Multi-angle reflection type PPG non-invasive cerebral venous sinus blood oxygen measurement method and device

Abstract

Multi-angle reflection type PPG non-invasive cerebral venous sinus blood oxygen measurement method and device, wherein the blood oxygen measurement method comprises the following steps: S1, the wavelength incidence angle and the light source-probe interval predetermined optical probe is pasted on the occipital region of the measured person, in at least including the calibration period and the monitoring period multiple period, drive the multi-wavelength light source in the optical probe to emit detection light at different incidence angles, and the optical reflection signal of the shallow tissue and the deep blood flow is collected by multiple photodetectors simultaneously;S2, the optical reflection signal is interfered and inhibited, the optical reflection signal collected in the calibration period is matched and fitted with the pre-established tissue optical parameter model library, and the individualized skull parameter set of the occipital region of the measured person is estimated based on the optimization algorithm. The blood oxygen measurement method and device can continuously, real-timely and reliably reflect the monitoring of the cerebral venous sinus blood oxygen state under non-invasive conditions.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, specifically to a multi-angle reflective PPG non-invasive method and device for measuring intracranial venous sinus blood oxygenation. Background Technology

[0002] The brain is one of the most metabolically active organs in the human body, consuming approximately 20%–25% of the body's total oxygen, and is highly sensitive to hypoxia. Studies have shown that a 4–6 minute interruption of cerebral blood flow can lead to irreversible neuronal damage. Therefore, continuous, real-time, and reliable monitoring of brain oxygenation is crucial for assessing surgical outcomes, guiding treatment decisions, providing early warning of ischemic and hypoxic events, and improving patient prognosis in neurosurgery, neurointerventional therapy, and the diagnosis and perioperative management of diseases such as moyamoya disease and traumatic brain injury. However, current brain oxygenation monitoring technologies still have significant limitations in clinical application, forming a technical bottleneck that restricts precise neuromonitoring.

[0003] Currently, invasive methods are the most commonly used for cerebral oxygenation monitoring in clinical practice. While arterial blood gas analysis is considered a reference standard for blood oxygenation measurement, it only provides data at discrete time points and cannot reflect the continuous trend of oxygenation status. Furthermore, repeated arterial punctures increase the risks of pain, infection, and hematoma. More importantly, arterial blood gas analysis reflects systemic arterial blood oxygenation levels, which may differ significantly from the actual oxygenation status of local brain tissue. Brain tissue oxygen partial pressure monitoring, achieved through intracranial probe implantation, allows for continuous monitoring. However, this method is highly invasive, carrying risks such as intracranial hemorrhage, infection, and brain tissue damage. Its monitoring range is limited to a small area around the probe, making it difficult to reflect overall or multi-regional brain perfusion. Additionally, the high cost of the equipment limits its clinical application.

[0004] In the field of noninvasive blood oxygen monitoring, measurement methods based on photoplethysmography (PPG) or near-infrared spectroscopy (NIRS) have been widely used. Blood oxygen monitoring of peripheral sites such as fingertips and earlobes usually employs transmission or reflection optical methods, primarily used to reflect the overall oxygenation status of the systemic circulatory system, and is relatively mature in clinical monitoring.

[0005] However, the aforementioned non-invasive blood oxygenation monitoring methods primarily measure peripheral soft tissue blood vessels, and their optical paths do not involve skull structures, making it difficult to reflect the true oxygenation status of blood supply vessels in the paracranial or cerebral venous sinuses. In applications such as cerebrovascular diseases or neurosurgery, situations often arise where peripheral blood oxygenation levels are normal while local cerebral blood supply has become abnormal; traditional peripheral blood oxygenation monitoring lacks targeted early warning capabilities.

[0006] To achieve non-invasive assessment of the oxygenation status of the cerebral venous sinuses, near-infrared spectroscopy has been introduced into the field of cerebral oxygenation monitoring. However, it faces significant challenges in the presence of the skull. The skull has high scattering and absorption characteristics, and its thickness and structure vary from person to person, leading to severe attenuation of light signals during propagation. Signals from the scalp and extracranial tissues tend to dominate, masking the blood oxygenation information of the target vessels behind the skull. Furthermore, increasing the distance between the light source and the detector to improve the detection depth often introduces a larger area of ​​tissue signal superposition, reducing spatial resolution and exacerbating quantitative errors. Therefore, there is an urgent need for a non-invasive blood oxygenation monitoring method specifically for the cerebral venous sinuses, capable of effectively modeling the light propagation characteristics in the presence of the skull and accurately extracting blood oxygenation signals from the cerebral venous sinuses to meet the practical needs of cerebrovascular disease monitoring and intraoperative cerebral oxygenation assessment.

[0007] To address these issues, academia and industry have proposed various improvement schemes, such as multi-range near-infrared spectroscopy, time-resolved or frequency-resolved near-infrared spectroscopy, diffusion optical tomography, and multimodal joint monitoring. However, these schemes generally suffer from problems such as complex equipment, large size, high cost, large computational load of algorithms, or high difficulty in clinical operation, making it difficult to meet the comprehensive requirements of continuity, real-time performance, portability, and ease of use in perioperative and intensive care. Summary of the Invention

[0008] The purpose of this invention is to propose a multi-angle reflective PPG non-invasive method and device for measuring intracranial venous sinus blood oxygenation, which can continuously, in real time and reliably reflect and monitor the intracranial venous sinus blood oxygenation status under non-invasive conditions.

