A method and device for detecting local cerebral tissue oxygen saturation by near-infrared

Abstract

The application discloses a kind of local brain tissue blood oxygen saturation near-infrared detection method and device, to solve the problem of the difference between the group average head model parameter and patient individual parameter in prior art, especially severe patients;Existing device also has the problems of positioning deviation, poor repeatability, insufficient accuracy and difficult to quantify indicators. Through the accurate and repeatable positioning of skin texture double-mode positioning sticker, the individual head three-dimensional anatomical model and individual multi-layer light transmission simulation model are constructed by querying the patient individual image data, and the light transmission loss is calculated layer by layer, combined with the patient individual skull density, based on the patient individual image head blood flow map analysis of cerebral blood perfusion process and individual neck arteriovenous biochemical blood gas detection, to complete the individualized detection of local brain tissue blood oxygen saturation. The application calls the existing multi-source detection data of the patient to improve the detection accuracy of local brain tissue oxygen saturation by algorithm calculation, without any additional examination and cost.

Description

Technical Field

[0001] This invention relates to the field of near-infrared spectroscopy detection and medical testing data processing technology, specifically to a near-infrared detection method and device for local brain tissue blood oxygen saturation. Background Technology

[0002] Local brain tissue oxygen saturation Regional cerebral tissue oxygen saturation (RTS) refers to the percentage of actual oxygen-binding capacity of hemoglobin within the cerebral microcirculation capillaries, reflecting the balance between oxygen supply and consumption in brain cells and serving as a crucial indicator for evaluating cerebral blood oxygen metabolism and brain function. Near-infrared spectroscopy (NIRS), referred to as NIRS in this invention, is used to calculate blood oxygen levels. The accuracy of this device's blood oxygenation calculations is highly dependent on the fitting accuracy of the actual optical model within the head tissue. The fitting accuracy of the head optical model directly determines the accuracy and stability of the final detection data. Medical and imaging personnel have found that existing equipment suffers from problems such as localization bias, significant individual variability, poor repeatability, insufficient detection accuracy, difficulty in quantifying indicators, and a lack of effective integration of NIRS with other detection methods. The change in regional cerebral tissue oxygen saturation is insufficient to meet the needs of high-precision individual diagnosis and treatment. Although medical staff can intuitively perceive the error phenomenon, their work focuses on data interpretation and disease assessment, and they lack the understanding of optical transmission mechanisms and underlying modeling. Therefore, they cannot trace the physical causes of the error, nor do they have the technical conditions to modify the equipment model.

[0003] Those skilled in the art (R&D engineers) typically improve instrument accuracy through hardware optimization, including signal denoising, hardware gain adjustment, and digital filter calibration. In recent years, the focus of R&D has shifted to algorithm optimization, aiming to improve data processing performance, and significant progress has been made in the extraction of weak target signals. For example, patent CN118806259A combines Transformer and 3DCNN to build a PM Conformer model, identifying abnormal brain blood oxygenation based on fMRI images, fully exploring spatial and temporal information in the images to improve detection results. Patent CN120114050A integrates near-infrared light attenuation signals and head motion signals, preprocessing and filtering them before calculating blood oxygen saturation using a dual-wavelength ratio model, reducing detection errors caused by head shaking. Patent CN121176903A, based on a CycleGAN network and combined with denoising and multi-scale iterative optimization strategies, transforms low-quality brain blood oxygenation signals collected by wearable devices into high-quality signals that meet standards. Patent CN118902448A combines skin pigmentation and ambient light brightness parameters with a BP neural network to correct monitoring data, improving the accuracy of cerebral blood oxygenation detection while controlling costs. Patent CN112043287A utilizes near-infrared light absorption characteristics to achieve non-invasive cerebral blood oxygenation monitoring, adding correction factors to eliminate tissue interference and using a predictive model to achieve continuous real-time monitoring, resulting in better signal-to-noise ratio and monitoring stability. Patent CN111035396A uses algorithms such as Green's function and Monte Carlo simulation to construct a photoacoustic measurement model, establishing a reference standard for cerebral functional blood oxygenation monitoring. Patent CN118749966A uses wavelet transform and empirical mode decomposition for multi-scale signal decomposition, combined with regression algorithms to scale-calibrate errors, achieving accurate correction of cerebral blood oxygenation detection data. Patent CN117442200B discloses a high-precision device for measuring the blood oxygenation status of cerebral cortex tissue, solving the expression for the change in optical density after noise removal at three near-infrared wavelengths to obtain the blood oxygenation status of cerebral cortex tissue. Patent CN1540314A discloses a non-invasive method for monitoring blood oxygen metabolism in biological tissues based on diffuse light. A photodetector located on one side of three light-emitting diodes sequentially detects the light intensity of diffused light passing through deep tissue from the three diodes, and then calculates the optical density accordingly. And the blood oxygen saturation of the tissue to be tested.

[0004] Theoretical researchers have conducted in-depth studies on the layered optical transmission mechanisms of biological tissues and the scattering and absorption laws of multilayer media, establishing mature multilayer optical transmission simulation models and accurately quantifying the differences in optical path length between layers. The human head is composed of layers of scalp, skull, cerebrospinal fluid, and brain tissue, with significant differences in optical parameters, thickness characteristics, and optical loss features among these tissues. However, these sophisticated models are computationally complex and remain primarily at the theoretical research stage. Research and development engineers struggle to deeply understand and apply these models, hindering their application in the iterative optimization of detection equipment. Therefore, the average head model and homogeneous medium model have long been used as the basic modeling paradigm, modified by incorporating DPF (Differential Path Length Factor), such as... In the formula The coefficients in this formula change with the detection wavelength, taking into account the subject's age. For the same subject, at the same location, and within a short period, the variation will be [not specified]. Only then can it be reliable. Patent CN116421183A uses an individualized optical cap constructed from the subject's brain tissue to achieve individualized blood oxygen response signal acquisition, thereby spatially improving the accuracy of target brain region localization and blood oxygen response signal acquisition. However, patent CN116421183A continues to use a population average head model and a uniform medium model without applying a multilayer optical transmission simulation model. Overall, current engineering optimizations mostly focus on signal post-processing, noise suppression, and external interference correction surface optimization.

[0005] In summary, medical staff, R&D engineers, and theoretical researchers belong to different fields. Although there is regular communication between them, differences in professional systems, knowledge structures, and technical perspectives significantly hinder in-depth communication and collaborative integration, thus limiting research and application thinking. Medical staff can directly identify detection errors in equipment, and theoretical researchers have built mature multi-layer optical transmission simulation models that can accurately adapt to the optical transmission characteristics of layered head tissues. However, R&D engineers struggle to identify innovative points and apply high-precision multi-layer optical transmission simulation models to engineering R&D and equipment iteration. For a long time, cross-disciplinary integration and optimization solutions in the field of near-infrared brain oxygenation detection have remained a key target for technological breakthroughs in the industry. Currently, R&D engineers can only observe superficial problems of insufficient equipment detection accuracy. Limited by inherent R&D systems and disciplinary barriers, they cannot individually distinguish the actual optical path loss and scattering differences of different tissue layers in the human head, making it difficult to overcome the technical bottleneck of mismatched underlying physical models between the average head model, the homogeneous medium model, and the layered structure of the head. This industry situation has resulted in inherent errors in the modeling of near-infrared cerebral oxygenation detection. The industry has consistently lacked effective technical solutions to correct accuracy defects from a physical perspective, and conventional engineering optimization methods have been unable to address these core issues. To overcome these limitations, it is essential to quantify individual anatomical differences into optical path factors and incorporate them into classical calculation algorithms. This fundamentally eliminates the oxygenation calculation bias caused by differences in head tissue structure and thickness. Summary of the Invention

