Knowledge graph construction system for brain function abnormalities based on near-infrared brain function imaging

Abstract

The application discloses a brain function abnormal knowledge graph construction system based on near-infrared brain function imaging, and relates to the technical field of knowledge graphs.The system comprises the following steps: an acquisition module, which is used to acquire reference brain oxygen data of a target object in an initial examination time period and reexamination brain oxygen data of the target object in a reexamination time period; a determination module, which is used to determine a spatial distance normalization value based on a pixel distance between reference grid data and reexamination grid data, and construct a cost matrix based on an association relationship between the reference grid data and the reexamination grid data and the spatial distance normalization value; an input module, which is used to input the cost matrix into a global distribution algorithm to obtain homologous brain region mapping data; and a construction module, which is used to determine a metabolic difference based on the homologous brain region mapping data, and construct a brain function knowledge graph based on the metabolic difference.The application achieves the technical effect of reducing the occurrence of matching failure or one-to-many confusion of homologous brain regions.

Description

Technical Field

[0001] This application relates to the field of knowledge graph technology, specifically to a system for constructing a knowledge graph of brain functional abnormalities based on near-infrared brain functional imaging. Background Technology

[0002] Near-infrared brain imaging (NIBMI) is commonly used to monitor local blood oxygen metabolism changes in the cerebral cortex. Long-term follow-up with NIBMI requires multiple administrations of the target subject wearing an NIBMI cap for optical density data acquisition. However, repeated cap use inevitably introduces two-dimensional planar positional deviations. When constructing medical knowledge graphs, existing inter-period comparison methods using in-situ grid truncation introduce geometric misalignment errors. Furthermore, relying on absolute blood oxygen concentration for proximity matching can easily lead to matching failures or many-to-one errors in homologous brain regions due to sudden changes in local oxygen consumption caused by disease evolution. Summary of the Invention

[0003] To address the technical problem in related technologies where in-situ mesh cutting introduces geometric misalignment errors, which can easily lead to matching failures or many-to-one errors in homologous brain regions, this application provides a brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging.

[0004] The specific technical solution adopted is as follows: The acquisition module is used to acquire the baseline brain oxygenation data of the target object during the initial examination period and the re-examination brain oxygenation data during the re-examination period. The baseline brain oxygenation data includes multiple baseline grid data, and the re-examination brain oxygenation data includes multiple re-examination grid data. The determination module is used to determine the spatial distance normalization value based on the pixel distance between the reference grid data and the review grid data, and to construct the cost matrix based on the correlation between the corresponding waveform sequences of the reference grid data and the review grid data, as well as the spatial distance normalization value. The input module is used to input the cost matrix into the global allocation algorithm to obtain the homologous brain region mapping data between the baseline grid data and the review grid data; The module is used to determine the metabolic difference between the baseline grid data and the review grid data based on the homologous brain region mapping data, and to construct a brain function knowledge graph based on the evolutionary tags corresponding to the metabolic difference.

[0005] In one possible implementation of this application, the system further includes: The partitioning module is used to obtain the first optical density data stream of the target object during the initial inspection period and the second optical density data stream during the re-inspection period, and to divide the image data corresponding to the first optical density data stream and the second optical density data stream into equal amounts of baseline grid data and re-inspection grid data, respectively. The generation module is used to perform waveform analysis processing on the baseline grid data and the review grid data to generate baseline cerebral blood oxygenation data and review cerebral blood oxygenation data.

[0006] In one possible implementation of this application, the generation module includes: The first generation submodule is used to perform waveform analysis processing on the baseline grid data to generate baseline cerebral blood oxygenation data; The second generation submodule is used to perform waveform analysis processing on the re-examination grid data to generate re-examination cerebral blood oxygenation data.

[0007] In one possible implementation of this application, the first generation submodule is specifically used for: Based on the blood oxygen concentration values ​​of pixels in each benchmark grid data, a blood oxygen time series is constructed, and the mean blood oxygen concentration and the standard deviation of physiological fluctuations of the blood oxygen time series are calculated. The blood oxygen time series was filtered to obtain a low-frequency fluctuation signal; The low-frequency fluctuation signal is standardized to obtain a standardized waveform data sequence. The baseline cerebral blood oxygenation data is obtained by integrating the standardized waveform data sequence, the center coordinates of the baseline grid data, the mean baseline blood oxygenation concentration, and the standard deviation of physiological fluctuations.

[0008] In one possible implementation of this application, the determining module is specifically used for: Extract the first center coordinates of each baseline grid data in the baseline cerebral blood oxygenation data, and the second center coordinates of each review grid data in the review cerebral blood oxygenation data; Based on the difference between the first center coordinates and the second center coordinates, the pixel distance between the reference grid data and the review grid data is calculated; When the pixel distance meets the preset conditions, the pixel distance is normalized to obtain the spatial distance normalized value.

[0009] In one possible implementation of this application, the determining module is further configured to: Compare the pixel distance with a preset maximum offset threshold; If the pixel distance is greater than the preset maximum offset threshold, it is determined that the current re-examined grid data and its corresponding matching baseline grid data do not belong to the same anatomical brain region, and the calculation is skipped; If the pixel distance is less than or equal to the preset maximum offset threshold, it is determined that the current review grid data and its corresponding matching baseline grid data belong to the same anatomical brain region, and the pixel distance meets the preset condition.

[0010] In one possible implementation of this application, the determining module is further configured to: The standardized waveform data sequence of each baseline grid data in the baseline cerebral blood oxygenation data and the re-examination waveform data sequence of each re-examination grid data in the re-examination cerebral blood oxygenation data were extracted. Calculate the Pearson correlation coefficient between the standardized waveform data sequence and the reviewed waveform data sequence, and convert the Pearson correlation coefficient into the waveform difference cost between the baseline grid data and the reviewed grid data. The joint cost is obtained by weighted summation of the waveform difference cost and the normalized spatial distance value. The cost matrix is ​​constructed based on the joint cost value between the corresponding combinations of each baseline grid and the review grid.

