A preoperative image-based neurosurgical functional brain area positioning method based on hemodynamics

Abstract

The present application relates to the technical field of region segmentation, and particularly relates to a preoperative image brain function area positioning method based on hemodynamics in neurosurgery. The present application determines a blood vessel response lag value by acquiring blood oxygen signals and diffusion tensors of each voxel of the brain, quantifies neurovascular coupling damage into a specific numerical value, and solves the problem that traditional technology cannot quantify blood vessel compliance damage. The present application determines a flow direction structure deviation index by deviation of a gradient vector and an anatomical vector, captures geometric mismatch characteristics of a tumor spatial diffusion direction and a fiber bundle direction, and identifies pathological blood vessel tortuosity. The present application fuses the deviation index and the lag value, determines whole brain perfusion values in combination with diffusion tensors, accurately simulates blood transmission loss in pathological media, and objectively reflects cumulative energy attenuation. The present application determines brain function area confidence by correcting a reference offset of a signal, effectively restores pathological masked neural activity, reveals hidden functional areas that cannot be identified by traditional methods, and significantly improves brain function area positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of region segmentation technology, specifically to a method for locating functional brain regions in preoperative neurosurgical imaging based on hemodynamics. Background Technology

[0002] In neurosurgery, for space-occupying lesions such as gliomas, accurately defining the spatial relationship between the tumor boundary and brain functional areas preoperatively is fundamental to developing a surgical resection plan. Resting-state functional magnetic resonance imaging (fMRI), based on the physiological coupling mechanism between blood oxygen level-dependent signals and neural activity, is often used for preoperative functional area localization. However, neovascularization in brain tumors often lacks smooth muscle structures, and the brain tissue surrounding the tumor is subjected to long-term mechanical compression. These pathological changes disrupt the local neurovascular coupling mechanism. This disruption leads to distortion of the hemodynamic response function in the affected area, specifically manifested as a significant amplitude attenuation and phase lag of the blood oxygen signal relative to actual neural activity.

[0003] Most existing localization methods rely on the assumption of signal synchronicity, failing to recognize asynchronous signals caused by vascular pathology. This results in functional areas surrounding the tumor appearing as disconnected, false-negative regions on imaging. Current technologies struggle to reverse-engineer abnormal signals caused by pathology based on conventional imaging data, leading to low accuracy in the delineation and localization of brain functional areas. Summary of the Invention

[0004] To address the technical problem that existing technologies struggle to reverse-engineer abnormal signals caused by pathology based on conventional imaging data, resulting in low accuracy in the delineation and localization of brain functional areas, the present invention aims to provide a hemodynamic-based method for preoperative neurosurgical imaging-based localization of brain functional areas. The specific technical solution adopted is as follows: This invention first provides a hemodynamic-based method for preoperative imaging localization of brain functional areas in neurosurgery, the method comprising: In the human brain medical database, a whole-brain standardized voxel grid of the target brain is obtained; each voxel of the whole-brain standardized voxel grid includes blood oxygenation time-series signal and diffusion tensor. Based on the temporal phase shift of the blood oxygenation time-series signal of each voxel, the corresponding vascular response hysteresis value is determined; based on the spatial variation trend of the vascular response hysteresis value in the neighborhood of each voxel, the corresponding gradient vector is determined; component analysis is performed on the diffusion tensor of each voxel to determine the corresponding anatomical vector; based on the mutual deviation between the gradient vector and the anatomical vector at each voxel, the corresponding flow structure deviation index is determined. The flow structure deviation index at each voxel is fused with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability value; based on the local permeability value at each voxel and the diffusion tensor at the corresponding voxel, the corresponding whole brain perfusion value is determined; based on the whole brain perfusion value at each voxel and the vascular response hysteresis value at the corresponding voxel, the corresponding blood oxygen correction signal is determined. Based on the relative baseline offset of the blood oxygen correction signal at each voxel, the confidence level of the corresponding brain functional area is determined; based on the magnitude of the confidence level of the brain functional area, the brain functional areas are divided.

[0005] Furthermore, the method for obtaining the vascular response hysteresis value includes: The blood oxygen time-series signal at each voxel is used as input, and the corresponding complex blood oxygen signal is output using Hilbert transform. The complex blood oxygen signal is then standardized to determine the corresponding standard complex blood oxygen signal. The sum of all principal values ​​of the standard complex blood oxygen signal at each voxel is averaged to determine the vascular response hysteresis value at each voxel.

[0006] Furthermore, the method for obtaining the gradient vector includes: The gradient vector is determined by calculating the gradient of all vascular response hysteresis values ​​within the preset neighborhood window of each voxel using the central difference method.

[0007] Furthermore, the method for obtaining the anatomical vector includes: The diffusion tensor at each voxel is inversely transformed to determine the guiding reference tensor; the guiding reference tensor is then standardized to determine the guiding tensor at each voxel. The guiding tensor at each voxel is decomposed into eigenvalues ​​to determine the minimum eigenvector corresponding to the minimum eigenvalue at each voxel; the minimum eigenvector is then used as the anatomical vector at each voxel.

[0008] Furthermore, the method for obtaining the flow direction structure deviation index includes: The absolute value of the dot product between the gradient vector at each voxel and the anatomical vector at the corresponding voxel is taken as the flow structure deviation value; the product between the magnitude of the gradient vector at each voxel and the magnitude of the anatomical vector at the corresponding voxel is taken as the flow structure deviation magnitude value; the ratio between the flow structure deviation value and the flow structure deviation magnitude value is negatively correlated to determine the corresponding flow structure deviation index.

[0009] Furthermore, the method for obtaining the local permeability value includes: The flow direction deviation index at each voxel is positively correlated and mapped to determine the corresponding mapping deviation index; the mapping deviation index at each voxel is weighted and fused with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability reference value; the local permeability reference value is negatively correlated and mapped with the sum of the preset permeability constants to determine the corresponding local permeability value.

[0010] Furthermore, the method for obtaining the whole brain perfusion value includes: A cost accumulation equation is constructed based on the local permeation value and the guidance tensor at each voxel; the calculation result of the cost accumulation equation is the cumulative cost at each voxel; the local permeation value is negatively correlated with the cumulative cost; the guidance tensor is positively correlated with the cumulative cost. The cumulative cost at each voxel is standardized to determine the whole-brain perfusion value at each voxel.

[0011] Furthermore, the method for obtaining the blood oxygen correction signal includes: The whole brain perfusion value at each voxel is positively correlated to determine the corresponding amplitude compensation factor; the vascular response lag value at each voxel is positively correlated to determine the corresponding phase compensation factor; the corresponding blood oxygen compensation factor is determined based on the product between the amplitude compensation factor and the phase compensation factor; and the blood oxygen correction signal at each voxel is determined based on the product between the blood oxygen compensation factor and the blood oxygen time-series signal.