[0006] This invention provides a near-infrared method and device for detecting local brain tissue oxygen saturation, addressing the problem of significant differences between the average head model parameters used in existing devices and individual patient parameters, especially for critically ill patients; and resolving issues such as positioning bias, poor repeatability, insufficient detection accuracy, and difficulty in quantifying indicators in existing detection devices. To achieve the above objectives, an individual patient head model and a multi-layer optical transmission simulation model are constructed by querying individual patient imaging data, and optical transmission loss is calculated layer by layer. Combined with individual patient skull density, analysis of cerebral blood flow perfusion process based on individual patient imaging head blood flow maps, and biochemical blood gas analysis of carotid arteries and veins, individual detection of local brain tissue oxygen saturation is completed.

[0007] A near-infrared method for detecting local brain tissue oxygen saturation includes four steps: establishing a streamline database, calculating concentration changes, calculating brain oxygen saturation, and generating a report. Establishing the streamline database involves four sub-steps: creating a head model set, Monte Carlo simulation, extracting the most probable streamlines, and storing the database. Calculating concentration changes involves five sub-steps: creating a patient head model, database retrieval and matching, calculating the length between intersections, skull density calibration, and calculating changes in blood oxygen concentration. Calculating brain oxygen saturation involves three sub-steps: biochemical blood gas verification, baseline concentration calibration, and calculating blood oxygen saturation.

[0008] S1. Establishing a streamline database: First, a head model set is established. After collecting brain maps of preset classification dimensions, a series of 3D head models covering the patient's anatomical and geometric features are generated. Finite element models or voxel models are preferred for these 3D head models. Anatomical features include scalp thickness, subcutaneous tissue thickness, skull thickness, cerebrospinal fluid thickness, gray matter thickness, gray matter radius of curvature, age, and gender. Geometric features include light source-detector pairs, light source-detector distances, detector layout positions, and relative angles. Each head model has different light source-detector distances and relative angles set for common detector layout positions. Subsequently, Monte Carlo simulation is performed. GPU-accelerated Monte Carlo simulation software simulates the transmission process of near-infrared photons in the head models composed of specific anatomical and geometric features, generating a statistical set of numerous photon simulation paths corresponding to the anatomical and geometric features. The more refined the preset classification dimensions, the greater the computational load of the Monte Carlo simulation; meeting the requirements serves as a benchmark for the level of classification refinement. Pre-classifying the head models to complete the Monte Carlo simulation avoids consuming a large amount of computation time in subsequent detection.

[0009] Next, the maximum probability streamline is extracted. Based on the statistical set of photon paths, the maximum probability streamline corresponding to specific anatomical and geometric features is extracted. The maximum probability streamline refers to the spatial trajectory with the highest probability density, representing the maximum probability transmission trend of photons, extracted from the statistical set of massive random photon transmission paths simulated by a specific head model with specific anatomical and geometric features. It is composed of a set of three-dimensional coordinate points. Then, each maximum probability streamline and its feature label are stored together in the maximum probability streamline database. The feature label includes anatomical and geometric features.

[0010] S2. Concentration Change Calculation: A patient head model is established using MRI (Magnetic Resonance Imaging) and CT (Computed Tomography) cross-sectional images of the patient's head. A finite element model or voxel model is created. Image processing methods are used to measure the scalp thickness, subcutaneous tissue thickness, skull thickness, cerebrospinal fluid thickness, gray matter thickness, and gray matter radius of curvature in the near-infrared brain oxygenation detection target area. These measurements, along with the patient's age and gender, constitute the patient's anatomical features. Based on the actual position of the near-infrared brain oxygenation detection cap worn by the patient, the light source-detector pair, light source-detector distance, detector layout position, and relative angle are extracted to constitute the patient's geometric features. A retrieval expression is formed using the patient's feature labels (anatomical and geometric features). Finally, database retrieval and matching are performed. The patient's feature labels are compared with feature labels in the maximum probability streamline database. Nearest neighbor search or multi-feature weighted matching is used to find the maximum probability streamline corresponding to the closest individual patient head model from the database.

[0011] Preferably, patients with stroke, traumatic brain injury, brain tumor, or post-neurosurgery undergo routine head MRI or CT scans upon admission. These imaging data are stored in the hospital's PACS (Picture Archiving and Communication System) and can be accessed at any time after authorization. Based on existing head MRI or CT images in the patient's medical record database, the following bony landmarks are selected as localization references: the junction of the frontal and nasal bones at the root of the nose, the external occipital protuberance, the root of the zygomatic arch at the bilateral preauricular points, the bony depression in front of the tragus, and the bony prominence above the orbit at the upper edge of the eyebrow. After marking these reference points on the images, the anterior-posterior midline of the head (the line connecting the root of the nose to the external occipital protuberance) and the lateral lines (the line connecting the bilateral preauricular points) can be determined. Then, according to the international 10-20 system ratio or direct measurement, the projection position of the prefrontal cortex detection area (such as Fp1 and Fp2 points) on the scalp surface is calculated. This position guides the wearing of the frontal, lateral, and posterior edges of the near-infrared detection cap, ensuring that the light source-detector array is accurately aligned with the target brain region. This method utilizes clearly identifiable bony landmarks on imaging as positioning anchors, which is more objective and repeatable than simply relying on surface palpation or fixed distances, and can fully utilize the patient's existing imaging data to achieve individualized positioning. This method does not require any additional examinations or procedures, reducing the financial burden on patients.

[0012] When patients undergo MRI and CT scans, combining skin texture features with the dual-mode positioning patch to establish the spatial relationship between head cross-sectional data and the near-infrared detection cap's light source-detector assembly yields even better results. Furthermore, this dual-mode positioning patch can also serve as a positioning reference for constructing a three-dimensional model of the patient's head based on MRI and CT cross-sectional images.

[0013] Preferably, the length between intersection points is calculated, and the selected maximum probability streamline is mapped onto the patient's individual head model space. This can be done using a linear scaling method, where the maximum probability streamline is anisotropically scaled and displaced based on the ratio of the patient's individual head model to head model instances in the library in terms of key anatomical features, and then superimposed on the patient's individual head model. Alternatively, a non-rigid registration method can be used, where the head contour of the head model instance is non-rigidly registered with the head contour of the patient's individual head model to obtain a deformation field, which is then applied to the maximum probability streamline to fit the anatomical structure of the patient's individual head model. After mapping, the maximum probability streamline is superimposed on the patient's individual head model, completing the drawing of the maximum probability streamline. Next, the intersection points of this maximum probability streamline with the boundaries of each layer of the patient's individual head model are marked, including the skin layer, subcutaneous tissue layer, galea aponeurotica and periosteum layer, compact bone layer of the skull, cancellous bone layer of the skull, internal periosteum and dura mater layer, arachnoid mater and subarachnoid space layer and pia mater, and the intersection points of the gray matter layer with the maximum probability streamline. Finally, the curve length between the intersection points of the maximum probability streamline and the tissue layers of the patient's head is calculated based on the physical dimensions corresponding to the pixels. This length is the path length of near-infrared light as it travels through this tissue layer.