[0009] The technical solution adopted in this invention is: a multi-angle reflective PPG non-invasive method for measuring intracranial venous sinus blood oxygenation, comprising the following steps: S1, the incident angle of the wavelength and the distance between the light source and the detector The predetermined optical probe is fitted to the skull region behind the ear of the subject. During multiple time periods, including at least a calibration period and a monitoring period, the multi-wavelength light source inside the optical probe is driven to emit detection light at different incident angles. Multiple photodetectors simultaneously collect optical reflection signals from superficial tissues and deep blood flow. ; S2, Optical reflection signal Interference suppression is performed, and the optical reflection signals collected during the calibration period are matched and fitted with a pre-established tissue optical parameter model library. Based on the optimization algorithm, an individualized set of skull parameters for the postauricular skull region of the subject is estimated in real time. And calculate the comprehensive skull attenuation factor for each channel-wavelength combination. and the average skull penetration efficiency of this channel under multiple wavelengths. Individualized skull parameter set At least including the thickness of the local skull ; For the optical reflection signals collected during the monitoring period, and based on the comprehensive skull attenuation factor Skull attenuation compensation was performed to obtain the compensated optical reflection signal. ; S3. Based on the variational mode decomposition algorithm, analyze the compensated optical reflection signal. Separate the components and obtain the static baseline components of each wavelength in each channel. and dynamic blood flow pulsation components And calculate the blood oxygen ratio. ; S4. Constructing signal quality evaluation indicators And combined with the average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing and local skull thickness Establish a depth contribution to characterize the ability of each channel to detect deep blood flow signals. ; S5, Based on Deep Contribution Blood oxygen ratio Average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing Construct a spatial feature matrix containing skull attenuation features ; S6. The spatial feature matrix The data is input into a pre-trained intelligent correction model, which adaptively increases the feature weights of deep blood flow and suppresses the feature weights of superficial tissues. It extracts and fuses features related to deep blood flow from multi-channel spatiotemporal signals and outputs a predicted value of cerebral venous sinus blood oxygen saturation. ; S7. The predicted blood oxygen saturation value Individualized dynamic correction and smoothing are performed to obtain and output the final cerebral venous sinus blood oxygen saturation value. .

[0010] As a preferred embodiment, the tissue optical parameter model library is established based on bidirectional reflectance distribution function and bidirectional transmission distribution function, combined with the anisotropy and multilayer structure characteristics of human tissue, through in vitro or in vivo calibration measurements.

[0011] As a preferred embodiment, suppressing background interference includes canceling ambient light interference in the optical reflection signal based on an ANC filter; and / or eliminating motion artifacts in the optical reflection signal based on a Kalman filter.

[0012] As a preferred option, the integrated skull attenuation factor Calculated using the following formula: in, The directional transmission function is determined by the tissue optical parameter model library. For the first The estimated skull thickness is obtained from each channel; For the first The wavelength incident angle of each channel; The relative density of the skull. λ is the wavelength.

[0013] As a preferred embodiment, the deep contribution of the skull optical properties is incorporated. Calculated using the following formula: in, Indicates channel Average skull penetration efficiency under dual wavelengths , , They represent the first Each channel has wavelength , The corresponding comprehensive skull attenuation factors; and These are the normalization coefficients; for Signal quality evaluation metrics for each channel , For signal-to-noise ratio, For signal stability, For the consistency of blood flow pulsation , , These are the weighting coefficients, and + + =1; For the first The light source-detector spacing of each channel.

[0014] As a preferred embodiment, the intelligent correction model is a hybrid model of convolutional neural network and long short-term memory network, wherein the convolutional neural network is used to extract the spatial features of multi-channel signals, and the long short-term memory network is used to extract the time series features of signals.

[0015] As a preferred embodiment, in step S6, the individualized dynamic correction is calculated using the following formula: in, The oxygen saturation of the cerebral venous sinus after dynamic correction; The value is the predicted value of cerebral venous sinus oxygen saturation before correction; δ(t) is the dynamic correction factor. It is the estimation of skull thickness through each channel. The standard deviation is used to correct for deviations caused by unevenness in the skull. For reference average thickness; For correction factors; This is a correction coefficient for the contribution of deeper layers; For time Next The deep contribution of each channel; For the first Reference values ​​for the deep contribution of each channel; This is the heart rate variability correction factor; For time The change in heart rate.

[0016] As a preferred embodiment, in step S7, the smoothing process includes moving average processing and outlier removal.

[0017] A multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measuring device, the measuring device comprising: An optical probe has an asymmetric fitting structure that matches the curvature of the cranial region behind the ear, and the optical probe integrates a light source array and a detector array. A light source array for emitting detection light of at least two wavelengths, and for at least one wavelength, multiple light sources are arranged at different incident angles; A detector array, arranged in a ring or array form at a predetermined distance from the light source array, is used to receive reflected light signals; The signal acquisition and processing unit is connected to the detector array and is used to condition the reflected light signal, perform analog-to-digital conversion, and execute any of the steps of the multi-angle reflective PPG non-invasive cranial venous sinus blood oxygenation measurement method. A storage unit is used to store the preset tissue optical parameter model library and the pre-trained intelligent correction model; The power supply and communication unit is used to supply power and transmit data to the optical probe and signal acquisition and processing unit.

[0018] As a preferred embodiment, the signal acquisition and processing unit is further configured as follows: During the calibration period, the light source array is controlled to perform multi-wavelength, multi-angle optical scanning, and the acquired baseline optical signals are matched and fitted with the tissue optical parameter model library in the storage unit to estimate the individualized skull optical parameters of the current subject online. ; During the continuous monitoring phase, the individualized skull optical parameters are used to compensate for skull attenuation in the real-time acquired signals, and the compensated signals are input into the pre-trained intelligent correction model.

[0019] Compared with the prior art, the beneficial effects of the present invention are: 1. Through multi-angle optical incident and receiving structures, optimized wavelength combination design, and signal processing algorithms for low perfusion and strong scattering environments, continuous and real-time monitoring of the oxygenation status of the cerebral venous sinuses can be achieved without arterial puncture or intracranial probe implantation. This effectively avoids the risks of infection and bleeding associated with invasive monitoring, reduces the burden on patients, and expands the applicable population.

[0020] 2. By employing a multi-channel, multi-angle optical measurement and deep learning signal fusion strategy, an end-to-end deep neural network model is introduced on the basis of traditional multi-channel signal separation. Utilizing the spatiotemporal characteristics of the optical signals from each channel, incident angle information, and individualized skull parameters, adaptive suppression of interference signals in the scalp and superficial skull is achieved, while simultaneously enhancing the proportion of blood oxygenation signals in the brain's microcirculation. This method not only improves the specificity and absolute accuracy of blood oxygenation measurement but also maintains stable and continuous monitoring performance under complex environments such as low perfusion or head micro-movement through adaptive model adjustment, providing direct support for subsequent deep learning-based algorithm design.