[0012] Furthermore, the method for obtaining the confidence level of the brain functional area includes: The Pearson correlation coefficient between the blood oxygen correction signal and the preset baseline signal at each voxel was used as the confidence level of the brain functional area at each voxel.

[0013] Furthermore, the method for dividing brain functional areas includes: If the confidence level of the brain functional area is greater than the preset confidence threshold, then the corresponding voxel is classified as a brain functional area. If the confidence level of the brain functional area is less than or equal to the preset confidence threshold, the corresponding voxel will be classified as an abnormal area.

[0014] This invention also provides a hemodynamic-based preoperative imaging brain functional area localization system for neurosurgery, the system comprising: The data acquisition module is used to obtain a whole-brain standardized voxel grid of the target brain from the human brain medical database; each voxel of the whole-brain standardized voxel grid includes a blood oxygenation time-series signal and a diffusion tensor. The vascular response and structural analysis module is used to determine the corresponding vascular response hysteresis value based on the temporal phase shift of the blood oxygenation time-series signal of each voxel; determine the corresponding gradient vector based on the spatial variation trend of the vascular response hysteresis value in the neighborhood of each voxel; perform component analysis on the diffusion tensor of each voxel to determine the corresponding anatomical vector; and determine the corresponding flow structure deviation index based on the mutual deviation between the gradient vector and the anatomical vector at each voxel. The original blood oxygen temporal signal correction module is used to fuse the flow structure deviation index at each voxel with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability value; based on the local permeability value at each voxel and the diffusion tensor at the corresponding voxel, the corresponding whole brain perfusion value is analyzed and determined; based on the whole brain perfusion value at each voxel and the vascular response hysteresis value at the corresponding voxel, the corresponding blood oxygen correction signal is analyzed and determined. The brain functional area segmentation module is used to determine the confidence level of the corresponding brain functional area based on the relative baseline offset of the blood oxygen correction signal at each voxel; and to segment the brain functional area based on the magnitude of the confidence level.

[0015] The present invention has the following beneficial effects: This invention determines vascular response hysteresis values ​​by acquiring the temporal signals and diffusion tensors of blood oxygenation at each voxel in the target brain. It transforms abstract neurovascular coupling damage into concrete, calculable vascular response hysteresis values ​​by quantifying the phase delay of each raw blood oxygenation signal relative to the whole-brain average level. This process solves the technical challenge of traditional techniques being unable to quantify the degree of local vascular compliance impairment. Furthermore, by determining the flow structure deviation index through the mutual deviation between gradient vectors and anatomical vectors, it captures for the first time the geometric mismatch between the direction of tumor neoplasia and diffusion and the anatomical direction of normal white matter fiber bundles. This enables precise identification of pathological vascular tortuosity and significantly improves the detection of tumors. The ability to characterize infiltration boundaries; by fusing the flow-direction structure deviation index with the vascular response hysteresis value, and based on this, determining the whole brain perfusion value with diffusion tensor analysis, accurately reflecting the propagation cost of abnormal signal features along anatomical structures, objectively reflecting the degree of cumulative energy attenuation from the healthy blood supply area to the tumor periphery, quantifying the abstract pathophysiological process into a visualized spatial distribution map; by correcting the relative baseline offset of the signal to determine the confidence level of brain functional areas, and finally delineating brain functional areas, effectively recovering the real neural activity signals masked by vascular pathology, thus successfully revealing the hidden functional areas around the tumor that cannot be identified by traditional methods, and significantly improving the accuracy of brain functional area delineation and localization. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of a hemodynamic-based method for locating brain functional areas in preoperative neurosurgery using imaging, provided as an embodiment of the present invention; Figure 2 This is a structural block diagram of a hemodynamic-based neurosurgical preoperative imaging brain functional area localization system provided in one embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a hemodynamic-based preoperative imaging brain functional area localization method for neurosurgery proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of a hemodynamic-based preoperative imaging brain functional area localization method provided by the present invention.

[0021] Please see Figure 1 The diagram illustrates a flowchart of a hemodynamic-based preoperative imaging brain functional area localization method for neurosurgery, according to an embodiment of the present invention. The method includes: Step S101: Obtain the whole-brain standardized voxel grid of the target brain from the human brain medical database; each voxel of the whole-brain standardized voxel grid includes the blood oxygen temporal signal and the diffusion tensor.

[0022] Since blood oxygen time-series signals and diffusion tensors are two completely different types of data, in order to ensure that subsequent calculations and analyses can be built in a unified mathematical space, it is necessary to first unify blood oxygen time-series signals and diffusion tensors on the same benchmark (within voxels). Therefore, a whole-brain standardized voxel grid of the target brain is obtained to provide a physical foundation for subsequent transmission modeling, ensuring the stability and reliability of subsequent analysis and calculations.

[0023] The reason for choosing to acquire the blood oxygen temporal signal and diffusion tensor in this embodiment of the invention is as follows: First, the blood oxygen level of normal brain tissue blood flow is different from that of abnormal brain tissue peripheral blood flow; second, abnormal brain tissue will change the surrounding tissue structure, thereby affecting the direction of blood flow (i.e., the blood flow orientation will change), resulting in a difference between the diffusion tensor of normal brain tissue and the diffusion tensor of abnormal brain tissue; based on this, the present invention acquires the blood oxygen temporal signal and diffusion tensor separately, which is beneficial for the subsequent reverse recovery of the blood oxygen temporal signal.

[0024] As an example, in a specific implementation of this invention, the method for obtaining a whole-brain standardized voxel grid of the target brain, as well as the blood oxygenation time-series signal and diffusion tensor within each voxel, from a human brain medical database is as follows: The system first defines a whole-brain standardized voxel grid, wherein the size of each voxel is set within the range of: length: [1mm, 3mm], width: [1mm, 3mm], and height: [1mm, 3mm], specifically set to a length of 2mm, a width of 2mm, and a height of 2mm; the total number of voxels contained in the whole-brain standardized voxel grid can be set to... arrive In the meantime, taking the commonly used MNI152 standard space (2mm resolution) as an example, the total number of voxels can be set to 902,629. Next, the system acquires resting-state functional magnetic resonance imaging (fMRI) data (including time-series blood oxygenation signals) and diffusion tensor imaging (DTI) data (including diffusion tensors) of the target brain. Since the original acquisition resolution and field of view center may differ between different modalities, rigid body transformation or affine transformation algorithms are used to resample and register all the above data into a whole-brain standardized voxel grid. After alignment, the system reads the time-series blood oxygenation level signal of each voxel and the diffusion tensor of the same voxel within the scan cycle (determined according to the actual scan time, e.g., 2 minutes). The symmetric positive definite matrix is ​​used, and the blood oxygen level time series signal of each voxel is used as the blood oxygen time series signal of each voxel; the blood oxygen time series signal within each voxel and the diffusion tensor within the same voxel are anatomically precisely locked, which improves the stability and accuracy of the subsequent calculation process.