[0014] Calculate the change in blood oxygen concentration and the curve length. Calculate the optical path factor by substituting the optical parameters. (i.e., differential path length factor DPF), combined with near-infrared light density change measurements Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

[0015] In addition, skull density calibration involves querying patient skull density data obtained from quantitative computed tomography (QCT) scans and replacing the bone density parameters of the population-average head model used in the optical path factor calculation process, thereby calibrating the change in local brain tissue oxyhemoglobin concentration. and changes in deoxyhemoglobin concentration Skull density is used as a calibration parameter to further quantify the transmission and attenuation of near-infrared light signals within each patient's skull, effectively separating the interference of the skull on near-infrared light signals and eliminating detection errors caused by individual differences in skull density.

[0016] The improved algorithm fundamentally enhances the accuracy of near-infrared detection of local brain tissue blood oxygen saturation; simultaneously, it reduces the cost of each Monte Carlo simulation calculation and saves detection time by querying the maximum probability streamline database.

[0017] S3. Calculate cerebral oxygen saturation, query biochemical blood gas test data, and calibrate total hemoglobin concentration. Carotid artery oxygen saturation With venous oxygen saturation Baseline concentration was calculated by extracting local cerebral hemodynamic parameters. , Absolute concentration of oxyhemoglobin in local brain tissue and absolute concentration of deoxyhemoglobin The process is as follows:

[0018] Biochemical blood gas analysis was performed, and the patient's multi-sequence MRI data—high-resolution T1 / T2 structural images and cerebral blood flow perfusion maps—was retrieved to extract local cerebral hemodynamic parameters. The difference between arterial arrival time and venous outflow time in the target region was analyzed. Combined with a dual-compartment tracer kinetic model, the distribution density ratio of arterioles to venules was estimated, supplemented by susceptibility-weighted imaging or quantitative susceptibility mapping to enhance the spatial identification ability of venule density. Assuming a linear positive correlation between local vascular density and blood volume, the total cerebral blood volume was decomposed into arterial and venous blood volume components. Then, using carotid artery oxygen saturation and venous oxygen saturation as boundary conditions, and combining the aforementioned arterial and venous blood volume components, the instantaneous local brain tissue oxygen saturation was calculated. At the same time, the total hemoglobin concentration obtained from the patient's routine blood tests upon admission was checked. After unit conversion and correction for hematocrit, the total hemoglobin concentration in brain tissue was obtained. Total hemoglobin concentration Equal to the absolute concentration of oxyhemoglobin and absolute concentration of deoxyhemoglobin The sum. Baseline concentration calibration, calculated from the instantaneous local brain tissue oxygen saturation. and total hemoglobin concentration The baseline concentration can be calculated. , Subsequently, blood oxygen saturation was calculated at baseline concentration. , The change in oxyhemoglobin concentration is monitored in real time by superimposed near-infrared detection cap. and changes in deoxyhemoglobin concentration This allows us to obtain the absolute concentration of oxyhemoglobin at any given time. and absolute concentration of deoxyhemoglobin This allows for the calculation of local brain tissue oxygen saturation at any given time. The detailed calculation process is described in Example 3. This data processing method elevates near-infrared detection from merely reflecting relative changes to outputting absolute quantitative values ​​of local brain tissue oxygen saturation at any given time for each individual patient. This provides doctors with more reliable testing data for decision-making.

[0019] S4. Report generation: The calculated blood oxygen saturation is physiologically validated, and invalid results exceeding the physical range are removed. Values ​​exceeding the normal physiological reference range but still within the physical range are marked and anomalies are indicated in the report. Finally, the report generates local brain tissue blood oxygen saturation data including the detection geometry. Test report.

[0020] Correspondingly, a near-infrared detection device for local brain tissue oxygen saturation includes a streamline database establishment module, a concentration change calculation module, a brain oxygen saturation calculation module, and a report generation module.

[0021] The streamline database establishment module includes a head model set establishment submodule, a Monte Carlo simulation submodule, a maximum probability streamline extraction submodule, and a database storage submodule. The head model set establishment submodule collects brain maps of preset classification dimensions and generates a three-dimensional head model containing anatomical and geometric features. The Monte Carlo simulation submodule runs Monte Carlo simulation software to simulate the transmission process of near-infrared photons in the head model. The maximum probability streamline extraction submodule generates a large set of simulated photon paths for light source-detector pairs and extracts the maximum probability streamlines. The database storage submodule stores the streamlines and their features together in the database.

[0022] The concentration change calculation module includes a patient head model establishment submodule, a database retrieval and matching submodule, an intersection length calculation submodule, a skull density calibration submodule, and a blood oxygen concentration change calculation quantum module, used to calibrate the optical path factor to calculate the local brain tissue oxyhemoglobin concentration change. and changes in deoxyhemoglobin concentration .

[0023] The patient head model creation submodule queries the patient's head MRI and CT cross-sectional images to create a head model containing anatomical and geometric features; the database retrieval and matching submodule retrieves the database based on the patient's described features; through nearest neighbor search or multi-feature weighted matching, it finds the closest maximum probability streamline from the database and displays it on the individual patient head model using linear scaling or non-rigid registration.

[0024] The submodule for calculating the length between intersection points calculates the curve length between the intersection points of the maximum probability streamline and the tissue layers of the patient's head based on the physical dimensions corresponding to the pixels. The skull density calibration submodule is used to query the patient's recently measured skull density data to calibrate the optical path factor, thereby calibrating the change in local brain tissue oxyhemoglobin concentration. and changes in deoxyhemoglobin concentration .

[0025] The quantum module for calculating changes in blood oxygen concentration substitutes the curve length. The optical path factor is calculated using optical parameters, combined with measurements of near-infrared light density changes. Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

[0026] The cerebral blood oxygen saturation calculation module includes a biochemical blood gas verification submodule, a baseline concentration calibration submodule, and a blood oxygen saturation calculation submodule. After querying biochemical blood gas parameters and calibrating the baseline concentration, it calculates the local brain tissue blood oxygen saturation at any given time. .

[0027] The biochemical blood gas calibration submodule is used to query biochemical blood gas test data to calibrate total hemoglobin concentration. Carotid artery oxygen saturation With venous oxygen saturation The baseline concentration calibration submodule supports improving the absolute concentration of oxyhemoglobin in local brain tissue. and absolute concentration of deoxyhemoglobin The accuracy.

[0028] The baseline concentration calibration submodule is used to provide the individual baseline concentration for near-infrared detection calculations. , The change in the previously measured oxyhemoglobin concentration and changes in deoxyhemoglobin concentration Converted to the absolute concentration of oxyhemoglobin in local brain tissue and absolute concentration of deoxyhemoglobin This allows for the calculation of local brain tissue oxygen saturation at any given time. .

[0029] Querying the patient's multi-sequence MRI data—high-resolution T1 / T2 structural images and cerebral blood flow perfusion maps—extracted local cerebral hemodynamic parameters. The difference between arterial arrival time and venous outflow time in the target region was analyzed. Combined with a dual-compartment tracer kinetic model, the distribution density ratio of small arteries to small veins was estimated. Susceptibility-weighted imaging or quantitative susceptibility mapping was used to enhance the spatial identification ability of small vein density. Assuming a linear positive correlation between local vascular density and blood volume, the total cerebral blood volume was decomposed into arterial and venous blood volume components. Then, using carotid artery oxygen saturation and venous oxygen saturation as boundary conditions, combined with the aforementioned arterial and venous blood volume components, the instantaneous local brain tissue oxygen saturation was calculated. .