[0021] 3. A complete hardware system and real-time processing framework suitable for long-term continuous monitoring have been constructed, achieving high temporal resolution data output. This system can dynamically reflect changes in cerebral oxygenation and promptly identify abnormal events such as transient ischemic attacks. By balancing wearing comfort, ease of operation, and system cost control, this device has broad application potential in clinical scenarios such as neurosurgery, intensive care, and cerebrovascular disease management, providing key and objective physiological parameter support for precision medicine and personalized treatment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the application parts of the multi-angle optical measurement component and optical probe in this invention; Figure 2 This is a flowchart illustrating the control and processing module in this invention; Figure 3This is a schematic diagram of the blood oxygen measurement device in this invention; Figure 4 This is a schematic diagram of the architecture of the intelligent correction model in this invention; Figure 5 This is a schematic flowchart of the blood oxygen measurement method in this invention; Figure 6 This is a reflectance contour map of a whiteboard under incident conditions of 0°, -15°, -30°, and -45° in an embodiment of the present invention; wherein: (a) is a reflectance contour map under 0° incident conditions; (b) is a reflectance contour map under an incident angle of -15°. (c) is a reflectance contour map under an incident angle of -30°. (d) is a reflectance contour map under an incident angle of -45°. Figure 7 This is a reflectance contour map of bovine bone under incident conditions of 0°, -15°, -30°, and -45° in an embodiment of the present invention; wherein: (a) is a reflectance contour map under 0° incident conditions; (b) is a reflectance contour map under an incident angle of -15°. (c) is a reflectance contour map under an incident angle of -30°. (d) is a reflectance contour map under an incident angle of -45°. Figure 8 This is a two-dimensional angle-reflectivity curve of a whiteboard and cow bone in an embodiment of the present invention; wherein: (a) is a two-dimensional angle-reflectivity curve of the whiteboard; (b) is a two-dimensional angle-reflectivity curve of bovine bone; Figure 9 This is a three-dimensional surface plot of the reflectance of the whiteboard under incident conditions of 0°, -15°, -30°, and -45° in an embodiment of the present invention; wherein: (a) is a three-dimensional surface plot of reflectivity under 0° incident conditions; (b) is a three-dimensional surface plot of reflectance under an incident angle of -15°; (c) is a three-dimensional surface plot of reflectivity under an incident angle of -30°; (d) is a three-dimensional surface plot of reflectance under an incident angle of -45°. Figure 10 This is a three-dimensional surface plot of the reflectance of bovine bone under incident conditions of 0°, -15°, -30°, and -45°, as shown in this embodiment of the invention; wherein: (a) is a three-dimensional surface plot of reflectivity under 0° incident conditions; (b) is a three-dimensional surface plot of reflectance under an incident angle of -15°; (c) is a three-dimensional surface plot of reflectivity under an incident angle of -30°; (d) is a three-dimensional surface plot of reflectivity under an incident angle of -45°. Detailed Implementation

[0024] The present invention will now be described in detail through exemplary embodiments. However, it should be understood that, without further description, elements, structures, and features in one embodiment may be advantageously incorporated into other embodiments.

[0025] It should be noted that, unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "a," "an," or "the," etc., used in the specification and claims of this patent application do not express a limitation on quantity, but rather indicate the presence of at least one; the terms "first," "second," and "third," as used herein, should not be considered as a limitation on the order of components, but are merely for distinguishing different components; the terms "comprising," "including," etc., indicate that the elements or objects preceding "comprising" or "including" encompass the elements or objects listed following "comprising" or "including" and their equivalents, but do not exclude other elements or objects having the same function.

[0026] To more clearly describe this multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method and device, combined with the attached... Figures 1-10 This embodiment is described as follows: like Figures 1-4 As shown, this invention discloses a multi-angle reflective PPG non-invasive cranial venous sinus blood oxygenation measurement device. This device achieves stable, high-precision, and continuous blood oxygenation signal acquisition and processing through modular design. The overall system mainly comprises five functional modules: a dedicated probe for the skull behind the ear (optical probe), a signal acquisition and processing unit (including a signal acquisition and conditioning module, a control and processing module), a power supply module, and a communication and display module. These modules work synergistically through a rational layout and signal flow design, thereby meeting the continuous blood oxygenation monitoring requirements under conditions of low perfusion, complex bony structures, and minimal displacement during wear.

[0027] 1. Dedicated measurement probe for the skull behind the ear (optical probe) The dedicated probe for the postauricular skull is the core measurement unit of this device. Its design fully considers the complex anatomical curvature and high proportion of bony tissue in the postauricular skull, distinguishing it from traditional blood oxygen monitoring devices that clamp the auricle or earlobe. The probe adopts an asymmetrical fit structure, with the inner fitting surface customized according to the curvature of the postauricular skull to avoid the auricle and external auditory canal. At the same time, the optimized center of gravity improves wearing stability, ensuring that the probe will not be significantly displaced due to slight head movements during intraoperative, postoperative, and long-term continuous monitoring.

[0028] The optical probe integrates a multi-angle optical measurement component, which includes multiple light-emitting units. Each unit integrates one LED light-emitting element and one PD photodetector element. The LED and PD elements each form a light source array and a detector array, respectively. The light source array emits detection light of at least two wavelengths, and for at least one wavelength, multiple light sources are arranged at different incident angles. The detector array is arranged in a ring or array at a predetermined distance from the light source array to receive reflected light signals. The light-emitting units are arranged at different incident angles and spacings to form multi-angle, multi-depth optical paths, enabling simultaneous acquisition of superficial tissue and deep subskull blood flow signals. The probe also features an elastic fit and fixation structure, allowing for fine-tuning of the optical component position when the wearer's head moves slightly, maintaining stable contact between the optical window and the skin, effectively suppressing motion artifacts, and enhancing the reliability of deep blood oxygenation signals. A schematic diagram of the multi-angle optical measurement component and probe application area is shown below. Figure 1 As shown.

