the semantic vector and the target region features.

11. The system according to claim 10, characterized in that, The semantic vector determination submodule includes: A text label determining unit is configured to determine a text label for the category based on the category. A fixed-dimensional semantic vector determination unit is configured to determine a fixed-dimensional semantic vector based on the text tags; A semantic vector determination unit for a category is configured to perform feature space alignment on the semantic vectors of the fixed dimension to determine the semantic vectors of the category.

12. The system according to claim 10, characterized in that, The target region feature determination submodule includes: A spatial feature vector determination unit is configured to determine the spatial feature vector corresponding to the image based on the image. The cosine similarity determination unit is configured to determine the cosine similarity between the semantic vector and the spatial feature vector based on the semantic vector and the image. The feature similarity distribution determination unit is configured to determine the feature similarity distribution between the semantic vector and the spatial feature vector based on the cosine similarity. The matching degree determination unit is configured to determine the matching degree between the semantic vector and the spatial feature vector based on the feature similarity distribution. The target region and spatial location determination unit is configured to determine the target region and spatial location of the image based on the matching degree. The target region feature determination unit is configured to determine the target region features of the image based on the target region and its spatial location.

13. The system according to claim 10, characterized in that, The feature attribute determination submodule includes the following steps: The joint feature vector determination unit is configured to perform channel-dimensional concatenation of the semantic vector and the target region features based on the semantic vector and the target region features to determine the joint feature vector of the predetermined target; The attribute determination unit is configured to determine the spatial location attribute, morphological attribute, and energy intensity attribute of the predetermined target based on the joint feature vector. The feature attribute determination unit is configured to determine the feature attributes of the target region of the predetermined target based on the semantic vector and the attributes determined by the attribute determination unit.

14. The system according to any one of claims 9-13, characterized in that, The microwave signal processing module includes: The brain microwave signal acquisition submodule is configured to acquire brain microwave signals from a preset number of healthy samples. The noise filtering submodule is configured to determine the noise-filtered microwave signal based on the microwave signal obtained by the microwave signal receiving module and the brain microwave signal of the healthy sample obtained by the brain microwave signal acquisition unit. The normalization processing submodule is configured to determine the normalized microwave signal based on the microwave signal after noise filtering.

15. The system according to claim 14, characterized in that, The noise filtering submodule includes: The baseline mean signal determination unit is configured to determine the time-domain mean of the brain microwave signal of the healthy sample based on the brain microwave signal of the healthy sample obtained by the brain microwave signal acquisition submodule, for each transceiver channel of the transceiver antenna array, and determine the baseline mean signal corresponding to the transceiver channel based on the time-domain mean. The background clutter subspace determination unit is configured to determine a centered healthy sample matrix based on the brain microwave signal of the healthy sample and the baseline mean signal; perform singular value decomposition on the centered healthy sample matrix and determine the feature vectors corresponding to the first N principal components; and determine the background clutter subspace corresponding to the transceiver channel based on the feature vectors, where N is a preset number of principal components. The target microwave signal determination unit is configured to convert the microwave signal obtained by the microwave signal receiving module into a time domain signal, and determine the centered target microwave signal based on the baseline mean signal. The noise filtering unit is configured to determine the estimated clutter component in the target microwave signal based on the centered target microwave signal and the background clutter subspace; and to determine the noise-filtered microwave signal based on the centered target microwave signal and the estimated clutter component.

16. The system according to claim 14, characterized in that, The normalization processing submodule includes: The maximum absolute value determination unit is configured to determine the amplitude of the microwave signal based on the microwave signal after noise removal, and to determine the maximum absolute value of the amplitude in the frequency dimension. The normalized denominator determination unit is configured to determine the normalized denominator based on the maximum absolute value and a preset constant. A normalized microwave signal determination unit is configured to determine a normalized microwave signal based on the noise-filtered microwave signal and the normalization denominator.

17. An intracranial hemorrhage detection device, characterized in that, It includes: The receiving section forms a space to accommodate the patient's skull; A base, wherein the receiving portion is disposed on the base; A flexible headgear, disposed in the receiving space, is configured to be used to fill a liquid dielectric and to deform under pressure; The base is also configured to drain the leaked liquid when the liquid dielectric leaks. A signal transmission and acquisition component is used to transmit microwave signals to the flexible headgear and receive microwave signals from the flexible headgear. A processor configured to be communicatively connected to the signal transmission and acquisition component for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to be communicatively connected to the signal transmission and acquisition component for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to perform the method according to any one of claims 1-8.

18. An intracranial hemorrhage detection device, characterized in that, It includes: The receiving section forms a space to accommodate the patient's skull; A flexible headgear, disposed in the receiving space, is configured to be used to fill a liquid dielectric and to deform under pressure; A signal transmission and acquisition component is used to transmit microwave signals to the flexible headgear and receive microwave signals from the flexible headgear. The dielectric filling and discharging pressurization assembly is configured to fill the flexible headgear with dielectric or discharge the dielectric from the flexible headgear, and to pressurize the flexible headgear so that the flexible headgear can expand and fit against the patient's cranium; A processor configured to be communicatively connected to the signal transmission and acquisition component for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to be communicatively connected to the signal transmission and acquisition component for receiving microwave signals transmitted from the signal transmission and acquisition component, and further configured to perform the method according to any one of claims 1-8.

19. A computer storage medium, characterized in that, It stores a computer program thereon, which is executed by a processor to implement the method as described in any one of claims 1-8.

Images