[0030] At the same time, the total hemoglobin concentration was obtained from the patient's routine blood test upon admission. After unit conversion and correction for hematocrit, the total hemoglobin concentration in brain tissue was obtained. Total hemoglobin concentration Equal to the absolute concentration of oxyhemoglobin and absolute concentration of deoxyhemoglobin The sum. Calculated from the instantaneous local brain tissue oxygen saturation. and total hemoglobin concentration The baseline concentration can be calculated. , .

[0031] The blood oxygen saturation calculation submodule uses the near-infrared detection cap to monitor the real-time change in oxyhemoglobin concentration. and changes in deoxyhemoglobin concentration Superimposed on baseline concentration , By doing so, the absolute concentration of oxyhemoglobin at any given time can be obtained. and absolute concentration of deoxyhemoglobin This allows for the calculation of local brain tissue oxygen saturation at any given time. The calculation process is described in Example 3. This method elevates near-infrared detection from merely reflecting relative changes to outputting absolute quantitative indicators for individuals, providing doctors with more accurate detection data for decision-making.

[0032] The report generation module is used to verify the physiological rationality of the calculated blood oxygen saturation, remove invalid results that are outside the physical range, mark values ​​that are outside the normal physiological reference range but still within the physical range, and indicate and prompt the data sources used in the calculation process in the report.

[0033] Compared with existing methods, the present invention has the following advantages:

[0034] 1. This invention differs from existing instruments that rely on a population average head model. Approximately equal to However, this method suffers from significant engineering approximation errors. Based on meticulous layered reconstruction of the head's anatomical structure using MRI and CT images, and by constructing individualized optical models according to the differences in near-infrared spectral absorption and scattering characteristics of each tissue layer, the method accurately simulates the near-infrared light transmission attenuation process of the head. Optical path factor is achieved by separating optical path loss and extracting backscattering signals from brain tissue. The individualized solution of the differential path length factor (DPF) significantly reduces engineering approximation errors and completes individual calibration for near-infrared detection. This invention quantifies individual anatomical differences into optical path factors and integrates them into classical calculation algorithms, fundamentally eliminating the deviation in blood oxygenation calculation caused by differences in head tissue structure and thickness, thus laying a methodological foundation for high-precision and high-reliability individualized cerebral blood oxygenation detection.

[0035] 2. When patients undergo MRI and CT scans, a dual-mode positioning patch is used to match skin texture features, establishing the spatial relationship between head profile data and the light source-detector assembly on the near-infrared detection cap. This enables precise localization of the same area using multimodal detection, effectively avoiding the localization deviations and repeatability errors that easily occur between different detection methods, and significantly improving the localization accuracy of local brain tissue oxygen saturation detection.

[0036] 3. This invention introduces skull density as a key calibration parameter, further quantifying the transmission and attenuation characteristics of near-infrared spectroscopy within the skull, effectively eliminating the attenuation interference of the skull on the light signal, and eliminating detection errors caused by individual differences in skull density from a physical perspective.

[0037] 4. This invention determines the local brain tissue oxygen saturation based on local cerebral hemodynamic parameters obtained from MRI. Based on the results of biochemical blood gas analysis, baseline concentrations were established for the relative concentration changes obtained from near-infrared spectroscopy. , Calculation. This mechanism breaks through the limitation of traditional near-infrared detection technology, which can only reflect relative changes, and successfully realizes the calculation from relative change to absolute concentration. and The conversion transforms low-precision trend fluctuation detection into high-precision quantitative detection.

[0038] 5. This invention creatively integrates MRI and CT, ultrasound images, bone density data, biochemical blood gas test results, and dynamic detection data of local brain tissue oxygen saturation. By leveraging the strengths and avoiding the weaknesses, it retains the advantages of near-infrared detection—non-invasive, continuous, and convenient—while solving technical problems in existing detection methods such as inaccurate positioning, significant signal interference, lack of benchmarks, and difficulty in quantification, thus achieving synergistic effects from the integration of multiple technologies.

[0039] 6. This invention serves as a firmware upgrade component or software patch for existing near-infrared cerebral oximeters, enabling collaboration with existing manufacturers through OEM / licensing, leveraging their existing equipment and sales channels for rapid market penetration.

[0040] In summary, this invention has extremely high clinical adaptability and practicality: medical staff can completely follow the existing near-infrared detection cap wearing and operation procedures without changing their clinical work habits; the required head MRI / CT images, blood routine tests, and blood gas analysis are all readily available. Core data can be directly retrieved from PACS and LIS (Laboratory Information System) data generated during routine patient admission examinations, eliminating the need for patients to undergo additional specialized imaging or laboratory tests, and thus not increasing treatment costs or physical burden. This invention automatically completes multimodal data matching, individualized optical path factor calibration, and precise blood oxygen line anchoring in the background, completely solving the industry pain points of traditional near-infrared cerebral oximeters, such as "uniform zeroing upon startup, inability to adapt to individual anatomical differences, suspended detection baseline, and poor quantitative accuracy." It has extremely high application value and promising prospects in clinical departments requiring continuous and accurate cerebral oxygenation monitoring, such as ICUs, neurosurgery, and anesthesiology. Attached Figure Description

[0041] Figure 1 A schematic diagram illustrating the calibration principle of near-infrared detection for local brain tissue oxygen saturation.

[0042] Figure 2 A schematic diagram of multi-source data fusion from a near-infrared detection device for local brain tissue oxygen saturation.

[0043] Figure 3 This is a flowchart of a near-infrared detection device for local brain tissue oxygen saturation.

[0044] Explanation of reference numerals in the attached figures: 101 Light source, maximum probability streamline of 850nm near-infrared light; 102 Maximum probability streamline of 760nm near-infrared light; 103 Detector; 104 Scalp; 105 Subcutaneous tissue; 106 Skull; 107 Dura mater; 108 Cerebrospinal fluid; 109 Pia mater; 110 Gray matter; 111 Gray matter. Detailed Implementation

[0045] The present invention will be further described in detail below with reference to specific embodiments. The scope of protection of the present invention is not limited to the following embodiments.

[0046] Example 1:

[0047] A near-infrared method for detecting local brain tissue oxygen saturation, one aspect of which involves establishing a maximum probability streamline database of head model instances. Specifically, as follows... Figure 1 As shown, after collecting brain atlases with preset classification dimensions, a series of three-dimensional head models covering the patient's anatomical and geometric features are generated. These three-dimensional head models are preferably finite element models or voxel models. Anatomical features include scalp thickness (105), subcutaneous tissue thickness (106), skull thickness (107), cerebrospinal fluid thickness (109), gray matter thickness (111), gray matter radius of curvature (111), age, and sex. Geometric features include the distance between light source (101) and detector (104), the layout position of detector (104), and the relative angle. For each head model, different distances between light source (101) and detector (104) and different relative angles are set at common detector (104) layout positions. Subsequently, for each head model composed of anatomical and geometric features, GPU-accelerated Monte Carlo simulation software is used to simulate the transmission process of near-infrared photons in the head model, generating a statistical set of numerous photon simulation paths corresponding to the anatomical and geometric features. The more refined the preset classification dimensions, the larger the Monte Carlo simulation workload; meeting the requirements serves as a benchmark for classification refinement. By using a pre-defined classification head model to perform Monte Carlo simulation, the time spent on Monte Carlo simulation during subsequent detection is avoided.