[0029] A non-invasive, continuous, real-time, and reliable monitoring technology for the oxygenation status of the cerebral venous sinuses requires an ideal solution that avoids invasive procedures while optimizing the design based on the optical characteristics of the skull. This involves improving the specificity and stability of intracranial microvascular oxygenation signals through appropriate measurement site selection and optical path construction. This invention is proposed against this backdrop, providing a new technical path for practical non-invasive measurement of cerebral venous sinuses through the design of an optical measurement structure and method targeting the postauricular skull region. Specifically, the postauricular skull region was chosen for measurement primarily based on the following considerations: First, this region has the most developed venous sinuses, capable of collecting venous blood from the ipsilateral half of the brain, making it highly representative; second, this anatomical location is relatively fixed, with relatively little interference from superficial muscles and soft tissues; and more importantly, the adjacent venous sinuses in this region are located at the proximal end (downstream) of intracranial venous blood flow, and their oxygen saturation can integrate and reflect the metabolic status of multiple upstream brain regions, thus more reliably assessing the overall oxygenation and metabolic status of the brain tissue. Despite the complex skull structure and severe signal attenuation in this region, the oxygenation information of its venous sinus is of irreplaceable value for assessing the oxygenation and metabolic status of ipsilateral and even the whole brain tissue.

[0030] 2. Signal Acquisition and Conditioning Module The signal acquisition and conditioning module is responsible for converting the optical signals acquired by the probe into processable digital signals, ensuring high fidelity and high dynamic range. This module includes a photoelectric sensor interface, amplifier, filtering circuit, and analog-to-digital converter (A / D) module. The amplifier enhances the amplitude of weak light signals, the filtering circuit suppresses ambient light interference and high-frequency noise, and the A / D module digitizes the analog signals for accurate analysis by subsequent control and processing modules. The signal acquisition module also supports multi-channel synchronous acquisition, ensuring time synchronization between optical paths and guaranteeing the consistency and comparability of multi-angle measurement data. The implementation process is as follows: Figure 2 As shown.

[0031] 3. Control and Processing Module The control and processing module is the core computing unit of the system, mainly composed of a high-performance microcontroller (MCU) and embedded signal processing algorithms. The module inputs multi-channel digital optical signals from the signal acquisition module. Through a multi-channel fusion algorithm, it specifically extracts superficial tissue signals and deep blood flow signals, and combines this with a deep learning-driven intelligent correction model to dynamically compensate for signal drift caused by skull attenuation and individual differences, achieving high-precision calculation of blood oxygen saturation. Simultaneously, the module can evaluate the signal quality of each channel in real time and correct or remove low-quality or abnormal signals, thereby ensuring the reliability and continuity of the output data. This module outputs blood oxygen saturation results and signal quality indicators, providing directly referable data for clinical use. The implementation flowchart is as follows: Figure 3 As shown.

[0032] 4. Power Module The power module employs a rechargeable battery design, providing continuous and stable power to modules such as the probe, signal acquisition, control processing, and communication display. The power module features a power monitoring function, alerting the user when the battery is low, and supports flexible charging methods, ensuring reliable power supply during extended continuous monitoring.

[0033] 5. Communication and display module The communication and display module is responsible for system data transmission, visualization, and clinical early warning functions. The blood oxygenation results and signal quality information calculated by the control and processing module are transmitted wirelessly or via wired connection to external display devices, such as monitors, computers, or mobile terminals, enabling real-time data display and historical data recording. The module can also be configured with alarm thresholds; when the blood oxygenation level falls below a set value, a clinical early warning is triggered, providing immediate reference for intraoperative, postoperative, and neurocritical care.

[0034] 6. Storage unit Used to store a library of pre-set tissue optical parameter models and pre-trained intelligent correction models.

[0035] The device and system of this invention achieve high-precision non-invasive blood oxygen monitoring in the postauricular skull region through the coordinated operation of the above modules. The probe module is responsible for signal acquisition and ensuring physical fit stability; the signal acquisition and conditioning module ensures high-fidelity data; the control and processing module realizes multi-channel fusion, deep learning correction, and signal quality control; the power supply module provides continuous power supply; and the communication and display module completes real-time data output and alarm. The modules are closely linked through signal flow and control logic to form a closed-loop system. The overall system design takes into account the enhancement of deep blood oxygen information ratio, suppression of motion artifacts, adaptation to low perfusion environments, continuous real-time monitoring, and wearing comfort, which is significantly different from traditional blood oxygen monitoring devices that measure blood oxygen at the fingertips, wrists, or clipped to the ear, and has specialization, reliability, and irreplaceability.

[0036] The above-described system hardware configuration and design provide a complete and feasible technical solution for non-invasive blood oxygenation monitoring in the postauricular cranial region, laying a solid foundation for subsequent signal processing and blood oxygenation calculation algorithms, and providing reliable hardware support for continuous cerebral oxygenation monitoring in clinical practice.

[0037] like Figure 5 As shown, this invention also discloses a multi-angle reflective PPG non-invasive method for measuring intracranial venous sinus blood oxygenation. Based on a multi-channel, multi-angle hardware system of a blood oxygenation measurement device, it realizes the algorithm part of the technical framework of "multi-angle acquisition - multi-channel modeling - intelligent fusion - individualized correction," constructing a complete process of "preprocessing - spatial feature extraction - model fitting - dynamic correction - output smoothing." By deeply mining the spatial dimension information and time series features of multi-channel signals, and introducing individualized modeling and compensation technology for the optical characteristics of the skull, it accurately separates superficial interference signals from the scalp and skull from deep brain tissue blood oxygenation signals, achieving high-precision and continuous calculation of intracranial venous sinus blood oxygenation. The following will use a preferred embodiment based on a CNN-LSTM hybrid deep learning model to explain in detail the specific implementation of key steps such as "model fitting" and "individualized correction" in this framework.

[0038] Blood oxygen measurement methods include the following steps: S1, the incident angle of the wavelength and the distance between the light source and the detector The predetermined optical probe is fitted to the skull region behind the ear of the subject. During multiple time periods, including at least a calibration period and a monitoring period, the multi-wavelength light source inside the optical probe is driven to emit detection light at different incident angles. Multiple photoelectric detectors simultaneously collect optical reflection signals from superficial tissues and deep blood flow. ; S2. To address the strong scattering in the skull region behind the ear, motion artifacts caused by head micro-movements, ambient light interference, and severe signal attenuation due to the skull itself with significant individual differences, a cascaded scheme of "Adaptive Noise Cancellation (ANC) + Kalman Filtering (KF) + Online Skull Parameter Estimation" is adopted to perform multi-stage preprocessing on the raw PPG signals acquired from each channel, laying the foundation for subsequent feature extraction and signal fusion.