[0025] It should be noted that resting-state functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) are techniques well-known to those skilled in the art, and need not be elaborated upon or limited here. Furthermore, it should be noted that rigid body transformation and affine transformation algorithms are both well-known techniques, and implementers can choose them according to the specific implementation scenario; no further limitations or elaborations are needed here.

[0026] Step S102: Determine the corresponding vascular response hysteresis value based on the temporal phase shift of the blood oxygenation time-series signal at each voxel; determine the corresponding gradient vector based on the spatial variation trend of the vascular response hysteresis value in the neighborhood of each voxel; perform component analysis on the diffusion tensor at each voxel to determine the corresponding anatomical vector; determine the corresponding flow structure deviation index based on the mutual deviation between the gradient vector and the anatomical vector at each voxel.

[0027] On the one hand, in brain tumors and surrounding areas, the lack of smooth muscle structure in tumor angiogenesis and the mechanical compression caused by the tumor's space-occupying effect lead to damage to the local neurovascular coupling mechanism, meaning that blood flow encounters more resistance than normal. This pathological change causes a significant phase delay in the blood oxygenation time-series signal in this region relative to normal neural activity. Therefore, this embodiment of the invention, by analyzing the temporal phase shift of the blood oxygenation time-series signal at each voxel, can determine a vascular response hysteresis value, which reflects the phase hysteresis of the blood oxygenation level within the blood flow at that voxel, providing a reliable basis for subsequent recovery of abnormal signals caused by pathology.

[0028] On the other hand, in the region surrounding a brain tumor, pathological changes in blood vessels are manifested not only as increased resistance but also as disordered vascular orientation. This disorder leads to the asynchronous characteristics of local signals spreading in a spatially dominant gradient direction that is orthogonal to or deviates from the originally ordered direction of white matter fiber bundles. Therefore, in order to quantify this key geometric mismatch feature that causes signals in functional areas to be masked, this embodiment of the invention analyzes the mutual deviation between the direction of resistance to blood flow at each voxel (i.e., gradient vector) and the dominant direction of white matter fibers in the background tissue at each voxel (i.e., anatomical vector). If the deviation is greater, i.e., the two directions are less coincident, it indicates that the voxel may be in the region surrounding a brain tumor. Conversely, if the deviation is smaller, i.e., the two directions are more coincident, it indicates that blood flow is more constrained by white matter fibers, and the voxel may be a brain functional area. Thus, a flow structure deviation index can be determined, which can represent the pathological tortuosity specific to tumors.

[0029] Since the vascular response hysteresis value is determined based on the blood oxygenation time-series signal at each voxel, the spatial variation trend of the vascular response hysteresis value can accurately characterize the direction of resistance to blood flow at that voxel, i.e., determine a gradient vector that represents the main gradient direction of the spatial diffusion of the asynchronous signal characteristics, i.e., the potential pathological obstruction direction. Since the diffusion tensor is the information of background white matter fiber tissue obtained through diffusion tensor imaging (DTI) technology, it itself contains the guidance information of white matter fibers at that voxel. Therefore, by performing component analysis on the diffusion tensor, an anatomical vector can be determined, which can represent the direction of white matter fiber bundles at each voxel. Based on the mutual deviation between the gradient vector and the anatomical vector, the accuracy and reliability of subsequent flow direction structure deviation index calculation can be improved.

[0030] Step S103: Fuse the flow structure deviation index at each voxel with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability value; based on the local permeability value at each voxel and the diffusion tensor at the corresponding voxel, analyze and determine the corresponding whole brain perfusion value; based on the whole brain perfusion value at each voxel and the vascular response hysteresis value at the corresponding voxel, analyze and determine the corresponding blood oxygen correction signal.

[0031] As explained in step S102, blood flow through brain tissue is simultaneously inhibited by both static resistance and geometric deviation. Therefore, in order to accurately quantify the ease or difficulty of blood flow through a unit voxel, it is necessary to fuse the flow direction deviation index and the vascular response hysteresis value at each voxel. The vascular response hysteresis value represents the static resistance at each voxel, and the flow direction deviation index represents the geometric deviation at each voxel. This can determine a local permeability value that reflects the comprehensive obstruction effect of tortuous blood vessels on blood flow through each voxel.

[0032] Since brain tissue is a continuous medium, blood is transported throughout the entire brain. Furthermore, blood transport within the microvascular network follows anisotropic osmotic laws. Therefore, the actual flow of blood is constrained by the combined obstruction of tortuous vessels and the orientation information of white matter fibers. Specifically, by analyzing the local osmotic value and diffusion tensor at each voxel, the overall resistance of blood flow through each voxel can be determined. The overall resistance of blood flow is greater in voxels in the region surrounding brain tumors, while it is smaller in voxels in normal brain functional areas. This allows for the determination of a whole-brain perfusion value, reflecting not only the obstructive effect of all pathological regions (such as the area surrounding the tumor) on blood flow, but also the constraint of the anatomical path determined by the orientation of white matter fiber bundles. This provides a reliable data basis for the inverse correction of blood oxygenation time-series signals.

[0033] The original blood oxygenation time series signal is an abnormal signal caused by pathological changes (specifically, the blood oxygenation time series signal shows significant amplitude attenuation and phase lag relative to the actual neural activity). When analyzing and classifying brain tissue regions, it will reduce the accuracy of the classification results. Therefore, it is necessary to combine it with whole brain perfusion values ​​for reverse recovery, which can effectively achieve synchronous correction of the original signal and is also conducive to improving the accuracy of subsequent brain functional area classification and localization.

[0034] Step S104: Determine the confidence level of the corresponding brain functional area based on the relative baseline offset of the blood oxygen correction signal at each voxel; divide the brain functional area based on the numerical value of the confidence level of the brain functional area.

[0035] After obtaining the blood oxygen correction signal, it is necessary to determine whether the blood oxygen correction signal reflects a normal brain tissue region. That is, the blood oxygen correction signal needs to be compared with a normal baseline blood oxygen signal. Based on the magnitude of the offset between the two, the confidence level of a normal brain functional area is determined, which provides an accurate basis for subsequent division and localization of brain functional areas.

[0036] Based on the confidence scores of the brain functional areas, it is possible to adaptively determine whether each voxel belongs to an abnormal region or a truly normal brain functional area, effectively identifying functional areas around tumors that appear as unconnected false negative regions on images, and significantly improving the accuracy of brain functional area division and localization.