[0048] Next, the maximum probability streamlines are extracted. Based on the photon path statistics set, the maximum probability streamlines corresponding to the relevant anatomical and geometric features of each pair of light sources (101-104) are extracted, such as... Figure 1The maximum probability streamlines 102 for 850nm near-infrared light and 103 for 760nm near-infrared light are included. A maximum probability streamline refers to the spatial trajectory with the highest probability density, representing the maximum probability transmission trend of photons, extracted from a statistical set of massive random photon transmission paths simulated by a specific head model, specific anatomical features, and geometric features. It is composed of a set of three-dimensional coordinate points. Subsequently, each maximum probability streamline and its feature labels are stored in a maximum probability streamline database. The feature labels include anatomical and geometric features.

[0049] Secondly, an individual patient head model is established, and the patient's anatomical and geometric features are extracted. The individual patient head model is established from the patient's head MRI and CT cross-sectional images, preferably using a finite element model or voxel model. The anatomical feature values ​​are obtained by measuring the thickness of the scalp 105, subcutaneous tissue 106, skull 107, cerebrospinal fluid 109, gray matter 111, and the radius of curvature of gray matter 111 in the detection target area through image processing. Based on the actual position of the near-infrared detection cap detector 104 worn by the patient, the parameters of the light source 101-detector 104 pair, as well as their distance and relative angle, are extracted as geometric feature values. The retrieval expression is formed using the value of the feature label. Finally, the patient's feature label is compared with the feature label in the maximum probability streamline database. The maximum probability streamline corresponding to the closest individual patient head model is found from the database through nearest neighbor search or multi-feature weighted matching.

[0050] Preferably, patients with stroke, traumatic brain injury, brain tumor, or post-neurosurgery undergo routine head MRI or CT scans upon admission. These imaging data are stored in the hospital's PACS system and can be accessed at any time after authorization. Based on existing head MRI or CT images in the patient's medical record database, the following bony landmarks are selected as localization benchmarks: the junction of the frontal and nasal bones at the nasal root, clearly visible on sagittal images, corresponding to the palpable depression at the root of the nasal bridge; the most prominent bony prominence behind the external occipital protuberance, easily identifiable on axial or sagittal images, corresponding to the most prominent bony point on the posterior occipital bone; the bony depressions at the root of the zygomatic arch and in front of the tragus on bilateral preauricular points, visible on axial CT scans, corresponding to the palpable depression in front of the external tragus; and the bony prominence above the orbit at the upper edge of the eyebrow, clearly visible on coronal or sagittal images, corresponding to the bony contour of the upper edge of the external eyebrow. After marking these reference points on the images, the anterior-posterior midline of the head (the line connecting the root of the nose to the external occipital protuberance) and the lateral lines (the line connecting the preauricular points on both sides) can be determined. Then, according to the international 10-20 system ratio or by direct measurement, the projection position of the prefrontal cortex detection area (such as Fp1 and Fp2 points) on the scalp surface can be calculated. Marking this position on the patient's actual head guides the wearing of the front, lateral, and posterior edges of the near-infrared detection cap, ensuring that the light source-detector array is accurately aligned with the target brain region. This method uses clearly identifiable bony landmarks on images as positioning anchors for wearing the infrared detection cap, which is more objective and repeatable than simply relying on surface touch or fixed distances, and can fully utilize the patient's existing imaging data for individualized positioning. Compared to positioning methods estimated by surface touch, using bony landmarks on MRI or CT to position and wear the infrared detection cap offers an order of magnitude higher accuracy and repeatability. As the patient's condition changes, such as postoperative edema subsiding or skull defect repair, follow-up images can be retrieved at any time to update the localization, truly achieving dynamic individual monitoring of local brain tissue oxygen saturation. This method requires no additional examinations or procedures and incurs zero extra cost. Existing patient imaging data can be directly reused, especially head images taken during hospitalization as needed. If, during the acquisition of head MRI and CT cross-sectional images and near-infrared brain oxygenation data, the spatial relationship between the head cross-sectional data and the light source 101-detector 104 pair of the near-infrared detection cap is established using skin texture features and a dual-mode positioning sticker, the localization effect is even better. Specific operating procedures: Record the skin texture features of the positioning points and affix the dual-mode positioning sticker; conduct MRI and CT scans to acquire head cross-sectional data, which must include information on at least three non-collinear positioning points; wear the near-infrared detection cap according to the marked positions on the dual-mode positioning sticker, acquire the raw near-infrared light signal, and record the geometric positions of the light source and photodetector on the detection cap.

[0051] Third, database retrieval and matching compares the patient's anatomical and geometric features with the feature labels in the maximum probability streamline database. Through nearest neighbor search or multi-feature weighted matching, the closest head model instance and its corresponding maximum probability streamlines 102 (850nm near-infrared) and 103 (760nm near-infrared) are found in the database. Curve deformation and mapping: The selected maximum probability streamlines 102 (850nm near-infrared) and 103 (760nm near-infrared) are mapped from the head model instance to the patient's individual head model space. Linear scaling: Based on the ratio of the patient's individual head model to the head model instance in key anatomical features (such as skull thickness 107), the scaling method is applied. For example, after anisotropic scaling and displacement of the maximum probability streamline, it is superimposed on the patient's individual head model; non-rigid registration method: or non-rigid image registration is performed between the head contour of the head model instance and the head contour of the patient's individual head model to obtain a deformation field; this deformation field is applied to the maximum probability streamline to make it "fit" onto the anatomical structure of the patient's individual head model; visualization and output: the mapped maximum probability streamline 102 of 850nm near-infrared light and maximum probability streamline 103 of 760nm near-infrared light are displayed and superimposed on the patient's individual head model to complete the simulation of drawing the maximum probability streamline 102 of 850nm near-infrared light and the maximum probability streamline 103 of 760nm near-infrared light.

[0052] Fourth, calculate the length between intersection points. Calculate the intersection points of the maximum probability streamlines 102 and 103 of 850nm near-infrared light with the boundaries of each layer of the patient's individual head model. If the clarity of the MRI and CT cross-sectional images is high, the intersection points of the skin layer, subcutaneous tissue layer, galea aponeurotica and periosteum layer, compact bone layer of the skull, cancellous bone layer of the skull, periosteum and dura mater layer, arachnoid mater and subarachnoid space layer and pia mater, and gray matter layer with the maximum probability streamlines can be obtained. If the clarity of the MRI and CT cross-sectional images is average, the intersection points of the scalp layer 105, subcutaneous tissue layer 106, skull layer 107, cerebrospinal fluid layer 109, and gray matter layer 111 with the maximum probability streamlines 102 and 103 of 850nm and 760nm near-infrared light can be obtained. Calculate the curve length segment by segment. Preferably, the number of pixels between two adjacent intersection points is calculated along the maximum probability streamline 102 of 850nm near-infrared light and the maximum probability streamline 103 of 760nm near-infrared light, respectively. Then, the actual size of the pixels is calibrated, and the length of the segmented curve of the maximum probability streamline of near-infrared light in the patient's head profile is calculated based on the physical size corresponding to the pixels. Based on curve length Substitute the optical parameters to calculate the optical path factor. (i.e., differential path length factor, DPF).