[0039] S2.1, Optical reflection signal Interference suppression is performed, including ambient light and motion artifact suppression, as well as motion artifact removal: Ambient light cancellation: Using the reference light sensor signal, ambient light interference is cancelled in real time through an ANC filter.

[0040] Motion artifact removal: Based on the PPG signal state space model, KF dynamic estimation is used to remove motion artifact components.

[0041] S2.2 Modeling and Parameter Estimation of Skull Transmission and Scattering Characteristics In human tissues, both light transmission and scattering exhibit distinct spatial distribution characteristics, making the accurate acquisition of the tissue's bidirectional reflectance distribution function (BRDF) and bidirectional transmission distribution function (BTDF) crucial. Combining the anisotropy and multilayered structural characteristics of human tissues, a tissue optical parameter model library (BRDF / BTDF parametric model database) of skull tissues is established through in vitro or in vivo calibration measurements. Based on this library, individualized skull parameter sets for the target patient can be rapidly estimated. .

[0042] During the initial wear of the device or during periodic calibration, the control light source array performs a rapid multi-wavelength operation. Multiple angles Optical scanning, and based on the acquired baseline optical signals. The individualized skull optical parameters of the current subject are estimated online by matching and fitting the model library of tissue optical parameters in the storage unit and using optimization algorithms (such as least squares method). .

[0043] in, This refers to the thickness of the local skull. The relative density of the skull; and These are the wavelength-dependent scattering and absorption coefficients, respectively. Indicates the channel number. .

[0044] Furthermore, the comprehensive skull attenuation factor for each channel-wavelength combination is calculated. This factor quantifies the efficiency with which light signals penetrate the skull of a specific individual: in, The directional transmission function is determined by the tissue optical parameter model library. For the first The estimated skull thickness is obtained from each channel; For the first The wavelength incident angle of each channel; The relative density of the skull. λ is the wavelength.

[0045] Average skull penetration efficiency The calculation is as follows: in, , They represent the first Each channel has wavelength , The corresponding comprehensive skull attenuation factors.

[0046] After pretreatment, based on the comprehensive skull attenuation factor Optical reflection signals collected during the monitoring period Skull attenuation compensation was performed to obtain the compensated optical reflection signal. The original signal is about to be sent. Corrected to be closer to the ideal tissue signal : in: For the estimated artifact signal, For the first Channels, wavelengths Optical transmission efficiency coefficient under the following conditions For the first Channels, wavelengths static baseline components under For the first Channels, wavelengths Dynamic blood flow pulsation components, This is the corrected noise term.

[0047] S3. Using Variational Mode Decomposition (VMD) algorithm to replace traditional differential or filtering methods for the compensated optical reflection signal. Separate the components and obtain the static baseline components of each wavelength in each channel. and dynamic blood flow pulsation components And calculate the blood oxygen ratio. ; VMD determines the center frequency and bandwidth of modal components through iterative optimization, avoiding static baseline drift and dynamic component distortion. The separation formula is as follows: in: This is the VMD penalty factor (value range 1000-5000, adaptively adjusted according to the signal noise level, the higher the noise, the larger the value). The modal decomposition order is fixed at 2, corresponding to the static and dynamic components respectively. The low-frequency modes after VMD decomposition (corresponding to the static baseline components) ), The high-frequency modes after VMD decomposition (corresponding to dynamic blood flow pulsation components) ) This is the function for calculating the mean.

[0048] The blood oxygen ratio is corrected based on the wavelength characteristics of the skull: This formula uses the estimated wavelength-dependent attenuation factor ratio to directly compensate for the wavelength difference attenuation caused by the skull.

[0049] S4. Constructing signal quality evaluation indicators Combined with average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing and local skull thickness Establish a depth contribution to characterize the ability of each channel to detect deep blood flow signals. ; S4.1 Construct signal quality evaluation metrics that include signal-to-noise ratio, stability, and ripple consistency. It replaces the traditional single signal-to-noise ratio evaluation and comprehensively reflects signal reliability: in: , , Weighting coefficients (satisfying) + + =1); For the first Signal-to-noise ratio of each channel ( rms is the root mean square function, normalized to [0,1]). Signal stability coefficient ( , (The coefficient of variation reflects the degree of static baseline drift and is normalized to [0,1]). The coefficient of consistency of blood flow pulsation ( , The average reference signal for all channel AC components reflects the consistency of the pulsating signal and is normalized to [0,1].

[0050] S4.2, Define the deep contribution of the fusion of cranial physical properties. : in, Indicates channel Average skull penetration efficiency under dual wavelengths; and These are the normalization coefficients; for Signal quality evaluation metrics for each channel; For the first The light source-detector spacing of each channel.

[0051] S5, Based on Deep Contribution Blood oxygen ratio Average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing Construct a spatial feature matrix containing skull attenuation features ; S6. Spatial feature matrix The data is input into a pre-trained intelligent correction model, which adaptively increases the feature weights of deep blood flow and suppresses the feature weights of superficial tissues. It extracts and fuses features related to deep blood flow from multi-channel spatiotemporal signals, and outputs a predicted value of cerebral venous sinus blood oxygen saturation. .

[0052] To achieve intelligent fusion and correction within the technical framework, this invention provides a preferred embodiment. This embodiment introduces a CNN-LSTM hybrid deep learning model as the model fitting algorithm, enabling the synergistic fusion of spatial and temporal features. It adaptively learns the mapping relationship between individual skull differences, superficial tissue interference, and blood oxygenation signals. Simultaneously, an individualized dynamic correction step is designed to fine-tune the model output based on physiological parameters. This embodiment effectively improves the specificity and accuracy of blood oxygenation calculation.

[0053] S6.1 Model Structure Design The model consists of three core modules and adopts an end-to-end architecture, with the specific structure as follows: 1. Input layer: The input data is a spatiotemporal fusion matrix. , dimension ,in: Time window length (30 sampling points, corresponding to 0.3s at a sampling frequency of 100Hz, balancing real-time performance and stability). : Number of optical channels (consistent with the number of channels in the hardware system, ranging from 4 to 8); Feature dimensions of each channel ( ).

[0054] 2. CNN Spatial Feature Extraction Module: This module is responsible for capturing the spatial correlation between multiple optical channels, especially learning the skull penetration efficiency factor. How does it affect other traits (e.g.) , This allows for adaptive weighting and fusion at the feature level.