[0037] In summary, this invention, by acquiring the temporal blood oxygenation signal and diffusion tensor at each voxel in the target brain, determines the vascular response hysteresis value. It can transform abstract neurovascular coupling damage into concrete, calculable vascular response hysteresis values ​​by quantifying the phase delay of each raw blood oxygenation signal relative to the average level of the whole brain. This process solves the technical problem of traditional techniques being unable to quantify the degree of local vascular compliance impairment. Furthermore, by determining the flow structure deviation index through the mutual deviation between gradient vectors and anatomical vectors, it captures for the first time the geometric mismatch between the direction of tumor neoplasia and diffusion and the anatomical direction of normal white matter fiber bundles, enabling precise identification of pathological vascular tortuosity and significantly improving the ability to accurately identify pathological vascular tortuosity. The ability to characterize tumor infiltration boundaries; by fusing the flow-direction structure deviation index with the vascular response hysteresis value, and based on this, determining the whole brain perfusion value with diffusion tensor analysis, accurately reflecting the propagation cost of abnormal signal features along anatomical structures, objectively reflecting the degree of cumulative energy attenuation from the healthy blood supply area to the tumor periphery, quantifying the abstract pathophysiological process into a visualized spatial distribution map; by correcting the relative baseline offset of the signal to determine the confidence of brain functional areas, and finally delineating brain functional areas, effectively recovering the real neural activity signals masked by vascular pathology, thus successfully revealing the hidden functional areas around the tumor that cannot be identified by traditional methods, and significantly improving the accuracy of brain functional area delineation and localization.

[0038] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the vascular response hysteresis value includes: The blood oxygen time-series signal at each voxel is used as input, and the corresponding complex blood oxygen signal is output using Hilbert transform. The complex blood oxygen signal is then standardized to determine the corresponding standard complex blood oxygen signal. The sum of all principal values ​​of the standard complex blood oxygen signal at each voxel is averaged to determine the vascular response hysteresis value at each voxel.

[0039] Since the acquired blood oxygenation time-series signal is in real number form, and the phase delay of the blood oxygenation time-series signal caused by pathology is reflected in the phase, in order to accurately separate and extract the required key information, namely the instantaneous phase of the signal, this embodiment of the invention considers using Hilbert transform to convert the blood oxygenation time-series signal into a complex blood oxygenation signal with complex terms, separating the amplitude and phase information of the original blood oxygenation time-series signal, and accurately and stably extracting the instantaneous phase information of each voxel. This allows for the rapid separation of the blood oxygenation phase difference of each voxel through complex number operations.

[0040] The blood oxygenation time-series signals of different target subjects' brains are different. If absolute values ​​are used, the calculation results will not have universality and practicality. Therefore, it is necessary to standardize the blood oxygenation complex signal and determine a standard blood oxygenation complex signal, which effectively improves the reliability of analysis and calculation.

[0041] In one specific implementation of this invention, the method for obtaining the standard complex signal of blood oxygen is as follows: First, the system uses Hilbert transform to convert the blood oxygen time-series signal into the corresponding complex blood oxygen signal. Specifically: ,in, This represents the complex oxygenation signal of the x-th voxel at the t-th scan time. This represents the blood oxygenation time-series signal of the x-th voxel at the t-th scan time. Indicates the Hilbert transform operation; Represents the imaginary unit; then calculates the overall complex signal of blood oxygenation in the brain: ,in, This represents the complex oxygenation signal in the whole brain at time t during the scan. This represents the temporal signal of blood oxygenation in the whole brain at time t during the scan; finally, the standard complex signal of blood oxygenation is calculated: , This represents the standard complex signal of blood oxygen at the t-th scan time for the x-th voxel; To represent a very small constant, for example This is intended to prevent the denominator from being zero; it should be noted that the Hilbert transform is a well-known technique and will not be elaborated upon here.

[0042] Blood oxygenation signals in the brain are not static, but rather rhythmic physiological oscillations. In healthy brain regions, these oscillations are generally consistent. When blood vessels undergo pathological changes (lack of smooth muscle, tortuosity, compression), leading to impaired neurovascular coupling mechanisms, the most direct physical manifestation of this damage is a sluggish local hemodynamic response, i.e., a phase delay. In this embodiment of the invention, the principal value of the amplitude is the mathematical quantification of this phase delay, reflecting the magnitude of the local hemodynamic lag caused by neurovascular coupling damage, unique to each voxel after eliminating physiological fluctuations common to the entire brain. It should be noted that the blood oxygenation time-series signal is a temporally continuous signal, meaning that each scan time corresponds to an amplitude value. Therefore, the standard complex blood oxygenation signal also contains multiple principal values ​​of amplitude corresponding to multiple scan times. Averaging the sum of all principal values ​​of amplitude in the standard complex blood oxygenation signal at each voxel transforms the dynamically changing phase delay information into a stable, statistically significant static characteristic value, i.e., the vascular response lag value.

[0043] In one specific implementation of this invention, the method for obtaining the vascular response hysteresis value is as follows: The system first calculates the principal value of the amplitude of each blood oxygen standard complex signal, specifically: ,in, This represents the principal value of the amplitude of the standard complex signal of blood oxygen at time t for the x-th voxel, with a value range of... ; This represents the principal argument function taking complex numbers. Calculating the argument eliminates the influence of amplitude fluctuations, retaining only the phase difference. Then, the absolute deviation of the principal argument value of each voxel relative to the mean of the principal argument values ​​of all voxels in the whole brain is calculated, and a physiologically normal deviation threshold (e.g., 0.1) is set. (representing approximately 0.314 radians), the vascular response hysteresis values ​​of voxels with an absolute deviation less than this threshold are directly set to 0. Only voxels with deviations greater than this threshold are normalized to their maximum and minimum values, and the principal argument values ​​of the corresponding voxels are normalized to the range [0, 1] to improve the stability of subsequent calculations. Finally, the vascular response hysteresis values ​​are calculated: ,in, The vascular response hysteresis value of the x-th voxel ranges from [0, 1]. T represents the number of scan moments within a scan cycle (e.g., 2 minutes), and the averaging operation aims to obtain stable pathological features. A larger vascular response hysteresis value indicates more severe impairment of vascular compliance and a slower neurovascular response at that voxel. For healthy brain tissue, this vascular response hysteresis value typically approaches zero. The maximum and minimum values ​​used in the maximum-minimum normalization can be obtained based on the actual situation. For example, when multiple values ​​can be obtained during implementation and it is necessary to compare the magnitudes of different values, multiple values ​​can be statistically analyzed to obtain the maximum and minimum values. Conversely, when only a single value can be obtained during implementation or it is not necessary to compare the magnitudes of different values, the maximum and minimum values ​​can be statistically analyzed based on a large amount of historical experimental data or prior data.

[0044] It should be noted that the maximum and minimum value normalization algorithm can also be replaced by the maximum value normalization algorithm or the decimal scaling normalization algorithm. These are all well-known technologies, and implementers can adjust them according to the specific implementation scenario. They will not be elaborated or limited here.