[0011] In one possible implementation of this application, the input module is specifically used for: The cost matrix is ​​reduced by a global allocation algorithm to obtain a bijective set that includes multiple matching results; The bijective set is resolved into multiple homologous brain region mapping data between the baseline grid data and the review grid data, wherein any homologous brain region mapping data includes a set of matched baseline grid data and review grid data.

[0012] In one possible implementation of this application, the construction module is specifically used for: For any homologous brain region mapping data, extract the mean baseline blood oxygen concentration of the baseline grid data in the homologous brain region mapping data, and the mean re-examination blood oxygen concentration of the re-examination grid data; The metabolic difference between the baseline grid data and the repeat grid data is determined based on the difference between the mean blood oxygen concentration of the follow-up examination and the mean blood oxygen concentration of the baseline.

[0013] In one possible implementation of this application, the building module is further configured to: The standard deviation of physiological fluctuations in the baseline cerebral blood oxygenation data is extracted, and the adaptive interval range is calculated based on the standard deviation of physiological fluctuations. If the metabolic difference is within the adaptive range, then the target object's metabolic difference is determined to be within the normal physiological range, and the brain function knowledge graph is not updated. If the metabolic difference is higher than the upper limit of the adaptive range, then the evolution label of the metabolic difference in the baseline grid data is determined as the compensatory evolution label. If the metabolic difference is lower than the lower limit of the adaptive range, then the evolution label of the metabolic difference in the baseline grid data is determined as the degenerative evolution label. Each baseline grid data or review grid data is used as an anatomical entity node, and the anatomical entity nodes in the same brain region are connected by directed edges based on the compensatory evolution label or the degenerative evolution label. The metabolic difference is written to each directed connection edge to construct a brain function knowledge graph.

[0014] This application has, but is not limited to, the following technical effects: In this application, baseline cerebral oxygenation data of the target object during the initial examination period and baseline cerebral oxygenation data during the follow-up examination period are obtained. Based on the pixel distance between the baseline and follow-up grid data, a spatial distance normalization value is determined. A cost matrix is ​​constructed based on the correlation between the corresponding waveform sequences of the baseline and follow-up grid data, and the spatial distance normalization value. This cost matrix is ​​then input into a global allocation algorithm to obtain homologous brain region mapping data between the baseline and follow-up grid data. Furthermore, based on the homologous brain region mapping data, the metabolic difference between the baseline and follow-up grid data is determined. Based on the evolutionary labels corresponding to the metabolic difference, a brain function knowledge graph is constructed. Through joint bijective allocation using pixel distance and waveform similarity, accurate locking of homologous anatomical brain regions is achieved without rigid body registration, ensuring that the finally extracted metabolic difference truly reflects the pathological evolution trajectory. Furthermore, based on the evolutionary labels corresponding to the metabolic difference, a brain function knowledge graph is constructed, reducing the occurrence of matching failures or many-to-one errors in homologous brain regions. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the first embodiment of the brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging of this application. Figure 2 This is a schematic diagram of the system architecture involved in the brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging in this application; Figure 3 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application. Detailed Implementation

[0016] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0017] This application provides a system for constructing a knowledge graph of brain dysfunction based on near-infrared brain functional imaging. In the first embodiment of this system, referring to... Figure 1 and Figure 2 The system includes: The acquisition module 101 is used to acquire the baseline brain blood oxygen data of the target object during the initial examination period and the re-examination brain blood oxygen data during the re-examination period. The baseline brain blood oxygen data includes multiple baseline grid data, and the re-examination brain blood oxygen data includes multiple re-examination grid data.

[0018] As an example, the initial examination period can be a time period for the first diagnosis of the target object. The data obtained during this period serves as the benchmark for the subsequent knowledge graph. The benchmark brain blood oxygenation data is obtained by analyzing and calculating the optical density signal stream with discrete channel distribution output by the near-infrared imaging device. The benchmark grid data is obtained by dividing the optical density signal stream into a two-dimensional dynamic blood oxygenation image sequence. Specifically, the effective bounding box of the dynamic blood oxygenation image sequence within the anatomical projection range of the brain is divided into multiple square grids of equal size, thus obtaining multiple benchmark grid data. The benchmark grid data is also the multiple image grids obtained after dividing an image. The same applies to the re-examination grid data.

[0019] As an example, the follow-up period is the time period for follow-up of the target object, used for same-source comparison and updating. The data types contained in the follow-up brain oxygenation data are the same as those in the baseline brain oxygenation data, only the time period in which the data is acquired is different. The method of acquiring the follow-up grid data is the same as the division method of the baseline grid data mentioned above, and will not be elaborated here.

[0020] Prior to the steps of acquiring the baseline cerebral blood oxygenation data of the target subject during the initial examination period and the follow-up cerebral blood oxygenation data during the follow-up examination period, the system also includes: The segmentation module is used to acquire the first optical density data stream of the target object during the initial inspection period and the second optical density data stream during the re-inspection period, and to divide the image data corresponding to the first optical density data stream and the second optical density data stream into equal amounts of baseline grid data and re-inspection grid data, respectively.

[0021] As an example, to ensure the frequency periodicity integrity of subsequent low-frequency signal feature extraction, the system extracts signals with a continuous sampling duration greater than 300 seconds as the first optical density data stream. The second optical density data stream is acquired in a similar manner, only with a different acquisition time period. The acquired first or second optical density data stream is a phase-locked event-related blood oxygenation signal under task-state stimulation. The blood oxygenation time series is truncated and aligned along the time axis using the trigger timestamp of the external task stimulus. A spherical spline interpolation algorithm is then called to perform spatial two-dimensional interpolation processing on the input source. Through this interpolation processing, the system transforms the discrete optical density data into a continuous two-dimensional dynamic blood oxygenation image sequence. Subsequently, the system extracts the effective bounding box of this two-dimensional dynamic blood oxygenation image sequence within the brain's anatomical projection range. Within this effective bounding box, the system delineates... The system uses a grid of equal-sized squares, with each square defined as the baseline grid data. Each baseline grid data is assigned a natural number index. ( The method for dividing the grid data for review is the same as described above.