[0055] Conv1D Layer 1: 32 kernels, kernel size 3, ReLU activation function, output dimension... ; MaxPool1D layer: pooling kernel size 2, stride 1, output dimension... (Reducing computational load while preserving key features); Conv1D layer 2: 64 convolutional kernels, kernel size 3, activation function ReLU, output dimension... .

[0056] 3. LSTM Time Series Modeling Module: Employs a single bidirectional LSTM layer (Bi-LSTM) to capture the temporal dynamics of blood oxygenation signals (such as rapid fluctuations during transient cerebral ischemia). Hidden layer units: 128; Dropout rate: 0.2 (to prevent overfitting); Output dimension: (The forward and backward outputs of a bidirectional LSTM are spliced ​​together).

[0057] 4. Fully Connected Output Module: Contains two fully connected layers (Dense), outputting preliminary predicted values ​​of cerebral venous sinus blood oxygen saturation. Dense layer 1: 64 units, ReLU activation function, output dimension... ; Dense layer 2: 1 unit, activation function Sigmoid (normalizes the output to [0.7, 1.0], corresponding to a clinically effective range of 70%-100% oxygen saturation), output dimension... Preliminary predicted value of cerebral venous sinus blood oxygenation .

[0058] Model structure as follows Figure 4 As shown: S7. Predicted blood oxygen saturation value Individualized dynamic correction and smoothing are performed to obtain and output the final cerebral venous sinus blood oxygen saturation value. .

[0059] S7.1, Individualized Dynamic Correction To compensate for errors caused by individual differences in skull parameters (thickness, density) and low perfusion conditions (such as hypotension, microcirculatory disturbances), a dynamic correction factor δ(t) is introduced to fine-tune the output of the deep learning model, specifically incorporating parameters reflecting local heterogeneity of the skull: in, The oxygen saturation of the cerebral venous sinus after dynamic correction; The value is the predicted value of cerebral venous sinus oxygen saturation before correction; δ(t) is the dynamic correction factor. It is the estimation of skull thickness through each channel. The standard deviation is used to correct for deviations caused by unevenness in the skull. For reference average thickness; For correction factors; This is a correction coefficient for the contribution of deeper layers; For time Next The deep contribution of each channel; For the first Reference values ​​for the deep contribution of each channel; This is the heart rate variability correction factor; For time The change in heart rate.

[0060] S7.2 Output Smoothing Processing A combination of moving average and outlier removal is used to ensure output stability and avoid false alarms caused by transient noise. in: The sliding window length is 20 sampling points, corresponding to 0.2s. This is an outlier removal function (which prunes values ​​outside the range [70%, 100%] to that range to avoid invalid output). The final output is the cerebral venous sinus blood oxygen saturation value (time resolution 10ms, measurement error ≤±2%). This represents the clinically effective range of blood oxygen saturation (70%, 100%).

[0061] It should be noted that the above example, using a CNN-LSTM hybrid model and a specific correction formula, details one specific implementation of the "model fitting" and "individualized correction" steps within the technical framework of this invention. The scope of protection of this invention is not limited to this specific embodiment. Any scheme based on the "multi-angle acquisition-multi-channel modeling-intelligent fusion-individualized correction" technical framework of this invention, employing different model fitting algorithms (such as other neural network architectures, traditional machine learning models, etc.) or correction strategies to achieve non-invasive cerebral venous sinus blood oxygenation measurement, falls within the scope of protection of this invention.

[0062] The following describes the system's workflow and modes, using the blood oxygen measurement methods and devices as examples: The system uses "hardware startup - signal acquisition - algorithm processing - result output - early warning feedback" as its core logic, connecting hardware modules and algorithm processes to achieve non-invasive continuous and real-time monitoring of intracranial blood oxygenation in the postauricular skull region. The specific process is as follows: System initialization and hardware deployment: After powering on the power module, the system automatically completes self-tests of hardware such as the probe optical components, signal acquisition module, and communication link. If no faults are found, it enters standby mode. Medical staff place the special skull probe behind the ear onto the patient's ear, adjust its position through the elastic fixing structure to avoid the auricle and external auditory canal, and achieve stable fixation with the help of the customized curvature fitting surface to ensure good contact between the optical window and the skin.

[0063] Multi-channel signal synchronous acquisition: The control and processing module sends a start command to the probe, triggering a multi-wavelength light source with 660nm red light and 940nm near-infrared light to emit light signals with different incident angles and multiple spacings; the photoelectric receiving array synchronously acquires the reflected PPG signal, and after the signal acquisition and conditioning module amplifies the signal and filters to suppress interference, it is converted into a digital signal through analog-to-digital conversion to ensure multi-channel time synchronization.

[0064] Algorithm-side signal processing and blood oxygen calculation: The control and processing module calls a preset algorithm, first using an ANC+KF cascade scheme to cancel ambient light and head micro-motion interference; then using the VMD algorithm to separate the static baseline and dynamic blood flow pulsation components, evaluating channel quality and selecting reliable channels through multi-dimensional indicators; modeling the deep contribution of channels to enhance brain tissue signal weights, constructing a spatial feature matrix input to a CNN-LSTM hybrid model, completing spatiotemporal feature fusion to output preliminary predicted values; combining individualized skull parameters and dynamic correction of perfusion status, and smoothing with a moving average to obtain the final intracranial blood oxygen saturation value.

[0065] Data transmission, display and early warning: The communication and display module transmits the final blood oxygen value and signal quality indicators to the monitor, computer and other terminals in real time through wireless or wired links to realize visual display and historical record storage; the system continuously compares the blood oxygen value with the preset threshold (such as ≤85%), and immediately triggers a clinical early warning through dual sound and light when it is lower than the threshold.

[0066] Continuous monitoring and system closed loop: The system executes the "signal acquisition-algorithm processing-result output" process in a loop with a time resolution of 10ms / frame; the power module monitors the power in real time, issues a prompt when the power is low, and supports hot-swappable charging or battery replacement to ensure battery life; after the monitoring is completed, the system automatically exports complete data for clinical review and medical record archiving.