[0045] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the gradient vector includes: The gradient vector is determined by calculating the gradient of all vascular response hysteresis values ​​within the preset neighborhood window of each voxel using the central difference method.

[0046] The vascular response hysteresis value represents the magnitude of the phase lag of blood at each voxel. A larger vascular response hysteresis value indicates a larger phase lag, meaning greater resistance at that voxel, and vice versa. Therefore, the spatial gradient of the vascular response hysteresis value within a preset neighborhood can reflect the direction in which local signal asynchrony features spread the fastest (phase lag increases the most), i.e., the direction of potential pathological obstruction. This allows for accurate quantification of pathological features, providing data support for subsequent feature calculations. It should be noted that the gradient at a voxel location in space depends on the data distribution within its surrounding neighborhood. In this embodiment of the invention, the preset neighborhood window is set to be centered on each voxel. Within the spatial range, the aim is to ensure that the calculated spatial gradient can reflect the main gradient direction of the spatial diffusion of local signal asynchronous characteristics to the greatest extent. It should be further noted that for voxels at the boundary positions of the brain normalized voxel grid, there may be empty spaces in their preset neighborhood windows. In this case, the vascular response hysteresis value at the voxel position is used to fill in the gaps to ensure the stability of the calculation.

[0047] In one specific implementation of this invention, the gradient vector at each voxel is calculated using the central difference method. The central difference method is a well-known technique and will not be described in detail here.

[0048] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the anatomical vector includes: The diffusion tensor at each voxel is inversely transformed to determine the guiding reference tensor; the guiding reference tensor is then standardized to determine the guiding tensor at each voxel. The guiding tensor at each voxel is decomposed into eigenvalues ​​to determine the minimum eigenvector corresponding to the minimum eigenvalue at each voxel; the minimum eigenvector is then used as the anatomical vector at each voxel.

[0049] Since the diffusion tensor characterizes conduction capacity, while blood flow focuses on transmission resistance or path cost when searching for loss paths, and since peritumoral edema and extracellular matrix changes can simultaneously cause variations in DTI diffusion characteristics and distortion of blood oxygenation signals, using the inverse DTI matrix as a spatial conduction constraint can more accurately confine abnormal neurovascular decoupling signals to the actual anatomical lesion area. Therefore, this embodiment of the invention performs an inverse transformation on the diffusion tensor. Furthermore, to facilitate subsequent data processing, the guiding reference tensor is standardized after the inverse transformation. Specifically, the inverse matrix of the diffusion tensor for each voxel is calculated, and its trace is used for normalization, which can be expressed by the formula: ,in, Let x be the guiding tensor of the x-th voxel; Let x represent the diffusion tensor of the x-th voxel; The `tr()` function represents the operation of finding the inverse of a matrix; `tr()` represents the trace function of a matrix. Represents a pre-defined minimum constant, for example The purpose is to prevent the denominator from being zero; the guide tensor defines the anisotropic constraint property of the local space: the transmission cost defined by the guide tensor is minimized in the direction compliant with the main direction of the nerve fiber bundle, while the transmission cost is maximized in the direction perpendicular to the fiber bundle.

[0050] The guided tensor is a The matrix contains the directional information of all white matter fibers at the voxel location. To facilitate subsequent analysis and calculation, this embodiment of the invention performs component analysis on the guidance tensor to extract the directional information that best represents the anatomical direction of the white matter fibers at the voxel location. Therefore, eigenvalue decomposition is performed on the guidance tensor at each voxel location to obtain multiple eigenvalues ​​and corresponding eigenvectors. Since the guidance tensor is obtained by inverting the diffusion tensor, it can also reflect the direction of resistance at the voxel location. Therefore, in this embodiment of the invention, the minimum eigenvector corresponding to the minimum eigenvalue represents the direction of minimum conduction resistance, i.e., the direction of the white matter fiber bundle (because blood flows with the least resistance along the fiber bundle). Using the minimum eigenvector as the anatomical vector at each voxel location, the local orientation information of the white matter fiber bundle at each voxel location can be accurately extracted, providing a clear directional reference for subsequent quantification of the geometric mismatch between the blood flow direction and the anatomical structure direction. It should be noted that eigenvalue decomposition uses existing calculation methods, which will not be elaborated here.

[0051] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the flow direction structure deviation index includes: The absolute value of the dot product between the gradient vector at each voxel and the anatomical vector at the corresponding voxel is taken as the flow structure deviation value; the product between the magnitude of the gradient vector at each voxel and the magnitude of the anatomical vector at the corresponding voxel is taken as the flow structure deviation magnitude value; the ratio between the flow structure deviation value and the flow structure deviation magnitude value is negatively correlated to determine the corresponding flow structure deviation index.

[0052] In normal brain tissue, the growth of microvessels follows the anatomical direction of white matter fiber bundles. That is, under normal physiological conditions, changes in blood flow resistance are distributed along the nerve conduction path, and the direction of the resistance gradient (gradient vector) and the fiber direction (anatomical vector) tend to be parallel. However, the morphology of neovascularization in brain tumors is chaotic, disregarding the original anatomical channels. The tumor's space-occupying and compression cause vascular torsion, resulting in increased blood flow resistance perpendicular to the direction of white matter fiber bundles. The direction of the resistance gradient and the fiber direction no longer tend to be parallel. Therefore, in this embodiment of the invention, the absolute value of the dot product between the gradient vector at each voxel and the corresponding anatomical vector at the voxel is calculated to determine a flow structure deviation value, which can characterize the degree of pathological mismatch caused by vascular malformation and mechanical compression. The reason for calculating the absolute value of the dot product instead of the dot product is that the cases where the direction of the resistance gradient (gradient vector) is parallel to the fiber direction (anatomical vector) in the opposite direction and the cases where the direction of the resistance gradient (gradient vector) is parallel to the fiber direction (anatomical vector) in the forward direction are consistent. Both can indicate that the growth of microvessels follows the anatomical direction of white matter fiber bundles, which is a normal physiological state. Therefore, taking the absolute value of the dot product can conform to the reality and ensure the rationality of the calculation.

[0053] Because the vascular response hysteresis values ​​are uniformly distributed in normal brain tissue and the gradient vectors are close to 0, directly substituting them into the subsequent formula for calculating the flow-to-structure deviation index could lead to normal voxels being misjudged as pathological states. Therefore, before calculating the structure deviation index, it is necessary to determine whether the magnitude of the gradient vector at the voxel is less than a preset minimum threshold, such as one-thousandth of the magnitude of the gradient vectors of all voxels in the whole brain, which can accommodate the numerical range of different target objects. If it is less than 0, the flow-to-structure deviation index of that voxel is directly set to 0; otherwise, the subsequent calculation of the flow-to-structure deviation index is performed normally.