[0022] The generation module is used to perform waveform analysis processing on the baseline grid data and the review grid data to generate baseline cerebral blood oxygenation data and review cerebral blood oxygenation data.

[0023] As an example, waveform analysis is performed on the baseline grid data to obtain baseline brain blood oxygenation data, and waveform analysis is performed on the review grid data to obtain review brain blood oxygenation data.

[0024] The generation module includes: The first generation submodule is used to perform waveform analysis on the baseline grid data to generate baseline brain blood oxygenation data.

[0025] The first generation submodule is specifically used for: Based on the blood oxygen concentration values ​​of pixels in each benchmark grid, a blood oxygen time series is constructed, and the mean blood oxygen concentration and the standard deviation of physiological fluctuations of the blood oxygen time series are calculated.

[0026] As an example, for each reference grid data, the system extracts all image pixels falling within that grid. Over all consecutive sampling times, the system extracts the blood oxygen concentration value for each pixel. For each individual sampling time, the system calculates the arithmetic mean of the blood oxygen concentration values ​​for all pixels within the grid. The system then combines these continuously generated arithmetic means along the time axis in chronological order to generate a one-dimensional blood oxygen time series. .in, This indicates the specific data sampling time point.

[0027] As an example, based on the generated blood oxygen time series The system calculates the arithmetic mean and statistical standard deviation of the data in the time domain, specifically: The system sets the calculated time-domain arithmetic mean as the baseline mean blood oxygen concentration. The higher the mean baseline blood oxygen concentration, the better it is within the initial consultation time period (i.e., the initial examination time period). The higher the overall metabolic and oxygen consumption level of the local grid region, the more the system sets the calculated statistical standard deviation of the data as the physiological fluctuation standard deviation. The larger the standard deviation of this physiological fluctuation, the wider the range of normal fluctuations and ranges of blood oxygen in the local grid area during task execution.

[0028] The blood oxygen time series was filtered to obtain a low-frequency fluctuation signal.

[0029] The low-frequency fluctuation signal is standardized to obtain a standardized waveform data sequence.

[0030] As an example, during the development of mental illness over several months, changes in the condition can lead to a significant increase or decrease in the absolute value of blood oxygen concentration in local brain regions. Therefore, bandpass filtering and standardization operations are used to extract pure waveform features that are not affected by the increase or decrease in absolute values.

[0031] As an example, a one-dimensional blood oxygen time series The input is fed into a preset bandpass filter. The system sets the lower frequency passband limit of the bandpass filter to [value missing]. The upper limit of the frequency passband is set to The system uses this filter to remove high-frequency heartbeat and respiratory physiological interference signals from the sequence, outputting a low-frequency fluctuation signal. .

[0032] Furthermore, regarding this low-frequency fluctuation signal The system uses the Z-score algorithm to perform standardization. For the low-frequency waveform data point corresponding to the current sampling time t, the calculation process is shown in the following formula: in, Indicates the current sampling time The corresponding reference normalized low-frequency waveform data points; This represents the original value of the filtered low-frequency fluctuation signal at the current moment; This represents the arithmetic mean of the low-frequency fluctuation signal over the entire sampling period; This represents the standard deviation of the low-frequency fluctuation signal over the entire sampling period. This calculation method ensures that interference from variations in metabolic scale is eliminated. When the value is less than 0.01, it is determined to be an invalid detection signal (dead zone), and the signal is directly... Sequences marked as all zeros are not included in subsequent correlation calculations.

[0033] Specifically, the numerator By subtracting its own baseline mean, the signal is shifted to a relative fluctuation state around zero, eliminating the base effect of absolute blood oxygen concentration; denominator term As a scaling reference, the intensity of fluctuations between different grids is uniformly mapped to a standard scale. The ratio of these two factors generates a feature sequence that reflects only the functional fluctuation rhythm without containing absolute numerical information. The system sets this continuously generated sequence as a standardized waveform data sequence. .

[0034] The baseline cerebral blood oxygenation data is obtained by integrating the standardized waveform data sequence, the center coordinates of the baseline grid data, the mean baseline blood oxygenation concentration, and the standard deviation of physiological fluctuations.

[0035] As an example, for those who have assigned indexes For each reference grid data, the system extracts the center point of the grid, obtains its longitudinal and transverse coordinates in a two-dimensional plane, and sets these values ​​as the center coordinates of the reference grid data. The standardized waveform data sequence, the center coordinates of the baseline grid data, the mean baseline blood oxygen concentration, and the standard deviation of physiological fluctuations are integrated into the baseline brain blood oxygen data.

[0036] The second generation submodule is used to perform waveform analysis processing on the re-examination grid data to generate re-examination cerebral blood oxygenation data.

[0037] As an example, forcibly applying the valid bounding box from the initial investigation period to the re-examination image, which has a physical offset, will inevitably cause some edge regions of the bounding box to exceed the effective signal coverage range actually interpolated from the re-examination image (i.e., outside the detection field of view of the current photoelectric cap sensor). If features are directly extracted from the grid falling into this blind spot, the system will return an empty set or a non-numeric error result because no pixels can be collected, thus triggering a system-level crash in subsequent correlation coefficient calculations and matrix allocation.

[0038] Based on this, an index was allocated for each Before extracting features, the system performs a validity check on the reviewed grid data. The system first counts the total number of pixels contained in the reviewed grid data. Then, the system traverses these pixels and counts the number of valid pixels that contain valid interpolated blood oxygen concentration data (non-null values).

[0039] The system divides the number of effective pixels by the total number of pixels to obtain the effective interpolation area ratio of the review grid data. The system pre-sets a preset effective area threshold in memory to characterize the bottom line of data reliability (for example, the system sets the empirical value of this threshold to 50%). The system compares the calculated effective interpolation area ratio of each review grid data with the preset effective area threshold. If the effective interpolation area ratio of a certain review grid data is lower than the preset effective area threshold, the system determines that the review grid data is within the follow-up review period. Once the data has been physically removed from the effective monitoring field of view, the system immediately marks the reviewed grid data as an invalid grid and suspends all subsequent feature extraction and time series processing operations for that grid.