[0067] Experimental verification: To verify the feasibility and effectiveness of the multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method and device proposed in this invention in penetrating the skull and extracting deep cerebral blood flow signals, this embodiment uses a standard diffuse reflection reference body (white board) and a skull simulated tissue sample (bovine bone) to conduct bidirectional reflectance distribution function (BRDF) optical property tests. The experiment was set with incident angles of 0°, 15°, 30°, and 45°, and the range of reflection angles was -80° to 80° and 0° to 180°. The focus was on verifying the system's ability to capture deep venous sinus blood flow signals, its multi-angle detection adaptability, and its ability to distinguish signals from intracranial and extracranial tissues under conditions of strong skull scattering and high attenuation.

[0068] Reference sample: Standard white board (diffuse reflection reference body conforming to ISO 9050 standard), used to establish the reference reflection characteristics for optical testing.

[0069] Test sample: fresh bovine skull (simulating human skull biological tissue), used to verify the system's performance in real biological tissue scenarios.

[0070] Test conditions: Incident light wavelength is in the visible light band; incident angles are set to 0° (perpendicular incidence), -15°, -30°, and -45° respectively; reflection angle... , The test range is -80° to 80°, the sampling step size is 1°, and the reflectivity test accuracy is 1×10⁻⁶. -4 .

[0071] Experimental Results and Analysis: Comparison of reflectivity distribution characteristics: This experiment obtained the two-dimensional contour distribution of reflectivity of a whiteboard and cow bone under different incident angles, among which... Figure 6The reflectance contour color cloud map is a white board under the incident conditions of 0°, -15°, -30° and -45°. The reflectance is represented by the color gradient from dark blue to bright red: the central bright red / orange-yellow area corresponds to the high reflectance peak, and the color gradually transitions to green, cyan and blue towards the periphery, with the reflectance gradually decreasing. The reflectance of the dark blue area at the edge is close to 0. Figure 7 The color contour map of the bovine bone under the same incident conditions shows that the color distribution pattern is consistent with that of the white board, with the central bright spot corresponding to the reflection peak and the edge area having lower reflectivity.

[0072] Two representative working conditions, 0° (vertical incidence) and -45° (large angle incidence), were selected for comparative analysis. The core parameters are shown in Tables 3-1 and 3-2. Table 3-1 Comparison of reflectivity characteristics of whiteboard and cow bone under 0° incident conditions Table 3-2 Comparison of reflectivity characteristics of whiteboard and cow bone under -45° incident conditions. From Table 3-1, Table 3-2 and Figure 6 , Figure 7 It can be known that: Reflection peak consistency verification: Under incident conditions of 0° and -45°, the center position of the reflection peak of the bovine bone perfectly matches that of the white board, proving that the optical detection system of the present invention can accurately capture the specular reflection direction of biological tissues and that the optical acquisition logic of the system is effective. Biological tissue characteristic verification: The maximum reflectance of bovine bone was reduced by 5.9% (0° incident) and 8.2% (-45° incident) compared to the white plate, respectively, and the half width at half maximum (FWHM) of the reflection peak was narrower, which is consistent with the optical absorption and scattering characteristics of biological tissue (bone), proving that the system can effectively distinguish the optical differences between the reference body and biological tissue. Multi-angle adaptability verification: Under both vertical and large-angle incident conditions, the system can stably acquire effective reflectivity data, and the reflectivity distribution trend is consistent, verifying the reliability of the system under complex incident conditions.

[0073] Three-dimensional reflectivity surface characteristic analysis in Figure 9 The image shows the three-dimensional color surface plots of reflectance of a white board under incident conditions of 0°, -15°, -30°, and -45°. The reflectance is represented by both the surface height and the color gradient. The Z-axis represents the reflectance value. The reflectance increases from purple / dark blue (0.000) to bright red (the highest value is about 0.008 to 0.010). The sharp bright red peak in the center corresponds to the high reflectance area. The surface height decreases and the color gradually transitions to green, cyan, and blue towards the edges. The reflectance decreases gradually, and the edge area is dominated by purple / dark blue with a reflectance close to 0. Figure 10The image shows the three-dimensional color surface of the bovine bone under the same incident conditions. The overall surface shape is consistent with that of the white board, except that the height of the central bright red peak is slightly lower and the reflectance attenuation gradient is steeper. The color distribution pattern is consistent with that of the white board.

[0074] Depend on Figure 9 , Figure 10 It can be known that: The three-dimensional curved surface of the whiteboard exhibits a sharp peak at the center and a symmetrical, gently decaying shape around the edges, with no obvious distortion, which verifies the stability and accuracy of the experimental system. The three-dimensional curved surface of the bovine bone is consistent with the height of the whiteboard, and the peak position is perfectly matched, with only the peak height being slightly lower. This further proves that the system of the present invention can accurately capture the reflectivity characteristics of biological tissues, providing reliable optical data support for clinical tissue identification and imaging detection.

[0075] Two-dimensional angular reflectivity curve analysis Figure 8 Two-dimensional angle-reflectivity curves of a whiteboard and a cow bone under different incident angles are shown. The peak value of the curve corresponds to the maximum reflectivity, and different colored curves correspond to different incident angles. The shape of the curve represents the trend of reflectivity with the reflection angle.

[0076] Depend on Figure 8 It can be known that: Both the whiteboard and the cow bone curves show a single-peak distribution with the same peak position, but the peak value of the cow bone curve is slightly lower than that of the whiteboard. The curve trends under different incident angles (15°, 30°, 45°) are completely matched, proving the consistency and reliability of the system under multi-angle incident conditions, which can meet the detection needs of complex clinical scenarios.

[0077] Experimental conclusions The above-mentioned comparative experiment on the reflectivity of a standard whiteboard and cow bone fully verified the technical effectiveness of the present invention: Feasibility verification: The optical detection system of this invention can accurately capture the reflectance distribution of biological tissue (bovine bone) under different incident angles. The position and distribution trend of the reflection peak are highly consistent with the standard reference body, proving that the optical acquisition and analysis logic of the system is effective and can be applied to the optical detection of biological tissue. Tissue differentiation capability verification: There are significant differences in the peak reflectance and attenuation characteristics between bovine bone and whiteboard. The system can effectively differentiate them, providing data support for the identification and detection of bone tissue in clinical practice. Multi-angle adaptability verification: The system can stably acquire effective reflectivity data at multiple incident angles such as 0°, -15°, -30°, and -45°, demonstrating good environmental adaptability and meeting the multi-angle detection needs of actual clinical applications.