[0054] Since the gradient vector and the anatomical vector have different magnitudes, directly using the absolute value of their dot product as the calculation result would introduce irrelevant magnitude information, making the final flow structure deviation index unable to fully reflect the directional angle information between the gradient vector and the anatomical vector. Therefore, in this embodiment of the invention, the product between the magnitude of the gradient vector at each voxel and the magnitude of the anatomical vector at the corresponding voxel is used as the flow structure deviation magnitude. By using the ratio operation, the interference of the absolute values ​​of the two vectors can be eliminated, and only the pure directional angle information is extracted, thus improving the accuracy of the calculation result.

[0055] According to the calculation principle of vector dot product, the more parallel two vectors are, the larger the absolute value of the dot product; the more perpendicular two vectors are, the smaller the absolute value of the dot product (when two vectors are perpendicular, the dot product is 0). Therefore, the larger the deviation value of the flow structure, the smaller the deviation between the blood flow direction and the direction of white matter fibers (the more parallel the blood flow direction and the direction of white matter fibers). That is, the deviation value of the flow structure is negatively correlated with the degree of deviation between the blood flow direction and the direction of white matter fibers. Therefore, the embodiment of the present invention uses negative correlation mapping to make the final flow structure deviation index positively reflect the degree of deviation between the blood flow direction and the direction of white matter fibers.

[0056] In one specific implementation of this invention, the flow direction structure deviation index can be expressed by the following formula: ,in, The flow structure deviation index of the x-th voxel is represented, and its value range is [0, 1]. Let x represent the gradient vector of the x-th voxel; Represents the anatomical vector of the x-th voxel; This represents the absolute value function, designed to eliminate the effect of negative values; This represents the modulus function of a vector; To represent a very small positive number, aiming to avoid a denominator of 0, it can be set as... When the gradient vector is parallel to the anatomical vector, the flow structure deviation index approaches 0, representing normal physiological resistance changes; when the gradient vector is perpendicular to the anatomical vector, the flow structure deviation index approaches 1, representing the pathological tortuosity specific to tumors.

[0057] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the local permeability value includes: The flow direction deviation index at each voxel is positively correlated and mapped to determine the corresponding mapping deviation index; the mapping deviation index at each voxel is weighted and fused with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability reference value; the local permeability reference value is negatively correlated and mapped with the sum of the preset permeability constants to determine the corresponding local permeability value.

[0058] The flow structure deviation index is relatively small in normal brain functional areas, while in the voxels of tumor boundary regions, although the flow structure deviation index is larger than that in normal brain functional areas, the difference between the two is small, making it very easy to misjudge. Therefore, firstly, a positive correlation mapping is applied to the flow structure deviation index, which makes the model more sensitive to significant pathological areas and less sensitive to normal or slightly abnormal areas, thereby forming a sharper contrast at the pathological boundary and effectively improving the ability to characterize the tumor invasion boundary.

[0059] The ease with which blood flows through a unit voxel is inhibited by both static resistance and geometric deviation, which are reflected in the vascular response hysteresis value (the larger the value, the greater the static resistance) and the mapping deviation index (the larger the value, the greater the geometric deviation), respectively. Therefore, by weighting and fusing the mapping deviation index and the vascular response hysteresis value, a local permeability reference value is determined, which can comprehensively quantify the dual pathological inhibition effect on blood flow at each voxel. The larger the local permeability reference value, the more difficult it is for blood to permeate at that voxel. Therefore, negatively mapping the local permeability reference value makes it easier for blood to permeate and flow at that voxel, which is expressed as a positive correlation.

[0060] In one specific implementation of this invention, the method for calculating local permeability is as follows: ,in, This represents the local osmotic value of the x-th voxel; This term represents a positive correlation mapping to the flow direction structure deviation index, determining the corresponding mapping deviation index; where... This represents a nonlinear factor designed to adjust the model's sensitivity to different pathological regions: when When the value is a positive integer 1, it indicates that the effect of the mapping deviation exponent is linear, because the range of the flow structure deviation exponent is [0, 1]; when When the value is greater than a positive integer 1, the model further amplifies larger flow structure deviation indices, while having almost no effect on smaller flow structure deviation indices; that is, it is more sensitive to significantly pathological areas and less sensitive to normal or slightly abnormal areas. When the value is less than a positive integer 1, it indicates that the model is not sensitive to significant pathological areas, but instead becomes more sensitive to normal or slightly abnormal areas. In the embodiments of this invention, this is typically... The value is set to be greater than a positive integer 1, for example, in the implementation of this invention. A value of 2 is intended to create a sharper contrast at the pathological boundary and improve the ability to depict the boundary of tumor invasion. The adjustment weight representing the vascular response lag value aims to control the contribution of vascular dysfunction (response lag) to blood flow obstruction. The larger the value, the greater the contribution of vascular dysfunction to blood flow obstruction. The value can be a positive integer 1. The moderating weight of the mapping deviation index is used to control the contribution of vascular structural disorder (geometric tortuosity) to blood flow obstruction. A higher value indicates a greater contribution of vascular structural disorder to blood flow obstruction. In tumor pathological regions, geometric mismatch is more specific, therefore it can be used to... The value is set to be greater than For example, the value is 2; the positive integer 1 in the denominator is intended to ensure that the local permeability value can return to a reasonable benchmark value, i.e., a positive integer 1, when both the vascular response hysteresis value and the mapping deviation index are 0, thus avoiding the mathematical undefinedness caused by the denominator being 0, and at the same time providing a normalized reference benchmark for the blood flow velocity of healthy tissues.

[0061] It should be noted that the above-mentioned nonlinear factors and two adjustment weights and All of these can be adjusted by the implementer according to the specific implementation scenario, and will not be elaborated or limited here.

[0062] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the whole brain perfusion value includes: A cost accumulation equation is constructed based on the local permeation value and the guidance tensor at each voxel; the calculation result of the cost accumulation equation is the cumulative cost at each voxel; the local permeation value is negatively correlated with the cumulative cost; the guidance tensor is positively correlated with the cumulative cost. The cumulative cost at each voxel is standardized to determine the whole-brain perfusion value at each voxel.

[0063] Brain tissue is a continuous medium, and blood transport in the microvascular network follows anisotropic permeability. Therefore, simple Euclidean distance cannot reflect the complex pathological transport process. To simulate the dynamic process of blood permeating from healthy brain regions to all parts of the brain, an anisotropic functional equation (i.e., cost accumulation equation) is constructed to solve for the cumulative cost required for blood to reach all voxels in the brain. Since the local permeability value obtained above represents the ease of blood permeation and flow at each voxel, the larger the value, the easier the blood flows and the smaller the cost required. Therefore, the local permeability value is negatively correlated with the cumulative cost to be solved. The guidance tensor is obtained by inverse transformation of the diffusion tensor. The guidance tensor characterizes the transmission resistance or path cost that blood flow focuses on when searching for loss paths. Therefore, the guidance tensor is positively correlated with the cumulative cost.