[0040] If the effective interpolation area ratio of a certain review grid data is greater than or equal to the preset effective area threshold, the system determines that the review grid data has a qualified amount of data to participate in inter-period comparison, marks it as valid review grid data, and allows it to enter the subsequent feature extraction stage. Then, the review waveform data sequence, the center coordinates of the review grid data, and the mean review blood oxygen concentration are extracted from the review grid data as review brain blood oxygenation data.

[0041] The determination module 102 is used to determine the spatial distance normalization value based on the pixel distance between the reference grid data and the review grid data, and to construct the cost matrix based on the correlation between the corresponding waveform sequences of the reference grid data and the review grid data, as well as the spatial distance normalization value.

[0042] As an example, during the months-long follow-up period, each re-wearing of the laser cap causes unpredictable geometric positional shifts in the two-dimensional plane. If image boundaries are generated independently and meshes are automatically created for each follow-up, the number and position of meshes generated in each follow-up period will be inconsistent with those in the initial diagnosis period. This would prevent the cost matrix used for subsequent comparison calculations from meeting the requirement of equal row and column dimensions. Therefore, the system must introduce the global boundary from the initial diagnosis baseline period to force spatial dimension alignment, and based on this, calculate the normalized value of spatial distance.

[0043] As an example, the spatial distance normalization value can represent the spatial distance between the normalized baseline grid data and the review grid data, reflecting the deviation between the two grid positions.

[0044] As an example, each element in the cost matrix represents the cost required to assign two grids to the same anatomical brain region. The smaller the element value, the higher the probability that the two grids belong to the same physical brain region.

[0045] The step of determining the spatial distance normalization value based on the pixel distance between the reference grid data and the review grid data in module 102 includes: Extract the first center coordinates of each baseline grid data in the baseline brain oxygenation data, and the second center coordinates of each review grid data in the review brain oxygenation data.

[0046] The pixel distance between the reference grid data and the review grid data is calculated based on the difference between the first center coordinates and the second center coordinates.

[0047] As an example, the first center coordinates of each baseline grid data and the second center coordinates of each review grid data in the review brain oxygenation data are first extracted from the baseline brain oxygenation data.

[0048] As an example, the x-coordinates and y-coordinates of the first center coordinates and the second center coordinates are determined, and the sum of the squares of the differences between the x-coordinates and the y-coordinates of the two center coordinates is calculated. The square root of this sum is then taken to obtain the Euclidean line pixel distance between the two points, which is the pixel distance. The larger the pixel distance value, the farther the physical position offset of the two grids on the two-dimensional image is.

[0049] When the pixel distance meets the preset conditions, the pixel distance is normalized to obtain the spatial distance normalized value.

[0050] As an example, after calculating the pixel distance, the pixel distance is verified by offset distance. If it is determined that the two grids do not belong to the same anatomical brain region, the calculation process is skipped. Otherwise, the two grids belong to the same anatomical brain region and the pixel distance meets the preset condition.

[0051] As an example, the normalized value of spatial distance The calculation method can be: in, This represents the normalized spatial distance value after dimensional alignment. and The x and y coordinates of the center coordinates of the reference grid data i and the review grid data j are respectively represented; the numerator constitutes the pixel distance between the two grids; the denominator is the preset maximum offset threshold. , is a constant greater than zero.

[0052] This calculation method ensures that absolute pixel distances are uniformly mapped to... Within the standard probability space. Specifically, the denominator term As a scaling benchmark, it unifies the image offset at different resolutions; the overall ratio term This reflects a positive distance penalty rule: the larger the actual pixel distance (numerator) and the closer it is to the physical tolerance limit (denominator), the closer the normalized value is to... This means that within the valid candidate region, the greater the deviation between two grid positions, the higher the cost during the final weighting.

[0053] Before the step of normalizing the pixel distance to obtain a normalized spatial distance value when the pixel distance meets a preset condition, the method further includes: Compare the pixel distance with the preset maximum offset threshold.

[0054] If the pixel distance is greater than the preset maximum offset threshold, it is determined that the current review grid data and its corresponding matching baseline grid data do not belong to the same anatomical brain region, and the calculation is skipped.

[0055] If the pixel distance is less than or equal to the preset maximum offset threshold, it is determined that the current review grid data and its corresponding matching baseline grid data belong to the same anatomical brain region, and the pixel distance meets the preset condition.

[0056] As an example, to prevent the algorithm from forcibly matching two grids that are extremely far apart (which is completely impossible to produce such a large wearing error) in pursuit of weak waveform similarity, a parameter is pre-set to characterize the upper limit of the maximum possible physical sliding distance of the optical cap, and it is defined as the preset maximum offset threshold. (For example, based on extensive engineering measurement experience, the system sets this constant to twice the sum of the side lengths of two adjacent grids).

[0057] If the system determines that the pixel distance of the Euclidean line is greater than the preset maximum offset threshold This indicates that the grid data is currently being reviewed. It has exceeded the baseline grid data. Within a reasonably plausible range of displacement in physical reality (i.e., the two cannot possibly be the same anatomical brain region), the system immediately performs a physical hard truncation, directly allocating the corresponding cost. The system is given a maximum penalty value, and the current inner loop is skipped, proceeding directly to the next iteration. The calculation.

[0058] If the system determines that the pixel distance of the Euclidean line is less than or equal to the preset maximum offset threshold This indicates that the grid data is currently being reviewed. Located in the baseline grid data If the pixel distance meets the preset conditions within the legal candidate space area, then the pixel distance is within the preset conditions.

[0059] The step of determining the correlation between the waveform sequences corresponding to the baseline grid data and the review grid data, and constructing the cost matrix based on the spatial distance normalization value in module 102, includes: The standardized waveform data sequences of each baseline grid data in the baseline brain oxygenation data and the re-examination waveform data sequences of each re-examination grid data in the re-examination brain oxygenation data were extracted.

