A tissue oxygen real-time monitoring system based on multispectral fusion

Abstract

This invention belongs to the field of tissue oxygen monitoring technology. It provides a real-time tissue oxygen monitoring system based on multispectral fusion, comprising: addressing the problems of low monitoring accuracy caused by the lack of differentiation between wavelength penetration depth and interference between shallow and deep signals in existing technologies; firstly, acquiring near-infrared reflectance spectral signals at three wavelengths: 730nm, 808nm, and 940nm; then calculating the tissue oxygen contribution weight based on the differences in wavelength scattering coefficients, generating and validating a spatial weight matrix, and optimizing substandard matrices using a BP neural network optimization model; next, using the validated matrix to weightedly superimpose the spectral signals, eliminating shallow interference to form a fused total tissue oxygen signal; finally, extracting the light absorption intensity from the total signal according to the Lambert-Beer law, and calculating and outputting the tissue oxygen saturation. This invention achieves wavelength-depth layered fusion, enhances effective deep signals, improves the accuracy and depth resolution of tissue oxygen monitoring, and is suitable for real-time monitoring of deep muscle tissue.

Description

Technical Field

[0001] This invention belongs to the field of tissue oxygen monitoring technology, specifically a real-time tissue oxygen monitoring system based on multispectral fusion. Background Technology

[0002] In the field of biomedical testing, tissue oxygen saturation monitoring technology based on multispectral imaging or spectral detection has been widely used to assess blood oxygenation and microcirculation status in tissues such as body surface tissues and muscle tissues. Existing conventional multispectral fusion algorithms typically assume that incident light of different wavelengths has the same or similar transmission path, penetration depth, and scattering distribution within biological tissues. Based on this, multi-band spectral signals are weighted, superimposed, or fused at the feature level to reconstruct and output tissue oxygen signals.

[0003] However, biological tissues are highly non-uniform scattering media, and the interaction between different wavelengths of light and tissues varies significantly: short-wavelength light scatters strongly, absorbs highly, and penetrates shallowly in the epidermis, mainly reflecting the oxygen signal of superficial tissues such as the epidermis; long-wavelength light scatters weakly and penetrates deeper, reflecting the oxygen signal of deep tissues such as subcutaneous muscles. Existing conventional multispectral fusion methods do not distinguish between the detection depth and tissue layer corresponding to different wavelengths, directly fusing superficial epidermal signals and deep muscle signals globally. This leads to severe dilution and masking of the oxygen signal in deep muscles by the oxygen signal in the epidermis, failing to accurately reflect the oxygen metabolism status of the target deep tissues. This results in low monitoring results, insufficient spatial and depth resolution, and is difficult to meet the needs for accurate and real-time monitoring of oxygen saturation in deep tissues such as muscles.

[0004] Therefore, given the shortcomings of existing multispectral fusion technologies, such as lack of depth stratification, easy signal dilution, and insufficient monitoring accuracy, there is an urgent need for a real-time tissue oxygen monitoring system capable of distinguishing wavelength penetration depth, suppressing epidermal interference, and achieving optimized fusion of stratified signals. To this end, this invention provides a real-time tissue oxygen monitoring system based on multispectral fusion. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is: One objective of this invention is to provide a real-time tissue oxygen monitoring system based on multispectral fusion, comprising: Spectral signal acquisition module: Acquires near-infrared reflectance spectral signals of at least three wavelengths from the tissue to be monitored; Weight matrix optimization module: Based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored, calculate the oxygen contribution weights of the three wavelength signals to tissues at different depth levels, generate a spatial weight matrix, verify the calculation accuracy of the spatial weight matrix, construct a spatial weight matrix optimization model, and optimize spatial weight matrices that do not meet the accuracy verification standards. Signal fusion and interference removal module: The near-infrared reflectance spectrum signal is weighted and superimposed using a spatial weight matrix that has been verified by calculation accuracy, so as to eliminate the interference of superficial tissues on the oxygen signals of transitional tissues and deep muscle tissues, and form a fused total tissue oxygen signal. Tissue oxygen saturation calculation module: Based on the fused total tissue oxygen signal, calculate and output tissue oxygen saturation.