[0064] To provide a reference benchmark for calculating the cumulative cost of the entire brain, a reference zero potential surface must first be determined, i.e., a healthy blood supply region in the brain where the neurovascular coupling mechanism remains intact. Without an accurate definition of the source point (healthy blood supply region), the subsequent cumulative cost calculation will lack a benchmark. Therefore, in a specific implementation of this invention, the source point (healthy blood supply region) is obtained as follows: A histogram distribution of the vascular response hysteresis values ​​of all voxels in the entire brain is statistically analyzed. Since the vascular response of healthy brain tissue is sensitive, its hysteresis value should be close to zero. A statistical screening criterion is set, for example, selecting voxels in the top 10% of the low-value range of the histogram as candidate voxels, which can be adjusted according to the specific implementation scenario. To avoid interference from isolated low-value points caused by image noise, three-dimensional spatial connected component analysis is performed on the above candidate voxels. The number of voxels contained in each connected component is calculated, and connected components with a voxel count greater than a preset voxel count threshold (e.g., one-thousandth of all voxels in the entire brain) are retained and defined as the source point set S. For all voxel coordinates in the source point set, their cumulative cost is initialized to zero.

[0065] In one specific implementation of this invention, the anisotropic equation (i.e., the cost accumulation equation) can be represented by the following equation: ,in, This represents the local osmotic value of the x-th voxel; Let x be the guiding tensor of the x-th voxel; This represents the cumulative cost of blood flow from the source set to the xth voxel; Let represent the gradient of the cumulative cost of the x-th voxel, which is a vector representing the direction of the fastest cost growth; Indicates the transpose operation; The dot product symbol is used; it should be noted that the above equations are solved using the anisotropic fast marching algorithm on a whole-brain normalized voxel grid, where the algorithm starts from the boundary conditions. Initially, updates are pushed forward to the whole brain according to the direction guided by the guiding tensor until the cumulative cost of all voxels has been calculated. It should be noted that the anisotropic fast travel algorithm is a well-known technology and will not be elaborated here.

[0066] Since the cumulative cost obtained from numerical solutions may include extreme outliers caused by necrotic regions, and its dimensions are not suitable for direct use in subsequent signal multiplication operations, it needs to be standardized to a dimensionless attenuation probability to ensure the stability of the reverse recovery of the original blood oxygenation time-series signal. In a specific implementation of this invention, the process of standardizing the cumulative cost at each voxel to determine the whole-brain perfusion value at each voxel is as follows: First, the cumulative costs of all voxels in the whole brain are sorted from smallest to largest. The system selects the 95th percentile, denoted as . To ensure the robustness of the model, all cumulative costs greater than the 95th percentile are truncated to the 95th percentile, which eliminates the influence of singular values ​​in necrosis centers. It should be noted that the 95th percentile can be replaced with the 98th percentile, etc., depending on the specific implementation scenario; this will not be elaborated upon or limited here. Then, all truncated cumulative costs are normalized using a maximum-minimum normalization algorithm, normalizing their range to [0, 1] to ensure computational stability. The normalized result is the whole-brain perfusion value at each voxel. A whole-brain perfusion value of 0 indicates that the voxel is in a healthy blood supply trunk. A brain perfusion value of 1 indicates that the voxel is located at the end of the blood supply or in a severely obstructed area, serving as the baseline data for the amplitude recovery of subsequent blood oxygenation time-series signals. The maximum and minimum values ​​used in the maximum and minimum value normalization algorithm can be obtained according to the actual situation. For example, when multiple values ​​can be obtained during the implementation process and it is necessary to compare the magnitude relationship between different values, multiple values ​​can be counted to obtain the maximum and minimum values. However, when only a single value can be obtained during the implementation process or it is not necessary to compare the magnitude relationship between different values, the maximum and minimum values ​​can be obtained by counting based on a large amount of historical experimental data or prior data obtained in the early stage.

[0067] In other specific implementations of this invention, the maximum and minimum value normalization algorithm can also be replaced by decimal scaling normalization, etc. The implementer can set it according to the specific implementation scenario.

[0068] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the blood oxygen correction signal includes: The whole brain perfusion value at each voxel is positively correlated to determine the corresponding amplitude compensation factor; the vascular response lag value at each voxel is positively correlated to determine the corresponding phase compensation factor; the corresponding blood oxygen compensation factor is determined based on the product between the amplitude compensation factor and the phase compensation factor; and the blood oxygen correction signal at each voxel is determined based on the product between the blood oxygen compensation factor and the blood oxygen time-series signal.

[0069] In the context of brain tumors, the failure of functional area localization is often not due to the disappearance of neural activity, but rather because vascular pathology causes amplitude attenuation and phase lag in the blood oxygen signal during transmission. Therefore, the inverse recovery of the original blood oxygen timing signal involves both amplitude and phase aspects. Whole-brain perfusion values ​​reflect the normalized resistance of blood oxygen flowing through each voxel; a larger value indicates greater resistance at that voxel, which inhibits the amplitude of the original blood oxygen timing signal. Therefore, the brain perfusion values ​​need to be positively correlated (positively compensated) to determine the corresponding amplitude compensation factor. The vascular response lag value represents the magnitude of the phase lag in the original signal, i.e., the original signal has shifted left on the time axis. Therefore, to bring the original signal back to the normal baseline on the time axis, the vascular response lag value needs to be positively correlated (positively compensated) to determine a phase compensation factor that can advance the shift of the originally lagging waveform.