[0060] As an example, after completing the spatial dimension legality screening, the system performs time series feature (waveform correlation) operations on the grids belonging to the legal candidate regions, extracts the waveform sequence data corresponding to each baseline grid data and the review grid data, calculates the correlation between any two grids, and standardizes the waveform data sequence to represent the inherent functional fluctuation fingerprint of the brain region, which is not affected by the disease and is used for homologous localization. Calculate the Pearson correlation coefficient between the standardized waveform data sequence and the reviewed waveform data sequence, and convert the Pearson correlation coefficient into the waveform difference value between the baseline grid data and the reviewed grid data.

[0061] As an example, before executing the Pearson correlation formula, the system performs a variance division-to-zero protection check. The system calculates the statistical variance of all sampling points within both the standardized waveform data sequence and the reviewed waveform data sequence. The system then checks whether the variances of these two sequences are zero or below a preset minimum normal number (e.g., ...). If the variance of any sequence is detected to be lower than the minimum normal number (indicating that the grid did not detect any effective blood oxygen fluctuations during the entire sampling period, presenting as a straight dead zone), in order to prevent the Pearson formula from reporting a division-by-zero error when calculating the denominator, the system forces the Pearson correlation coefficient between the grids to be directly assigned a value. If the variances of both sequences are normal, the system will perform a standard Pearson correlation coefficient calculation. The calculation method for the Pearson correlation coefficient is existing technology and will not be elaborated here.

[0062] As an example, after calculating the Pearson correlation coefficient, the system further converts this correlation into waveform difference cost. The calculation method is as follows: in, Represents a standardized waveform data sequence. This indicates a sequence of waveform data to be reviewed; The function output is between The Pearson correlation coefficient between the two normalized waveforms reflects the degree of linear synchronization in the undulation rhythm of the two waveforms.

[0063] This calculation method ensures that similarity is converted into a positive matching penalty cost. Specifically, a constant... Subtract the numerator of the correlation coefficient This achieves an inverted relationship: the more similar two sequences are (their correlation coefficients approach a certain value), the more inverted the relationship becomes. When the difference approaches ), (Indicating extremely low matching cost); when the fluctuations of two sequences are completely opposite (correlation coefficient approaches 0.5%) When the difference approaches ), (This indicates that the matching cost is extremely high). Denominator term As a scaling constant, it makes the overall ratio term Strictly mapped to Within the interval, the dimensions and scale are kept absolutely consistent with the aforementioned spatial distance normalization values, thus avoiding the failure of weights due to excessively large absolute values ​​of a certain feature during subsequent weighting.

[0064] The joint value is obtained by weighted summation of the waveform difference value and the spatial distance normalized value.

[0065] As an example, obtain the normalized values ​​of spatial distances at the same dimensional scale. Difference value from waveform Then, the system introduces a preset first weighting coefficient. With the second weighting coefficient (constraint Since the waveform features induced by the time-locking task are more effective at determining the identity of trans-temporal anatomical brain regions than finite planar distance offsets, the system in this embodiment will... Values ,Will Values .

[0066] The system performs a weighted summation calculation: in, Represents the baseline grid data With review of grid data The joint substitution value between them. This calculation method ensures that within the legal candidate range of spatial displacement, further local optimization is performed through waveform features. Specifically, the first weighted product term Provides the basic traction force for nearest matching; second weighted product term It provides consistency verification of functional attributes; the sum of the two constitutes the comprehensive cost. The smaller this cost value, the greater the probability that the two grids are not only physically close but also produce highly consistent neural functional responses to the same cognitive task, and that they belong to the same physical brain region.

[0067] The cost matrix is ​​constructed based on the joint cost value between the corresponding combinations of each baseline grid and the review grid.

[0068] As an example, the calculated joint cost value Fill the first blank matrix Line number Column, complete After the calculation, the system proceeds in its internal loop. Repeat the above verification, truncation, and calculation operations in sequence to calculate the result. to And fill in the matrix in sequence. Okay, complete the output of the joint value of all reviewed grid data corresponding to the first reference grid data. After the inner traversal loop ends, the system indexes the reference grid data in the outer traversal loop. Restart command The inner loop increments, and so on, until the outer loop reaches... At that time, the system will All joint cost values ​​are filled into an empty matrix, thus outputting a complete two-dimensional cost matrix.

[0069] Input module 103 is used to input the cost matrix into the global allocation algorithm to obtain the homologous brain region mapping data between the baseline grid data and the review grid data.

[0070] As an example, the global allocation algorithm could be the Hungarian algorithm. If a nearest-neighbor pairing rule is used in the cost matrix to find the minimum value for a single row of data, it can easily lead to multiple adjacent and waveform-similar reference grid data pointing to the same column (the same review grid data). This phenomenon causes many-to-one mapping conflicts, violating the objective property that the anatomical cortex of the brain is physically independent and non-overlapping. Therefore, the system introduces a globally coordinated allocation algorithm to ensure that a one-to-one bijective relationship must be formed between discrete grids. Before inputting the cost matrix into the global allocation algorithm, the difference in the number of reference grid data and review grid data is calculated, and virtual grid data corresponding to the difference in number is added to the side with fewer data, forming a... The matrix forces the joint computational value of all virtual grids involved in the calculation to be a preset global maximum floating-point constant.

[0071] Specifically, the system directly inputs the complete cost matrix into the system's preset Hungarian algorithm solution module. The Hungarian algorithm is a mature technical means to solve the allocation problem. The solution module takes the minimum extreme value of the sum of all selected elements in the matrix as the constraint objective, and performs row reduction and column reduction operations on the cost matrix to determine the homologous brain region mapping data between the baseline grid data and the review grid data.

[0072] As an example, homologous brain region mapping data represents the cross-period homologous brain region mapping tuples output by the global allocation algorithm. In any tuple, there is a set of matched baseline grid data and review grid data, indicating that after excluding displacement interference, the review grid data in this combination covers the same anatomical physical region as the baseline grid data.

[0073] Specifically, input module 103 is used for: The cost matrix is ​​reduced using a global allocation algorithm to obtain a bijective set containing multiple matching results.

[0074] The bijective set is resolved into multiple homologous brain region mapping data between the baseline grid data and the review grid data, wherein any homologous brain region mapping data includes a set of matched baseline grid data and review grid data.