[0007] As a further improvement of the present invention, the specific process for obtaining the near-infrared reflectance spectral signals of the at least three wavelengths is as follows: At least three different wavelengths of near-infrared light are emitted sequentially or simultaneously to the tissue to be monitored. After the near-infrared light is incident on the tissue to be monitored, the reflected light carrying tissue oxygen information is converted into an analog electrical signal. The analog electrical signal is filtered, amplified, and converted from analog to digital to obtain digitized multi-wavelength near-infrared reflectance spectral data. The signal-to-noise ratio and validity of the multi-wavelength near-infrared reflectance spectral data are verified to obtain near-infrared reflectance spectral signals of at least three wavelengths.

[0008] As a further improvement of the present invention, the three wavelengths of near-infrared reflectance spectral signals include: The three different wavelengths of near-infrared light include 730nm for superficial tissues, 808nm for transitional tissues, and 940nm for deep muscle tissues.

[0009] As a further improvement of the present invention, the process of calculating the tissue oxygen contribution weight is as follows: The scattering coefficients of each wavelength at the three depth levels of superficial tissue, transitional tissue, and deep muscle tissue are summed, and the scattering coefficient of each wavelength at a single depth level is divided by the sum of the scattering coefficients of the corresponding wavelength at all depth levels to form the tissue oxygen signal contribution weight of the corresponding depth level.

[0010] As a further improvement of the present invention, the specific process for generating the generated spatial weight matrix is ​​as follows: The contribution weights of superficial tissue, transitional tissue, and deep muscle tissue corresponding to 730nm, 808nm, and 940nm are arranged in a regularized manner according to the wavelength dimension and the depth dimension, forming a two-dimensional spatial weight matrix composed of wavelength and depth.

[0011] As a further improvement of the present invention, the specific process of verifying the accuracy of the spatial weight matrix calculation is as follows: Using a simulated tissue model or a standard tissue oxygen monitoring device as the true value of the validation benchmark, the tissue oxygen result calculated by the spatial weight matrix is ​​compared with the true value of the validation benchmark to obtain the calculation deviation between the tissue oxygen result and the validation benchmark. If the calculation deviation is less than or equal to the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix meets the usage requirements. If the calculation deviation is greater than the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix does not meet the usage requirements.

[0012] As a further improvement of the present invention, the specific process of constructing the spatial weight matrix optimization model is as follows: The spatial weight matrices with and without the required accuracy are collected and used together to form the model training sample set, which includes the corresponding scattering coefficient data, spatial weight matrix data, the verification benchmark true value, and the calculation deviation data. The BP neural network is used as the basic network model, and the BP neural network is trained using the model training sample set to obtain the trained spatial weight matrix optimization model.

[0013] As a further improvement of the present invention, the specific process of optimizing the spatial weight matrix that fails the accuracy verification is as follows: The spatial weight matrix whose current accuracy does not meet the requirements, the corresponding scattering coefficient data, and the calculation deviation are all input into the spatial weight matrix optimization model, and the corrected optimal tissue oxygen contribution weight is output.

[0014] As a further improvement of the present invention, the specific process for forming the fused tissue oxygen total signal is as follows: The acquired near-infrared reflectance spectral signals were layered and distinguished according to wavelength and depth dimensions to obtain the spectral signal components corresponding to shallow tissues at 730nm wavelength, the spectral signal components corresponding to transitional tissues at 808nm wavelength, and the spectral signal components corresponding to deep muscle tissues at 940nm wavelength. Based on the validated spatial weight matrix, the optimal contribution weights corresponding to 730nm / shallow tissue, 808nm / transitional tissue, and 940nm / deep muscle tissue were extracted and multiplied element-wise to obtain weighted spectral signals. All weighted spectral signals were linearly superimposed and summed to form the fused total tissue oxygen signal.