[0053] Fifth, calculate the change in blood oxygen concentration. Optical path factor. Combined with near-infrared light density change measurement values Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

[0054] Example 2:

[0055] The following describes the establishment of a multilayer optical transmission simulation model based on the exponential decay model according to the diffusion approximation theory, with a wavelength of... When, the change in optical density The change in blood oxygen concentration in the local brain tissue is driven by both changes in oxyhemoglobin and deoxyhemoglobin concentrations, and this driving effect is regulated by three factors: molar absorptivity (material nature), optical path length (device parameter), and optical path factor (environmental interference). The process of calculating the change in local brain tissue blood oxygen concentration by applying the modified Lambert-Beer law and substituting the optical path factor calibrated using the individual patient head model is as follows:

[0056] Step 1: Calculate the effective penetration depth of each layer:

[0057] , (1)

[0058] For the i-th layer of medium at wavelength The effective penetration depth below; The wavelength of near-infrared light signals, such as Equal to 760nm Equal to 850nm; For the i-th layer of medium at wavelength The absorption coefficient at the specified depth; For the i-th layer of medium at wavelength The reduced scattering coefficient under the given conditions, , These are the optical parameters mentioned above.

[0059] The influence of bone mineral density on the optical parameters of the skull: the denser the bone and the better the mineralization, the higher the bone density. The more intense the scattering of light by the bone's internal microstructure, such as trabeculae, the greater the impact of bone density. The larger the value, the lower the scattering. Conversely, in osteoporosis, bone density decreases, and scattering weakens. It also decreases accordingly. (Bone density) (unit The reduced scattering coefficient mapped to the skull layer (unit Empirical formula:

[0060] (2)

[0061] This formula describes the degree to which light deviates from its original direction due to scattering within the skull, establishing a linear relationship between bone density and scattering ability. 3.18 and 0.96 are empirical coefficients fitted from the literature. For example, patient 1 (healthy, (Approximately 1.6) Approximately 6.05 Patient 2 (Osteoporosis) (Approximately equal to 1.0): Approximately 4.14 Patient 3 (high bone density) (Approximately 2.0) Approximately 7.32 Quantitative computed tomography (QCT) uses standard phantom imaging data to calibrate trabecular bone X-ray CT imaging data, quantitatively detects the mineral content of trabecular bone, and then calculates bone mineral density. .

[0062] Step 2: Calculate the effective optical thickness of each single layer:

[0063] (3)

[0064] For the i-th layer of medium at wavelength Effective optical thickness below;

[0065] The maximum probability streamline segment length for near-infrared light in the head profile includes the optical path lengths at both the incident and exit ends; i equals 4, 5, or 8. MRI and CT profiles have high clarity and many layers, while MRI and CT profiles generally have low clarity and few layers. Skin layer: contains only epidermis and dermis, excluding hair; Subcutaneous tissue layer: pure subcutaneous fat and connective tissue; Galea aponeurotica and periosteum layer: dense connective tissue membrane-like structure; Compact bone layer of skull: the outer layer of skull with high mineralization hard tissue; Cancellous bone layer of skull: the middle layer of skull containing medullary cavity tissue, the core adapting layer of the bone density formula; Inner periosteum and dura mater layer: the fibrous membrane between the inner layer of skull and the cerebrospinal fluid layer; Arachnoid mater and subarachnoid space layer: pure cerebrospinal fluid layer; Pia mater and brain layer: pia mater and gray / white matter, with gray matter being the core target layer for near-infrared detection.

[0066] Step 3: Calculate the total optical thickness :

[0067] ; (4)

[0068] Step 4: Calculate the optical path factor :

[0069] , (5)

[0070] in: It is the optical path factor (i.e., differential path length factor, DPF). , , These are globally accepted fitting coefficients. , , ;

[0071] Step 5: Calculate the correction factor :

[0072] (6)

[0073] Step Six: Calculate the change in oxyhemoglobin concentration using the modified Lambert-Beer Law. and changes in deoxyhemoglobin concentration Calculate the local brain tissue oxygen saturation using formula 9. .in, The absolute concentration of oxyhemoglobin and This refers to the absolute concentration of deoxyhemoglobin. , Baseline concentration:

[0074] (7)

[0075] (8)

[0076] (9)

[0077] Solve the system of equations:

[0078] (10)

[0079] (11)

[0080] get:

[0081] (12)

[0082] (13)

[0083] in, wavelength Below, the molar absorptivity of oxyhemoglobin (describes a substance's ability to absorb light of a specific wavelength; it is a characteristic constant of the substance). wavelength Below, the molar absorptivity of deoxyhemoglobin; , These are also the optical parameters mentioned above. This represents the change in oxyhemoglobin concentration. The change in deoxyhemoglobin concentration; This refers to the change in optical density (change in absorbance). This represents the distance between the light source and the detector.

[0084] Existing instruments are based on the population average head model assumption. Approximately equal to The blood oxygen concentration change calculation of the present invention differs from existing instruments: it quantifies individual anatomical differences into optical path factors through a multilayer optical transmission simulation model. Not equal to By embedding it into the classic algorithm to correct the Lambert-Beer law, the calculation deviation caused by differences in tissue thickness and structure is eliminated in principle, thus providing a methodological basis for obtaining more reliable cerebral blood oxygenation test results.

[0085] Example 3:

[0086] like Figure 2 As shown, Examples 1 and 2 demonstrate a method for quantifying individual anatomical differences into optical path factors and calculating changes in local brain tissue oxygenation concentration using a multilayer optical transmission simulation model. The following describes an infrared detection device applying this method. Near-infrared light signal data and layered optical path factor data from the head are read onto a server, and the modified Lambert-Beer law is used to calculate the change in local brain tissue oxygenated hemoglobin concentration. and changes in deoxyhemoglobin concentration The optical path factor is calibrated by reading the patient's recent skull density data, thereby calibrating the change in local brain tissue oxyhemoglobin concentration. and changes in deoxyhemoglobin concentration Read in biochemical blood gas test data and calibrate total hemoglobin concentration. Carotid artery oxygen saturation With jugular venous oxygen saturation ; Read in cervical artery and vein blood oxygen data to calculate the absolute concentration of local brain tissue at any given time. , and blood oxygen saturation .

[0087] like Figure 3 As shown, the near-infrared detection device for local brain tissue oxygen saturation includes a streamline database establishment module, a concentration change calculation module, a brain oxygen saturation calculation module, and a report generation module;

[0088] The streamline database module comprises four sub-modules: the head model set sub-module collects brain maps of preset classification dimensions and generates a three-dimensional head model containing anatomical and geometric features; the Monte Carlo simulation sub-module runs Monte Carlo simulation software to simulate the transmission process of near-infrared photons in the head model; the maximum probability streamline extraction sub-module generates a large set of simulated photon paths for light source-detector pairs and extracts the maximum probability streamlines; and the database storage sub-module stores the streamlines and their features together in the database.