[0075] As an example, after reducing the cost matrix using the Hungarian algorithm, a result is output to the system containing... The system extracts the bijective set of the group matching results and parses it into... A series of homologous brain region mapping data across different periods In this mapping data, each baseline grid data index in the initial search phase Each one was uniquely assigned a corresponding grid index for the review phase. At this point, the system establishes at the algorithm level: after ruling out physical slippage of the equipment, the follow-up review time... Generate review grid data The location of the brain region, compared with the time period of the initial examination. Generate reference grid data The brain regions in question belong to the same physical brain region in objective anatomy.

[0076] Module 104 is used to determine the metabolic difference between the baseline grid data and the review grid data based on the homologous brain region mapping data, and to construct a brain function knowledge graph based on the evolutionary labels corresponding to the metabolic difference.

[0077] As an example, the metabolic difference is the difference in blood sample concentration between the baseline grid data and the review grid data. By determining the evolution results of the metabolic difference, the evolution label of each grid can be determined, and then a brain function knowledge graph can be constructed.

[0078] Module 104 is specifically used for: For any homologous brain region mapping data, extract the baseline mean blood oxygen concentration from the baseline grid data and the re-examination mean blood oxygen concentration from the re-examination grid data.

[0079] As an example, firstly, the mapping data of any homologous brain region is traversed, and the joint generation value corresponding to the mapping relationship is extracted from the cost matrix. The joint generation value is then compared with the set maximum floating-point constant (maximum penalty value) of the system for equality. Specifically: If the system determines that the allocation cost is equal to the maximum penalty value, it indicates that the baseline grid data... During this review, the device had physically moved out of the effective detection field of view of the photoelectric cap (no valid review grid data was available to match). The system immediately executed a circuit breaker operation, directly skipping the reference grid data. The metabolic difference is calculated and the subsequent map update steps are performed. If the system determines that the allocated metabolic value is not equal to the maximum penalty value, it indicates that the mapping relationship is legal and valid.

[0080] The metabolic difference between the baseline grid data and the repeat grid data is determined based on the difference between the mean blood oxygen concentration of the follow-up examination and the mean blood oxygen concentration of the baseline.

[0081] As an example, the "baseline / re-examination mean blood oxygen concentration" represents the absolute oxygen consumption metabolic level, which is affected by pathological evolution and is used to calculate the metabolic difference. The system retrieves the re-examination mean blood oxygen concentration based on the index in the homologous brain region mapping data. and the corresponding baseline mean blood oxygen concentration Metabolic difference The calculation process is shown in the following formula: in, This represents the metabolic difference after removing the influence of equipment position deviation; This represents the average absolute oxygen consumption of homologous brain regions found through algorithm matching within the current review period, which is also the average blood oxygen concentration during the review. This represents the average absolute oxygen consumption of the brain region at the time of initial diagnosis, which is also the baseline average blood oxygen concentration.

[0082] This calculation method ensures that the absolute numerical difference mixed with positional errors is reduced to a purely pathological metabolic variation. Specifically, the subtraction term... Provides the latest metabolic status for the current follow-up period; minuend. As a static baseline, it provides a reference origin; the difference term as a whole reflects the direction and scale of metabolic changes: when the difference is positive and the larger the value, it indicates that the blood oxygen consumption of the local brain region has increased sharply from the initial visit to the follow-up examination; when the difference is negative and the larger the absolute value, it indicates that the blood oxygen perfusion of the local brain region has seriously declined.

[0083] Module 104 is also used for: The standard deviation of physiological fluctuations in the baseline cerebral blood oxygenation data is extracted, and the adaptive interval range is calculated based on the standard deviation of physiological fluctuations.

[0084] As an example, the target subject's brain also experiences endogenous fluctuations in blood oxygen metabolism under normal conditions. Directly applying a uniform constant threshold across all subjects would misinterpret these normal fluctuations. Therefore, the system retrieves the target subject's own baseline standard deviation to construct an individual adaptive boundary.

[0085] As an example, the positive adaptive threshold is first calculated. The calculation method is as follows: in, This represents the preset dynamic boundary multiplier constant (strictly taken as a value in this embodiment). (This represents a 95% confidence interval that covers the vast majority of normal fluctuations in statistical analysis). This represents the standard deviation of physiological fluctuations. This calculation method ensures that the discrimination scale can be flexibly and adaptively adjusted according to the normal activity levels of different target subjects and different brain regions. Specifically, the cardinality term... The multiplier term determines the inherent fluctuation bandwidth of the current brain region. This bandwidth was further broadened to accommodate normal measurement noise. The product of the two terms constructs a discrimination scale: this positive adaptive threshold. The larger the value, the higher the system's tolerance for numerical changes in that brain region, thus effectively avoiding misjudging normal fluctuations in highly active brain regions as abnormal lesions.

[0086] As an example, positive adaptive thresholding Add a minus sign to generate the corresponding negative adaptive threshold. Based on these two adaptive thresholds, the adaptive interval range is obtained. .

[0087] If the metabolic difference is within the adaptive range, then the target object's metabolic difference is determined to be within the normal physiological range, and the brain function knowledge graph is not updated.

[0088] As an example, if the metabolic difference is within the adaptive range, the system determines that the metabolic increase or decrease in that anatomical brain region has not exceeded its inherent normal physiological fluctuation range. The system classifies this difference as normal fluctuation or measurement noise and directly abandons triggering subsequent map evolution updates for that grid.

[0089] If the metabolic difference is higher than the upper limit of the adaptive range, then the evolution label of the metabolic difference in the baseline grid data is determined as the compensatory evolution label.

[0090] As an example, the system determines the metabolic difference. Strictly greater than the positive adaptive threshold (The upper limit of the adaptive range) indicates that the brain region produced excessive oxygen consumption that significantly exceeded the normal baseline level at the time of re-examination. Based on this, the system generates compensatory evolutionary labels for brain region metabolism that characterize abnormal metabolic overload.