[0015] As a further improvement of the present invention, the specific process for calculating the output tissue oxygen saturation is as follows: Based on the fused total tissue oxygen signal, the light absorption intensity information corresponding to the three wavelengths of 730nm, 808nm, and 940nm was extracted respectively. According to the Lambert-Beer law and the principle of near-infrared spectroscopy for tissue oxygen detection, the correspondence between light absorption intensity and the concentrations of oxyhemoglobin and deoxyhemoglobin was established. The relative content ratio of oxyhemoglobin to deoxyhemoglobin in the tissue to be tested was obtained, and the final tissue oxygen saturation was calculated from the relative content ratio.

[0016] The second objective of this invention is to provide a method for real-time monitoring of tissue oxygenation based on multispectral fusion, comprising: Step 1: Acquire near-infrared reflectance spectral signals of at least three wavelengths from the tissue to be monitored; Step 2: Based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored, calculate the oxygen contribution weights of the three wavelength signals to tissues at different depth levels, generate a spatial weight matrix, verify the accuracy of the spatial weight matrix calculation, construct an optimization model for the spatial weight matrix, and optimize the spatial weight matrix that does not meet the accuracy verification. Step 3: Use the spatial weight matrix verified by calculation accuracy to perform weighted superposition of near-infrared reflectance spectral signals to eliminate the interference of superficial tissues on the oxygen signals of transitional tissues and deep muscle tissues, forming a fused total tissue oxygen signal; Step 4: Calculate the output tissue oxygen saturation based on the fused total tissue oxygen signal.

[0017] The beneficial effects of this invention are as follows: 1. Targeting the pain points of existing technologies, this approach differentiates the tissue penetration depth and corresponding layers of 730nm, 808nm, and 940nm near-infrared light, avoiding signal dilution in superficial epidermis and masking of oxygen signals in deep muscles, thus fundamentally addressing the problem of low monitoring results. A spatial weight matrix is ​​constructed and combined with a BP neural network optimization model to achieve precise optimization and verification of contribution weights, making the weighted superposition more closely reflect the actual optical characteristics of the tissue and improving the accuracy and reliability of monitoring.

[0018] 2. By employing a layered weighted signal fusion method, the effective oxygen signal components in the transition layer and deep muscle tissue are enhanced, while shallow scattering interference is weakened, significantly improving the depth and spatial resolution of the monitoring. This enables real-time monitoring of tissue oxygen, meeting the precise oxygen saturation monitoring needs of different body parts, including the body surface and muscles, especially deep muscle tissue. It is easily applicable to industrial applications and adaptable to parameter calibration for various monitoring requirements, demonstrating strong practicality and adaptability. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1This is a system module diagram of a real-time tissue oxygen monitoring system based on multispectral fusion according to the present invention; Figure 2 This is a flowchart of the steps of a real-time tissue oxygen monitoring method based on multispectral fusion according to the present invention. Detailed Implementation

[0021] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments. Example 1

[0022] like Figure 1 As shown in the embodiment of the present invention, a real-time tissue oxygen monitoring system based on multispectral fusion includes: Spectral signal acquisition module: Acquires near-infrared reflectance spectral signals of at least three wavelengths from the tissue to be monitored; In the spectral signal acquisition module, the specific process of acquiring near-infrared reflectance spectral signals of at least three wavelengths of the tissue to be monitored is as follows: at least three different wavelengths of near-infrared light are emitted sequentially or simultaneously to the tissue to be monitored. The three different wavelengths of near-infrared light include 730nm (corresponding to superficial tissue), 808nm (corresponding to transitional tissue), and 940nm (corresponding to deep muscle tissue). Near-infrared light is incident on the tissue to be monitored. After being scattered and absorbed by different depth layers such as the epidermis, dermis and muscle, it forms reflected light carrying tissue oxygen information. The reflected light signal carrying tissue oxygen information is received by a photodetector and converted into an analog electrical signal. The analog electrical signal is sequentially filtered, amplified, and converted from analog to digital to obtain digitized multi-wavelength near-infrared reflectance spectral data. Simultaneously, the signal-to-noise ratio (SNR) and validity of the multi-wavelength near-infrared reflectance spectral data are verified, and invalid data with an SNR below a preset threshold or abnormal light intensity are discarded. The verified reflectance spectral signals of at least three wavelengths can be used as input data for constructing the spatial weight matrix.