[0078] The parts not described in detail in the above embodiments are existing technologies.

[0079] It should be noted that although the present invention has been described through the above embodiments, the present invention may have many other embodiments. Without departing from the spirit and scope of the present invention, those skilled in the art can obviously make various corresponding changes and modifications to the present invention, but all such changes and modifications should fall within the scope of protection of the appended claims and their equivalents.

Claims

1. A multi-angle reflective PPG non-invasive method for measuring intracranial venous sinus blood oxygenation, characterized in that, Includes the following steps: S1, the incident angle of the wavelength and the distance between the light source and the detector The predetermined optical probe is fitted to the skull region behind the ear of the subject. During multiple time periods, including at least a calibration period and a monitoring period, the multi-wavelength light source inside the optical probe is driven to emit detection light at different incident angles. Multiple photodetectors simultaneously collect optical reflection signals from superficial tissues and deep blood flow. ; S2, Optical reflection signal Interference suppression is performed, and the optical reflection signals collected during the calibration period are matched and fitted with a pre-established tissue optical parameter model library. Based on the optimization algorithm, an individualized set of skull parameters for the postauricular skull region of the subject is estimated in real time. And calculate the comprehensive skull attenuation factor for each channel-wavelength combination. and the average skull penetration efficiency of this channel under multiple wavelengths. ; Individualized skull parameter set At least including the thickness of the local skull ; For the optical reflection signals collected during the monitoring period, based on the comprehensive skull attenuation factor... Skull attenuation compensation was performed to obtain the compensated optical reflection signal. ; S3. Based on the variational mode decomposition algorithm, analyze the compensated optical reflection signal. Separate the components and obtain the static baseline components of each wavelength in each channel. and dynamic blood flow pulsation components And calculate the blood oxygen ratio. ; S4. Constructing signal quality evaluation indicators And combined with the average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing and local skull thickness Establish a depth contribution to characterize the ability of each channel to detect deep blood flow signals. ; S5, Based on Deep Contribution Blood oxygen ratio Average skull penetration efficiency Wavelength and angle of incidence Light source-detector spacing Construct a spatial feature matrix containing skull attenuation features ; S6. The spatial feature matrix The data is input into a pre-trained intelligent correction model, which adaptively increases the feature weights of deep blood flow and suppresses the feature weights of superficial tissues. It also extracts and fuses features related to deep blood flow from multi-channel spatiotemporal signals, outputting a predicted value of cerebral venous sinus blood oxygen saturation. ; S7. The predicted blood oxygen saturation value Individualized dynamic correction and smoothing are performed to obtain and output the final cerebral venous sinus blood oxygen saturation value. .

2. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that: The tissue optical parameter model library is based on bidirectional reflectance distribution function and bidirectional transmission distribution function, combined with the anisotropy and multilayer structure characteristics of human tissue, and is established through in vitro or in vivo calibration measurements.

3. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that, Suppressing background interference includes canceling ambient light interference in the optical reflection signal based on an ANC filter; and / or eliminating motion artifacts in the optical reflection signal based on a Kalman filter.

4. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that: The comprehensive skull attenuation factor Calculated using the following formula: in, The directional transmission function is determined by the tissue optical parameter model library. For the first The estimated skull thickness is obtained from each channel; For the first The wavelength incident angle of each channel; The relative density of the skull. λ is the wavelength.

5. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that, Deep contribution of integrating the optical properties of the skull Calculated using the following formula: in, Indicates channel Average skull penetration efficiency under dual wavelengths , , They represent the first Each channel has wavelength , The corresponding comprehensive skull attenuation factors; and These are the normalization coefficients; for Signal quality evaluation metrics for each channel , For signal-to-noise ratio, For signal stability, For the consistency of blood flow pulsation , , These are the weighting coefficients, and + + =1; For the first The light source-detector spacing of each channel.

6. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that: The intelligent correction model is a hybrid model of convolutional neural network and long short-term memory network, in which the convolutional neural network is used to extract the spatial features of multi-channel signals, and the long short-term memory network is used to extract the time series features of signals.

7. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that: In step S6, the individualized dynamic correction is calculated using the following formula: in, The oxygen saturation of the cerebral venous sinus after dynamic correction; The value is the predicted value of cerebral venous sinus oxygen saturation before correction; δ(t) is the dynamic correction factor. It is the estimation of skull thickness through each channel. The standard deviation is used to correct for deviations caused by unevenness in the skull. For reference average thickness; For correction factors; This is a correction coefficient for the contribution of deeper layers; For time Next The deep contribution of each channel; For the first Reference values ​​for the deep contribution of each channel; This is the heart rate variability correction factor; For time The change in heart rate.

8. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measurement method according to claim 1, characterized in that: In step S7, the smoothing process includes moving average processing and outlier removal.

9. A multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measuring device, characterized in that, The blood oxygen measurement device includes: An optical probe has an asymmetric fitting structure that matches the curvature of the cranial region behind the ear, and the optical probe integrates a light source array and a detector array. A light source array for emitting detection light of at least two wavelengths, and for at least one wavelength, multiple light sources are arranged at different incident angles; A detector array, arranged in a ring or array form at a predetermined distance from the light source array, is used to receive reflected light signals; A signal acquisition and processing unit, connected to the detector array, is used to condition and convert the reflected light signal from analog to digital, and to perform the steps of the method according to any one of claims 1-8; A storage unit is used to store the preset tissue optical parameter model library and the pre-trained intelligent correction model; The power supply and communication unit is used to supply power and transmit data to the optical probe and signal acquisition and processing unit.

10. The multi-angle reflective PPG non-invasive cerebral venous sinus blood oxygenation measuring device according to claim 9, characterized in that: The signal acquisition and processing unit is further configured to: During the calibration period, the light source array is controlled to perform multi-wavelength, multi-angle optical scanning, and the acquired baseline optical signals are matched and fitted with the tissue optical parameter model library in the storage unit to estimate the individualized skull optical parameters of the current subject online. ; During the continuous monitoring phase, the individualized skull optical parameters are used to compensate for skull attenuation in the real-time acquired signals, and the compensated signals are input into the pre-trained intelligent correction model.

Images