[0091] If the metabolic difference is lower than the lower limit of the adaptive range, then the evolution label of the metabolic difference in the baseline grid data is determined as the degenerative evolution label.

[0092] As an example, if the system determines the metabolic difference Strictly less than the negative adaptive threshold (The lower limit of the adaptive range) indicates that the brain region has produced hypoxia perfusion that is significantly lower than the normal baseline level. Based on this, the system generates a degenerative evolution label characterizing the brain region metabolism of atrophic nerve blood supply.

[0093] Each baseline grid data or review grid data is used as an anatomical entity node, and the anatomical entity nodes in the same brain region are connected by directed edges based on the compensatory evolution label or the degenerative evolution label.

[0094] As an example, for baseline grid data that has generated explicit evolutionary labels The system performs structured writing to the atlas database. When constructing its topology, the atlas database relies on anatomical entity names with clear medical significance, and cannot directly use two-dimensional planar coordinates composed of numbers as entities. The system maps the planar coordinates to a standard medical anatomy template by calling a pre-calibrated transformation matrix. Specifically: The system first retrieves a pre-set standardized brain anatomy projection template. Then, the system retrieves the two-dimensional affine transformation matrix corresponding to the template. This two-dimensional affine transformation matrix is ​​generated in advance by spatial projection dimension reduction based on the three-dimensional coordinates of anatomical landmarks (such as the root of the nose, the occipital protuberance, and bilateral preauricular fossae) of the international 10-20 standard EEG system.

[0095] The system uses this two-dimensional affine transformation matrix to transform the coordinates of each of the aforementioned extracted reference center planes. Linear coordinate transformation and registration are performed. After registration, the system overlays and compares the transformed coordinates onto the specific region of the standardized brain anatomical projection template, and reads the name of the anatomical brain region that falls on the template. The system converts the read anatomical brain region name into a string identifier. The system uses this string identifier as the primary key to create the corresponding atlas entity node in the atlas database. After the node initialization is completed, the system extracts the calculated baseline mean blood oxygen concentration. Standard deviation of physiological fluctuations The system writes these two values ​​into the internal attribute fields of the graph entity node, thus obtaining the graph database.

[0096] As an example, the system indexes based on baseline grid data. The system locates the corresponding anatomical entity node in the graph database and uses standard graph theory triplet subject-predicate-object structure construction instructions. The system sets the found graph entity node as the starting subject of the relational structure. Since it represents the evolution of the same brain region over time, the system also sets the same graph entity node (corresponding to the review grid data within the review period) as the ending object of the relational structure. (Where, the subject node is defined as "anatomical region T0", the object node as "anatomical region Tk", and the predicate is the evolution label, thus forming a time-chain topological structure reflecting the disease progression in the graph.) Furthermore, a directed evolutionary relational connection edge is constructed in the graph database, with both the starting and ending points pointing to their respective nodes.

[0097] The metabolic difference is written to each directed connection edge to construct a brain function knowledge graph.

[0098] As an example, after constructing each directed connection edge, the metabolic difference is extracted. The absolute value of the absolute value objectively quantifies the severity of compensatory or degenerative lesions in the brain region. The system uses this absolute value as a weight attribute and writes it into the internal data structure of the evolutionary link. Subsequently, the system assigns the current follow-up examination time period. The timestamp attribute is appended to the connection. After the attribute is written, the system submits a persistent storage instruction to the graph database, completing the long-term follow-up graph update task for the target object in the current review period. Finally, the system completely releases all temporary review data in memory and resets the system state to await the next inflow of optical density data.

[0099] This application provides a brain function abnormality knowledge graph construction system based on near-infrared brain functional imaging. In this system, baseline brain oxygenation data of the target object during the initial examination period and baseline brain oxygenation data during the follow-up examination period are obtained. Based on the pixel distance between the baseline and follow-up grid data, a spatial distance normalization value is determined. A cost matrix is ​​constructed based on the correlation between the corresponding waveform sequences of the baseline and follow-up grid data, and the spatial distance normalization value. This cost matrix is ​​then input into a global allocation algorithm to obtain homologous brain region mapping data between the baseline and follow-up grid data. Furthermore, based on the homologous brain region mapping data, the metabolic difference between the baseline and follow-up grid data is determined. Based on the evolutionary labels corresponding to the metabolic difference, a brain function knowledge graph is constructed. Through joint bijective allocation of pixel distance and waveform similarity, homologous anatomical brain regions are accurately located without rigid body registration, ensuring that the finally extracted metabolic difference truly reflects the pathological evolution trajectory. Furthermore, based on the evolutionary labels corresponding to the metabolic difference, a brain function knowledge graph is constructed, reducing the occurrence of matching failures or many-to-one errors in homologous brain regions.

[0100] Reference Figure 3 , Figure 3 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application.

[0101] like Figure 3 As shown, the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging may include: a processor 1001, a memory 1003, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1003.

[0102] Optionally, the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging may also include a user interface, a network interface, a camera, RF (Radio Frequency) circuitry, sensors, a WiFi module, etc. The user interface may include a display screen and an input submodule such as a keyboard; optional user interfaces may also include standard wired or wireless interfaces. The network interface may include standard wired or wireless interfaces (such as a Wi-Fi interface).

[0103] Those skilled in the art will understand that Figure 3 The structure of the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging shown in the figure does not constitute a limitation on the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0104] like Figure 3 As shown, the memory 1003, serving as a storage medium, may include an operating system, a network communication module, and a brain function abnormality knowledge graph construction program based on near-infrared brain functional imaging. The operating system is a program that manages and controls the hardware and software resources of the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging, supporting the operation of the brain function abnormality knowledge graph construction program based on near-infrared brain functional imaging and other software and / or programs. The network communication module is used to enable communication between the various components within the memory 1003, as well as communication with other hardware and software in the brain function abnormality knowledge graph construction system based on near-infrared brain functional imaging.

[0105] exist Figure 3 In the brain functional abnormality knowledge graph construction device based on near-infrared brain functional imaging shown, the processor 1001 is used to execute the brain functional abnormality knowledge graph construction program based on near-infrared brain functional imaging stored in the memory 1003 to implement the steps of the brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described above.