[0023] Weight matrix optimization module: Based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored, calculate the oxygen contribution weights of the three wavelength signals to tissues at different depth levels, generate a spatial weight matrix, verify the calculation accuracy of the spatial weight matrix, construct a spatial weight matrix optimization model, and optimize spatial weight matrices that do not meet the accuracy verification standards. In the weight matrix optimization module, the specific process of calculating the oxygen contribution weights of the three wavelength signals to tissues at different depth levels based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored is as follows: Based on the tissue structure of the tissue to be monitored, the monitored tissue is divided into three depth levels along the depth direction: superficial tissue, transitional tissue, and deep muscle tissue, which correspond one-to-one with the three wavelengths of 730nm, 808nm, and 940nm in the spectral signal acquisition module. Subsequently, the scattering coefficients of three wavelengths, 730nm, 808nm, and 940nm, were obtained in the superficial tissue, transitional tissue, and deep muscle tissue, respectively. First, the scattering coefficients of each wavelength in all three depth levels were summed. Then, the scattering coefficient of that wavelength in a single depth level was divided by the sum of the corresponding scattering coefficients to obtain the contribution weight of each wavelength to the oxygen signal in each depth level of the tissue. For example, taking a 730nm wavelength as an example, the scattering coefficients of this wavelength at three depth levels—superficial tissue, transitional tissue, and deep muscle tissue—are summed to obtain the total scattering coefficient corresponding to this wavelength. Then, the scattering coefficient of the 730nm wavelength in the superficial tissue is divided by the total scattering coefficient to obtain the contribution weight of the 730nm wavelength to the superficial tissue; the scattering coefficient of the 730nm wavelength in the transitional tissue is divided by the total scattering coefficient to obtain the contribution weight of the 730nm wavelength to the transitional tissue; and the scattering coefficient of the 730nm wavelength in the deep muscle tissue is divided by the total scattering coefficient to obtain the contribution weight of the 730nm wavelength to the deep muscle tissue. Following the same method, the contribution weights of the 808nm and 940nm wavelengths to the superficial tissue, transitional tissue, and deep muscle tissue, respectively, are calculated. It should be noted that the scattering coefficient is obtained by querying existing biological tissue optical databases, measuring in vitro tissue experiments, or optical simulation. These are known methods for obtaining tissue optical parameters in the field and are only used to provide a data basis for calculating the contribution weight of this invention.

[0024] In the weight matrix optimization module, the specific process of generating a spatial weight matrix and verifying its accuracy is as follows: The contribution weights of the three wavelengths 730nm, 808nm, and 940nm corresponding to the superficial tissue, transitional tissue, and deep muscle tissue are arranged in a regularized manner according to the wavelength dimension and the depth dimension to form a two-dimensional spatial weight matrix composed of wavelength and depth. For example, the three sets of contribution weights corresponding to the superficial tissue, the transitional tissue, and the deep muscle tissue at the 730nm wavelength are arranged as the first row, the three sets of contribution weights corresponding to the three depth levels at the 808nm wavelength are arranged as the second row, and the three sets of contribution weights corresponding to the three depth levels at the 940nm wavelength are arranged as the third row. The weights are arranged in order with the rows as the wavelength dimension and the columns as the depth dimension to form a two-dimensional spatial weight matrix with three rows and three columns. Using a simulated tissue model or a standard tissue oxygen monitoring device as the true value of the validation benchmark, the tissue oxygen result calculated by the spatial weight matrix is ​​compared with the true value of the validation benchmark to obtain the calculation deviation between the tissue oxygen result and the validation benchmark. If the calculation deviation is less than or equal to the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix meets the usage requirements. If the calculation deviation is greater than the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix does not meet the requirements and the spatial weight matrix needs to be optimized and corrected. It should be noted that the preset deviation threshold is a value set in advance based on the actual monitoring needs, the type of organization to be monitored, and the system accuracy requirements. It can be determined through experimental calibration, empirical values, or adaptive algorithms. It should also be noted that the simulated tissue model (such as a phantom tissue model based on optical parameter matching) and the standard tissue oxygen monitoring equipment (such as a medical-grade near-infrared tissue oxygen detection device) are conventional tissue oxygen calibration and verification carriers in the field, which can provide stable and reliable tissue oxygen true data and are only used for the accuracy analysis and precision verification of the spatial weight matrix.