[0089] The concentration change calculation module includes a patient head model creation submodule, a database retrieval and matching submodule, an intersection length calculation submodule, a skull density calibration submodule, and a blood oxygen concentration change calculation quantum module, used to calibrate the optical path factor and calculate the local brain tissue oxyhemoglobin concentration change. and changes in deoxyhemoglobin concentration The system includes a patient head model creation submodule, which reads MRI and CT cross-sectional images of the patient's head to create a head model containing anatomical and geometric features; a database retrieval and matching submodule, which reads the patient's anatomical and geometric features and retrieves the closest maximum probability streamline from a maximum probability streamline database; and a method using nearest neighbor search or multi-feature weighted matching to find the closest maximum probability streamline from the database, which is then superimposed on the individual patient head model using linear scaling or non-rigid registration. Figure 1 As shown, the submodule for calculating the length between intersection points calculates the curve length between the intersection points of the maximum probability streamline and the tissue layers of the patient's head based on the physical dimensions corresponding to the pixels. The skull density calibration submodule is used to query a patient's recently measured skull density. Calibrate the optical path factor This allows for the calibration of changes in local brain tissue oxygenated hemoglobin concentration. and changes in deoxyhemoglobin concentration The quantum module calculates changes in blood oxygen concentration by substituting the curve length. Calculate the optical path factor using optical parameters Combined with near-infrared light density change measurement values Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

[0090] The cerebral blood oxygen saturation calculation module includes a biochemical blood gas verification submodule, a baseline concentration calibration submodule, and a blood oxygen saturation calculation submodule. After calibrating the biochemical blood gas parameters and baseline concentration, it calculates the local brain tissue blood oxygen saturation at any given time. .

[0091] The biochemical blood gas verification submodule is used to read the total hemoglobin concentration. Carotid artery oxygen saturation and venous oxygen saturation The biochemical blood gas analysis submodule reads total hemoglobin concentrations from peripheral veins immediately upon admission, daily after surgery, or when the patient's condition changes. Blood samples were collected after wearing a near-infrared detection cap; venous blood gas analysis values ​​were obtained. This test can be used as input for patients with chronic obstructive pulmonary disease, acute respiratory distress syndrome, heart failure, and carbon monoxide poisoning. Otherwise, consult the doctor using this software to determine if peripheral venous blood should be drawn for venous blood gas analysis to measure venous oxygen saturation. Otherwise, default values ​​derived from experience should be provided based on the patient's age, gender, and condition, and the data source should be noted in the test results.

[0092] The biochemical blood gas verification submodule reads blood samples continuously while wearing a near-infrared detection cap, and analyzes arterial blood oxygen saturation in arterial blood gas analysis. For this test, the following inputs should be used: for patients with chronic obstructive pulmonary disease, acute respiratory distress syndrome, heart failure, and carbon monoxide poisoning; otherwise, the peripheral oxygen saturation measured by the earlobe transillumination pulse oximeter at the moment the near-infrared detection cap is worn. Otherwise, read the value measured by the finger-clip pulse oximeter when the near-infrared detection cap is worn. Otherwise, read the value measured by the lip transilluminator at the moment the near-infrared detection cap is worn. The value is given when all of the above data sources are unreadable. Use the default value of 98%, and annotate the data source in the test results. Use a non-invasive pulse oximeter. The measured value is used as the carotid artery oxygen saturation. Under conditions of normal cardiopulmonary function and good peripheral perfusion, When blood oxygen saturation is greater than or equal to 90%, the error compared with arterial blood gas analysis is usually within ±2%, and in this invention... This error, which accounts for only about 25% of the weight in the weighted calculation of local brain tissue oxygen saturation, is propagated to the final result. The result is less than 1%, which is acceptable in engineering terms for detection error. However, in cases of hypoperfusion, hypothermia, anemia, or elevated carboxyhemoglobin, Deviating from reality The sources of data used in the calculation process should be indicated in the test report to help doctors understand the accuracy of the test data.

[0093] The baseline concentration calibration submodule is used to read the instantaneous local brain tissue oxygen saturation. and total hemoglobin concentration The absolute concentration was calculated. and Then, the baseline concentration of the individual was calculated. , Instantaneous local brain tissue oxygen saturation The measurement and calculation process is as follows: Based on the extraction of local cerebral hemodynamic parameters from MRI, cerebral blood flow perfusion maps are reconstructed using multi-sequence MRI combined with high-resolution T1 / T2 structural images. Within the target region (e.g., at the 10 mm³ voxel level), the arteriovenous blood volume ratio is estimated through the following steps to obtain the total cerebral blood volume. The ratio of arteriolar to venous distribution density was estimated by using the difference between arterial arrival time and venous outflow time, combined with a two-compartment tracer kinetic model, and deconvolved to obtain the relative contributions of the arterial and venous components. Simultaneously, magnetic susceptibility-weighted imaging or quantitative magnetic susceptibility mapping of deoxyhemoglobin-sensitive sequences was used to enhance the spatial identification of venous density. The arteriovenous blood volume components were calculated: under the assumption of a linear positive correlation between local vascular density and blood volume, they were decomposed using the following formula:

[0094] (14)

[0095] (15)

[0096] in Arterial blood vessel density index, It is the venous density index (dimensionless relative value). Arterial blood volume ratio, The ratio of venous blood volume is linearly applicable to microvascular beds within the normal physiological range, but requires further calibration in cases of severe edema or tumor angiogenesis. The carotid artery oxygen saturation is also considered. With venous oxygen saturation As a boundary condition, the instantaneous local brain tissue oxygen saturation at the moment of blood sampling is obtained by substituting it into Formula 16. :

[0097] (16)

[0098] Total hemoglobin concentration The measurement and calculation process is as follows: The total hemoglobin concentration is obtained from the patient's routine blood test upon admission. The unit is g / dL or g / L. Convert it to molar concentration:

[0099] (17)

[0100] in: Unit conversion factor: 1 g / dL Hb is approximately equal to 155 μM (as a tetramer). This is a correction factor for cerebral hematocrit. Due to the Fåhraeus effect, the hematocrit of cerebral capillaries is lower than that of large vessels. Therefore, directly checking the hemoglobin value in peripheral blood reveals it to be higher than the total hemoglobin concentration in brain tissue. The recommended correction factor is 0.8, which can be adjusted according to the patient's specific condition such as anemia and dehydration.

[0101] Baseline concentration , The measurement and calculation process is as follows:

[0102] (18)

[0103] (19)

[0104] (20)

[0105] (twenty one)

[0106] Instantaneous local brain tissue oxygen saturation Substituting into Equation 19, we solve the system of equations 19 and 18 simultaneously to obtain the absolute concentration at the time of blood collection. and Then, substitute the values ​​into formula 20 or 21 to obtain the baseline concentration at the time of blood collection. , .

[0107] The blood oxygen saturation calculation submodule is used to calculate the blood oxygen saturation of local brain tissue at any given time. The calculation process is as follows: the change in oxyhemoglobin concentration measured in real time by the near-infrared detection cap at the moment of blood sampling. and changes in deoxyhemoglobin concentration The baseline concentration is superimposed from formulas 20 and 21. , The above gives the absolute concentration at any given time. and From absolute concentration and Substitute into Formula 19 to calculate the local brain tissue oxygen saturation at any given time. This allows us to obtain only the change in oxyhemoglobin concentration, whereas previously we could only do so. and changes in deoxyhemoglobin concentration Converted to local brain tissue oxygen saturation at any given time for an individual patient. Quantitative detection value.