[0106] The specific implementation of the brain function abnormality knowledge graph construction device based on near-infrared brain functional imaging in this application is basically the same as the various embodiments of the brain function abnormality knowledge graph construction system based on near-infrared brain functional imaging described above, and will not be repeated here.

[0107] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0108] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0110] The above are merely preferred embodiments of this application and do not limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

[0111] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A system for constructing a knowledge graph of brain functional abnormalities based on near-infrared brain functional imaging, characterized in that, The system includes: The acquisition module is used to acquire the baseline brain blood oxygen data of the target object during the initial examination period and the re-examination brain blood oxygen data during the re-examination period. The baseline brain blood oxygen data includes multiple baseline grid data, and the re-examination brain blood oxygen data includes multiple re-examination grid data. The determination module is used to determine a spatial distance normalization value based on the pixel distance between the reference grid data and the review grid data, and to construct a cost matrix based on the correlation between the corresponding waveform sequences of the reference grid data and the review grid data, and the spatial distance normalization value. The input module is used to input the cost matrix into the global allocation algorithm to obtain the homologous brain region mapping data between the baseline grid data and the review grid data; A construction module is used to determine the metabolic difference between the baseline grid data and the review grid data based on the homologous brain region mapping data, and to construct a brain function knowledge graph based on the evolutionary tags corresponding to the metabolic difference.

2. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The system also includes: The segmentation module is used to acquire the first optical density data stream of the target object during the initial inspection period and the second optical density data stream during the re-inspection period, and to divide the image data corresponding to the first optical density data stream and the second optical density data stream into equal amounts of baseline grid data and re-inspection grid data, respectively. The generation module is used to perform waveform analysis processing on the baseline grid data and the review grid data to generate baseline cerebral blood oxygenation data and review cerebral blood oxygenation data.

3. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 2, characterized in that, The generation module includes: The first generation submodule is used to perform waveform analysis processing on the reference grid data to generate reference brain blood oxygenation data; The second generation submodule is used to perform waveform analysis processing on the re-examination grid data to generate re-examination brain blood oxygenation data.

4. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 3, characterized in that, The first generation submodule is specifically used for: Based on the blood oxygen concentration values ​​of the pixels in each of the aforementioned benchmark grid data, a blood oxygen time series is constructed, and the mean baseline blood oxygen concentration and the standard deviation of physiological fluctuations of the blood oxygen time series are calculated. The blood oxygen time series was filtered to obtain a low-frequency fluctuation signal; The low-frequency fluctuation signal is standardized to obtain a standardized waveform data sequence; The standardized waveform data sequence, the center coordinates of the reference grid data, the mean reference blood oxygen concentration, and the standard deviation of physiological fluctuations are integrated to obtain the reference brain blood oxygen data.

5. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The determining module is specifically used for: Extract the first center coordinates of each baseline grid data in the baseline brain blood oxygenation data, and the second center coordinates of each review grid data in the review brain blood oxygenation data; Based on the difference between the first center coordinates and the second center coordinates, the pixel distance between the reference grid data and the review grid data is calculated; When the pixel distance meets the preset conditions, the pixel distance is normalized to obtain a spatial distance normalized value.

6. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 5, characterized in that, The determining module is further configured to: The pixel distance is compared with a preset maximum offset threshold; If the pixel distance is greater than the preset maximum offset threshold, it is determined that the current re-examined grid data and its corresponding matching reference grid data do not belong to the same anatomical brain region, and the calculation is skipped; If the pixel distance is less than or equal to the preset maximum offset threshold, it is determined that the current review grid data and its corresponding matching reference grid data belong to the same anatomical brain region, and the pixel distance meets the preset condition.

7. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The determining module is further configured to: Extract the standardized waveform data sequence of each benchmark grid data in the benchmark brain blood oxygenation data, and the re-examination waveform data sequence of each re-examination grid data in the re-examination brain blood oxygenation data; Calculate the Pearson correlation coefficient between the standardized waveform data sequence and the reviewed waveform data sequence, and convert the Pearson correlation coefficient into the waveform difference cost between the baseline grid data and the reviewed grid data; The joint cost is obtained by weighted summation of the waveform difference cost and the normalized spatial distance value. The cost matrix is ​​constructed based on the joint cost value between the corresponding combinations of each baseline grid and the review grid.

8. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The input module is specifically used for: The cost matrix is ​​reduced using a global allocation algorithm to obtain a bijective set containing multiple matching results. The bijective set is parsed into multiple homologous brain region mapping data between the baseline grid data and the review grid data, wherein any homologous brain region mapping data includes a set of matched baseline grid data and review grid data.

9. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The building module is specifically used for: For any homologous brain region mapping data, extract the mean baseline blood oxygen concentration of the baseline grid data in the homologous brain region mapping data, and the mean re-examination blood oxygen concentration of the re-examination grid data; The metabolic difference between the baseline grid data and the re-examination grid data is determined based on the difference between the mean re-examination blood oxygen concentration and the mean baseline blood oxygen concentration.

10. The brain functional abnormality knowledge graph construction system based on near-infrared brain functional imaging as described in claim 1, characterized in that, The building module is also used for: The physiological fluctuation standard deviation is extracted from the baseline cerebral blood oxygenation data, and the adaptive interval range is calculated based on the physiological fluctuation standard deviation. If the metabolic difference is within the adaptive range, then the metabolic difference of the target object is determined to be within the normal physiological range, and the brain function knowledge graph is not updated. If the metabolic difference is higher than the upper limit of the adaptive interval range, then the evolution label of the metabolic difference in the baseline grid data is determined to be the compensatory evolution label. If the metabolic difference is lower than the lower limit of the adaptive range, then the evolution label of the metabolic difference in the baseline grid data is determined to be a degenerative evolution label. Each baseline grid data or review grid data is used as an anatomical entity node, and the compensatory evolution label or degenerative evolution label is used as a directed connection edge for the anatomical entity nodes of the same brain region. The metabolic difference is written to each of the directed connection edges to construct a brain function knowledge graph.

Images