[0025] In the weight matrix optimization module, the specific process of constructing a spatial weight matrix optimization model and optimizing spatial weight matrices that fail accuracy verification is as follows: The spatial weight matrices that meet the accuracy standards and those that do not meet the accuracy standards during historical detection processes are collected to form the model training sample set; each training sample contains corresponding scattering coefficient data, spatial weight matrix data, verification benchmark true value, and calculated deviation data; Using a BP neural network as the basic network model, the BP neural network is trained with the training sample set of the model, so that the network learns the mapping relationship between the scattering coefficient, spatial weight matrix, calculation bias and the optimal tissue oxygen signal contribution weight, and obtains the trained spatial weight matrix optimization model. The spatial weight matrix whose accuracy does not meet the requirements, the corresponding scattering coefficient data, and the calculation deviation are input into the pre-trained spatial weight matrix optimization model. The spatial weight matrix optimization model iteratively corrects the tissue oxygen signal contribution weights corresponding to each wavelength and each depth level, and outputs the corrected optimal tissue oxygen contribution weights. The spatial weight matrix is ​​regenerated based on the corrected optimal contribution weights, and the accuracy analysis and deviation comparison are performed again on the regenerated spatial weight matrix. The iterative correction and accuracy verification steps are repeated until the calculation deviation is less than or equal to the preset deviation threshold, thus completing the optimization and correction of the spatial weight matrix.

[0026] Signal fusion and interference removal module: The near-infrared reflectance spectrum signal is weighted and superimposed using a spatial weight matrix that has been verified by calculation accuracy, so as to eliminate the interference of superficial tissues on the oxygen signals of transitional tissues and deep muscle tissues, and form a fused total tissue oxygen signal. In the signal fusion and interference removal module, the near-infrared reflectance spectral signals are weighted and superimposed using a spatial weight matrix that has been verified by computational accuracy. This eliminates the interference of superficial tissues on the oxygen signals of transitional and deep muscle tissues, forming the fused total tissue oxygen signal. The specific process is as follows: The near-infrared reflectance spectral signal acquired in the spectral signal acquisition module is called, and the near-infrared reflectance spectral signal is divided into layers according to the wavelength dimension and the depth dimension to obtain the spectral signal component corresponding to the shallow tissue at a wavelength of 730nm, the spectral signal component corresponding to the transition tissue at a wavelength of 808nm, and the spectral signal component corresponding to the deep muscle tissue at a wavelength of 940nm. Based on the spatial weight matrix optimized and verified in the weight matrix optimization module, the optimal contribution weights corresponding to 730nm / shallow tissue, 808nm / transition tissue, and 940nm / deep muscle tissue are extracted respectively. The spectral signal component of each layer is multiplied element-wise with the optimal contribution weight of the corresponding wavelength and the corresponding depth layer to obtain the weighted spectral signal at each wavelength-depth layer. All weighted spectral signals are linearly superimposed and summed to form a fused total tissue oxygen signal. This weighted superposition method enhances the effective oxygen signal components of the transition layer and deep muscle tissue in the superposition result, while weakening the strong scattering interference components from the superficial tissue, thereby effectively eliminating interference from the superficial tissue.

[0027] Tissue oxygen saturation calculation module: Based on the fused total tissue oxygen signal, calculate and output tissue oxygen saturation.