[0108] The report generation module is used to verify the physiological rationality of the calculated blood oxygen saturation, check whether the time and space of the multi-source data are accurately aligned, and eliminate invalid results that are outside the physical range. Values ​​that are outside the normal physiological reference range but still within the physical range are marked and indicated in the report. The data sources used in the calculation process are marked in the report. Finally, a report on the blood oxygen saturation of local brain tissue containing the detection geometry is generated.

Claims

1. A near-infrared method for detecting local brain tissue oxygen saturation, characterized in that, Includes the following steps: Step 1: Generate a set of three-dimensional head models containing anatomical and geometric features from the collected brain atlas set with preset classification dimensions; Step 2: Run Monte Carlo simulation software to simulate the transmission process of near-infrared photons in the head model and generate a set of simulated photon paths; Step 3: Extract the maximum probability streamline from the simulation path, and store the maximum probability streamline, anatomical features, and geometric features together into the maximum probability streamline database; Step 4: Query the patient's head MRI and CT cross-sectional images to create a patient individual head model that includes anatomical and geometric features. Based on the patient's anatomical and geometric features, retrieve the closest maximum probability streamline from the maximum probability streamline database and overlay it on the patient individual head model. Step 5: Calculate the curve length between the intersection points of the maximum probability streamline and the tissue layers of the patient's head based on the physical dimensions corresponding to the pixels. Calculate the optical path factor Combined with near-infrared light density change measurement values Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

2. The near-infrared detection method for local brain tissue oxygen saturation according to claim 1, characterized in that, In step 4, the closest maximum probability streamline is retrieved from the maximum probability streamline database, which is achieved through nearest neighbor search or multi-feature weighted matching; it is then overlaid on the patient's individual head model, which is achieved through linear scaling or non-rigid registration.

3. The near-infrared detection method for local brain tissue oxygen saturation according to claim 2, characterized in that, Step 5 also includes: querying the patient's recently measured cranial bone density data and calibrating the optical path factor. This allows for the calibration of changes in local brain tissue oxygenated hemoglobin concentration. and changes in deoxyhemoglobin concentration .

4. The near-infrared detection method for local brain tissue oxygen saturation according to claim 3, characterized in that, Step 5 also includes: querying biochemical blood gas test data and calibrating the total hemoglobin concentration. Carotid artery oxygen saturation and jugular venous oxygen saturation .

5. The near-infrared detection method for local brain tissue oxygen saturation according to claim 4, characterized in that, Step 5 also includes: Based on local brain tissue MRI structural images and cerebral blood flow perfusion maps, using carotid artery oxygen saturation and venous oxygen saturation As a boundary condition, combined with the arterial and venous blood volume components, the instantaneous local brain tissue oxygen saturation is calculated. ; From instantaneous and Calculate the baseline concentration , And then and Calculate the absolute concentration of oxyhemoglobin in local brain tissue relative to the absolute concentration of deoxyhemoglobin This allows for the calculation of local brain tissue oxygen saturation at any given time. .

6. The near-infrared detection method for local brain tissue oxygen saturation according to claim 5, characterized in that, Step 5 further includes: calculating the local brain tissue blood oxygen saturation. Physiological rationality verification is performed, and invalid results that exceed the physical range are eliminated. Values ​​that exceed the normal physiological reference range but are still within the physical range are marked and anomalies are indicated in the report. Finally, a local brain tissue blood oxygen saturation test report containing the detection geometry is generated.

7. A near-infrared detection device for local brain tissue oxygen saturation, used to implement the detection method according to any one of claims 1 to 6, characterized in that, This includes modules for establishing a streamline database, calculating concentration changes, calculating cerebral oxygen saturation, and generating reports. The streamline database module is used to generate a three-dimensional head model set containing anatomical and geometric features from the collected brain map set with preset classification dimensions, and to extract the maximum probability streamline and establish a streamline database through Monte Carlo simulation. The concentration change calculation module is used to calibrate the optical path factor to calculate the change in local brain tissue oxyhemoglobin concentration. and changes in deoxyhemoglobin concentration ; The cerebral oxygen saturation calculation module is used to calculate the local brain tissue oxygen saturation at any given time after calibrating biochemical blood gas parameters and baseline concentrations. ; The report generation module is used to verify the physiological rationality of the calculated local brain tissue oxygen saturation and generate a report containing the local brain tissue oxygen saturation at the detection geometric location. Test report.

8. The near-infrared detection device for local brain tissue oxygen saturation according to claim 7, characterized in that, The streamline database establishment module includes a header model set establishment submodule, a Monte Carlo simulation submodule, a maximum probability streamline extraction submodule, and a database storage submodule: The head model set submodule is used to generate a three-dimensional head model set containing anatomical and geometric features from the collected brain map set with preset classification dimensions. The Monte Carlo simulation submodule is used to run Monte Carlo simulation software to simulate the transmission process of near-infrared photons in the head model and generate a set of simulated photon paths. The maximum probability streamline extraction submodule is used to extract the maximum probability streamline by generating a set of simulated photon paths for a light source-detector pair. The database storage submodule is used to store the maximum probability streamlines, anatomical features, and geometric features together into the database to establish a maximum probability streamline database.

9. The near-infrared detection device for local brain tissue oxygen saturation according to claim 8, characterized in that, The concentration change calculation module includes a patient head model establishment submodule, a database retrieval and matching submodule, an intersection length calculation submodule, a skull density calibration submodule, and a blood oxygen concentration change calculation quantum module. A submodule for establishing a patient head model is used to query the patient's head MRI and CT cross-sectional images to create an individual patient head model that includes anatomical and geometric features. The database retrieval and matching submodule is used to retrieve the closest maximum probability streamline from the maximum probability streamline database based on the patient's anatomical and geometric features, and overlay it on the individual patient's head model. The submodule for calculating the length between intersection points is used to calculate the curve length between the intersection points of the maximum probability streamline and the tissue layers of the patient's head, based on the physical dimensions corresponding to the pixels. ; The cranial bone density calibration submodule is used to query the patient's recently measured cranial bone density data to calibrate the optical path factor. The quantum module for calculating changes in blood oxygen concentration is used to input the curve length. Calculate the optical path factor using optical parameters Combined with near-infrared light density change measurement values Based on the modified Lambert-Beer law, the change in local brain tissue oxyhemoglobin concentration was calculated. and changes in deoxyhemoglobin concentration .

10. The near-infrared detection device for local brain tissue oxygen saturation according to claim 9, characterized in that, The cerebral blood oxygen saturation calculation module includes a biochemical blood gas verification submodule, a baseline concentration calibration submodule, and a blood oxygen saturation calculation submodule: The biochemical blood gas calibration submodule is used to calibrate the total hemoglobin concentration. Carotid artery oxygen saturation and jugular venous oxygen saturation ; The baseline concentration calibration submodule is based on the calibrated total hemoglobin concentration. Carotid artery oxygen saturation and venous oxygen saturation Provides individualized baseline concentration for near-infrared detection calculations. , ; The blood oxygen saturation calculation submodule is based on the calibrated baseline concentration. , Changes in oxyhemoglobin concentration and changes in deoxyhemoglobin concentration This allows for the calculation of local brain tissue oxygen saturation at any given time. .

Images