[0028] In the tissue oxygen saturation calculation module, the specific process of calculating and outputting tissue oxygen saturation based on the fused total tissue oxygen signal is as follows: Based on the fused total tissue oxygen signal obtained by the signal fusion and de-interference module, the light absorption intensity information corresponding to the three wavelengths of 730nm, 808nm, and 940nm is extracted respectively. According to the Lambert-Beer law and the principle of near-infrared spectroscopy for tissue oxygen detection, the correspondence between the light absorption intensity of the above three wavelengths and the concentrations of oxyhemoglobin and deoxyhemoglobin is established. The fused total tissue oxygen signal is analyzed and calculated through the correspondence to obtain the relative content ratio of oxyhemoglobin to deoxyhemoglobin in the tissue to be tested, and the final tissue oxygen saturation is further calculated based on the relative content ratio. The calculated tissue oxygen saturation is then output and displayed.

[0029] For example, according to the Lambert-Beer law, the light absorption intensity of the test tissue to near-infrared light is positively correlated with the concentrations of oxyhemoglobin and deoxyhemoglobin and the optical path length. For three wavelengths, 730nm, 808nm, and 940nm, quantitative correspondences between light absorption intensity and the concentrations of oxyhemoglobin and deoxyhemoglobin at each wavelength are established. Under the premise of known optical path length and molar absorptivity, the actual content of oxyhemoglobin and deoxyhemoglobin in the test tissue is obtained by inversion calculation through the light absorption intensity in the fused tissue oxygen total signal. Then, the ratio of oxyhemoglobin to deoxyhemoglobin content is normalized to finally obtain the tissue oxygen saturation used to reflect the oxygen state of the test tissue.

[0030] Example 2 like Figure 2 As shown, based on the specific implementation process of Example 1, the present invention provides a method for real-time monitoring of tissue oxygen based on multispectral fusion, including: Step 1: Acquire near-infrared reflectance spectral signals of at least three wavelengths from the tissue to be monitored; Step 2: Based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored, calculate the oxygen contribution weights of the three wavelength signals to tissues at different depth levels, generate a spatial weight matrix, verify the accuracy of the spatial weight matrix calculation, construct an optimization model for the spatial weight matrix, and optimize the spatial weight matrix that does not meet the accuracy verification. Step 3: Use the spatial weight matrix verified by calculation accuracy to perform weighted superposition of near-infrared reflectance spectral signals to eliminate the interference of superficial tissues on the oxygen signals of transitional tissues and deep muscle tissues, forming a fused total tissue oxygen signal; Step 4: Calculate the output tissue oxygen saturation based on the fused total tissue oxygen signal.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time tissue oxygen monitoring system based on multispectral fusion, characterized in that: include: Spectral signal acquisition module: Acquires near-infrared reflectance spectral signals of at least three wavelengths from the tissue to be monitored; Weight matrix optimization module: Based on the differences in the scattering coefficients of the three wavelengths in the tissue to be monitored, calculate the oxygen contribution weights of the three wavelength signals to tissues at different depth levels, generate a spatial weight matrix, verify the calculation accuracy of the spatial weight matrix, construct a spatial weight matrix optimization model, and optimize spatial weight matrices that do not meet the accuracy verification standards. Signal fusion and interference removal module: The near-infrared reflectance spectrum signal is weighted and superimposed using a spatial weight matrix that has been verified by calculation accuracy, so as to eliminate the interference of superficial tissues on the oxygen signals of transitional tissues and deep muscle tissues, and form a fused total tissue oxygen signal. Tissue oxygen saturation calculation module: Based on the fused total tissue oxygen signal, calculate and output tissue oxygen saturation.

2. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for obtaining the near-infrared reflectance spectral signals of the at least three wavelengths is as follows: At least three different wavelengths of near-infrared light are emitted sequentially or simultaneously to the tissue to be monitored. After the near-infrared light is incident on the tissue to be monitored, the reflected light carrying tissue oxygen information is converted into an analog electrical signal. The analog electrical signal is filtered, amplified, and converted from analog to digital to obtain digitized multi-wavelength near-infrared reflectance spectral data. The signal-to-noise ratio and validity of the multi-wavelength near-infrared reflectance spectral data are verified to obtain near-infrared reflectance spectral signals of at least three wavelengths.

3. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 2, characterized in that, The three wavelengths of near-infrared reflectance spectral signals include: The three different wavelengths of near-infrared light include 730nm for superficial tissues, 808nm for transitional tissues, and 940nm for deep muscle tissues.

4. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The process for calculating the tissue oxygen contribution weight is as follows: The scattering coefficients of each wavelength at the three depth levels of superficial tissue, transitional tissue, and deep muscle tissue are summed, and the scattering coefficient of each wavelength at a single depth level is divided by the sum of the scattering coefficients of the corresponding wavelength at all depth levels to form the tissue oxygen signal contribution weight of the corresponding depth level.

5. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for generating the generated spatial weight matrix is ​​as follows: The contribution weights of superficial tissue, transitional tissue, and deep muscle tissue corresponding to 730nm, 808nm, and 940nm are arranged in a regularized manner according to the wavelength dimension and the depth dimension, forming a two-dimensional spatial weight matrix composed of wavelength and depth.

6. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for verifying the accuracy of the spatial weight matrix calculation is as follows: Using a simulated tissue model or a standard tissue oxygen monitoring device as the true value of the validation benchmark, the tissue oxygen result calculated by the spatial weight matrix is ​​compared with the true value of the validation benchmark to obtain the calculation deviation between the tissue oxygen result and the validation benchmark. If the calculation deviation is less than or equal to the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix meets the usage requirements. If the calculation deviation is greater than the preset deviation threshold, it means that the calculation accuracy of the spatial weight matrix does not meet the usage requirements.

7. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process of constructing the spatial weight matrix optimization model is as follows: The spatial weight matrices with and without the required accuracy are collected and used together to form the model training sample set, which includes the corresponding scattering coefficient data, spatial weight matrix data, the verification benchmark true value, and the calculation deviation data. The BP neural network is used as the basic network model, and the BP neural network is trained using the model training sample set to obtain the trained spatial weight matrix optimization model.

8. The real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for optimizing the spatial weight matrix that fails the accuracy verification is as follows: The spatial weight matrix whose current accuracy does not meet the requirements, the corresponding scattering coefficient data, and the calculation deviation are all input into the spatial weight matrix optimization model, and the corrected optimal tissue oxygen contribution weight is output.

9. A real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for forming the fused tissue oxygen total signal is as follows: The acquired near-infrared reflectance spectral signals were layered and distinguished according to wavelength and depth dimensions to obtain the spectral signal components corresponding to shallow tissues at 730nm wavelength, the spectral signal components corresponding to transitional tissues at 808nm wavelength, and the spectral signal components corresponding to deep muscle tissues at 940nm wavelength. Based on the validated spatial weight matrix, the optimal contribution weights corresponding to 730nm / shallow tissue, 808nm / transitional tissue, and 940nm / deep muscle tissue were extracted and multiplied element-wise to obtain weighted spectral signals. All weighted spectral signals were linearly superimposed and summed to form the fused total tissue oxygen signal.

10. A real-time tissue oxygen monitoring system based on multispectral fusion according to claim 1, characterized in that, The specific process for calculating the output tissue oxygen saturation is as follows: Based on the fused total tissue oxygen signal, the light absorption intensity information corresponding to the three wavelengths of 730nm, 808nm, and 940nm was extracted respectively. According to the Lambert-Beer law and the principle of near-infrared spectroscopy for tissue oxygen detection, the correspondence between light absorption intensity and the concentrations of oxyhemoglobin and deoxyhemoglobin was established. The relative content ratio of oxyhemoglobin to deoxyhemoglobin in the tissue to be tested was obtained, and the final tissue oxygen saturation was calculated from the relative content ratio.

Images