[0070] In one specific implementation of this invention, the amplitude compensation factor and the phase compensation factor can be expressed by the following formulas: Amplitude compensation factor: ,in, The amplitude compensation factor represents the x-th voxel; It represents an exponential function with the natural constant as its base, designed to represent the physical inverse process of fluid pressure (signal amplitude) decaying exponentially with the accumulation of resistance in a medium; This represents the whole-brain perfusion value of the xth voxel; This represents the compensation intensity coefficient, and its value is set to the reciprocal of the average signal-to-noise ratio of the original signal in the whole brain (for example, if the average signal-to-noise ratio of the whole brain is 50, then...). It can be set to 0.02), or it can be a constant based on experience (e.g., an empirical range). (The value can be 2). Phase compensation factor: ,in, The phase compensation factor represents the x-th voxel; This represents an exponential function with the natural constant as its base, designed to convert addition operations in the time domain into multiplication operations in the frequency domain, ensuring the stability of subsequent reverse recovery operations of the original blood oxygen time series signal in the frequency domain; Represents the imaginary unit; This represents the vascular response hysteresis value of the xth voxel; In summary, by combining the compensation factors of amplitude and phase, and multiplying them, a blood oxygen compensation factor can be determined. This factor can inversely recover the original blood oxygen time series signal, obtaining the final blood oxygen correction signal, thus providing reliable data support for brain functional area division. It should be noted that since phase lag manifests as a translation of the overall waveform in the time domain and a rotation of the phase angle of each frequency component in the frequency domain, operating in the frequency domain can achieve more accurate synchronization correction. Therefore, this embodiment of the invention applies the blood oxygen compensation factor to the frequency domain representation of the original blood oxygen time series signal. In a specific implementation of this embodiment, the method for obtaining the blood oxygen correction signal is as follows: First, for each voxel in the whole-brain standardized voxel grid, the following operations are performed: extract the original blood oxygen time series signal of that voxel within the entire scan cycle, and perform a Discrete Fourier Transform (DFT) on it to obtain the frequency domain signal, denoted as . , where x represents the xth voxel; then, the blood oxygen compensation factor is multiplied by the frequency domain signal, specifically: , This represents the blood oxygen-corrected frequency domain signal of the xth voxel; The oxygen compensation factor for the xth voxel; The dot product symbol represents the dot product operation. Physically, this operation accomplishes two things simultaneously: it raises the amplitude of the weak signal due to insufficient blood supply to a normal observation level, and it shifts the waveform lagging due to slow vascular response back to the synchronous moment of neuronal firing on the time axis. Finally, by performing an inverse discrete Fourier transform (IDFT) on the blood oxygen correction frequency domain signal and taking its real part, the final blood oxygen correction signal can be obtained.

[0071] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the confidence level of the brain functional area includes: The Pearson correlation coefficient between the blood oxygen correction signal and the preset baseline signal at each voxel was used as the confidence level of the brain functional area at each voxel.

[0072] The final blood oxygenation correction signal has removed spatiotemporal distortions caused by vascular pathology, returning to a baseline reflecting real neural activity. Therefore, the blood oxygenation correction signal can be used for subsequent segmentation and localization of brain functional areas. To improve the universality of subsequent brain functional area segmentation and localization, this embodiment of the invention uses the similarity (Pearson correlation coefficient) between the blood oxygenation correction signal and the preset baseline signal for judgment. This operation quantifies the probability that each voxel belongs to a normal brain functional area, thereby preserving the fuzzy transition information of functional area boundaries, allowing the final brain functional area localization to be flexibly set according to the specific implementation scenario. It should be noted that the Pearson correlation coefficient is a formula technique and will not be elaborated here; however, the Pearson correlation coefficient can also be replaced with other similarity indicators, such as the Spearman correlation coefficient or covariance. Implementers can adjust it according to the specific implementation scenario, and there is no need to limit it here.

[0073] In one specific implementation of this invention, the method for obtaining the preset reference signal is as follows: The system introduces a standard anatomical atlas (such as an AAL template) in the MNI space and resamples it to align with the standardized voxel grid of the whole brain; according to actual planning needs, a specific functional area (such as the left primary motor cortex) is selected from the atlas as a seed point region, and the average value of the blood oxygen correction signal of all voxels in the seed point region is extracted as the preset reference signal; in other specific implementations of this invention, the preset reference signal can also be set based on clinical experience, which is not limited here.

[0074] Preferably, in some possible implementations of the embodiments of the present invention, the method for dividing brain functional regions includes: If the confidence level of the brain functional area is greater than the preset confidence threshold, then the corresponding voxel is classified as a brain functional area. If the confidence level of the brain functional area is less than or equal to the preset confidence threshold, the corresponding voxel will be classified as an abnormal area.

[0075] By setting a confidence threshold, continuous probability values ​​(Pearson correlation coefficient) can be transformed into clear, clinically significant binary decision boundaries, providing a clear basis for dividing brain functional areas. At the same time, by flexibly adjusting the confidence threshold, a balance can be struck between detection sensitivity and false positive control, effectively improving the accuracy of dividing and locating brain functional areas.

[0076] In one specific implementation of this invention, the preset reliability threshold is set to 0.3. This is because the background noise of the resting-state blood oxygenation signal is high, and the correlation coefficient generated by real neural activity is generally low. Therefore, it is not necessary to set the preset reliability threshold too high. Setting it to 0.3 is the standard empirical lower limit for defining effective weak correlation in the field of neuroimaging, used to filter background speckle noise. Below this value, it is considered meaningless random noise, and above this value, it indicates the existence of definite functional connectivity. It should be noted that implementers can also adjust other preset reliability thresholds according to the specific implementation scenario, which is not limited here.

[0077] This invention also provides a hemodynamic-based preoperative imaging brain functional area localization system for neurosurgery; please refer to [link to relevant documentation]. Figure 2 The diagram illustrates a structural block diagram of an image analysis-based vascular repair assessment system for coronary interventional procedures provided by an embodiment of the present invention. The system includes: a data acquisition module S201, a vascular response and structural analysis module S202, a raw blood oxygenation time-series signal correction module S203, and a brain functional area division module S204.

[0078] The data acquisition module S201 is used to acquire a whole-brain standardized voxel grid of the target brain from the human brain medical database; each voxel of the whole-brain standardized voxel grid includes a blood oxygen temporal signal and a diffusion tensor. The vascular response and structure analysis module S202 is used to determine the corresponding vascular response hysteresis value based on the temporal phase shift of the blood oxygenation time-series signal of each voxel; determine the corresponding gradient vector based on the spatial variation trend of the vascular response hysteresis value in the neighborhood of each voxel; perform component analysis on the diffusion tensor of each voxel to determine the corresponding anatomical vector; and determine the corresponding flow structure deviation index based on the mutual deviation between the gradient vector and the anatomical vector at each voxel. The original blood oxygen temporal signal correction module S203 is used to fuse the flow structure deviation index at each voxel with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability value; based on the local permeability value at each voxel and the diffusion tensor at the corresponding voxel, the corresponding whole brain perfusion value is analyzed and determined; based on the whole brain perfusion value at each voxel and the vascular response hysteresis value at the corresponding voxel, the corresponding blood oxygen correction signal is analyzed and determined. The brain functional area division module S204 is used to determine the confidence level of the corresponding brain functional area based on the relative reference offset of the blood oxygen correction signal at each voxel; and to divide the brain functional area based on the value of the confidence level of the brain functional area.

[0079] It should be noted that the system provided in the above embodiments is only illustrated by the division of functional modules. In practical applications, the above functions can be assigned to different modules according to specific needs, that is, the internal structure of the computer device can be flexibly divided to complete all or part of the described functions. In addition, this system embodiment and the foregoing method embodiment are based on the same concept, and their specific implementation process can be found in the detailed description in the method embodiment, which will not be elaborated here.

[0080] It should be noted that the order of the above embodiments of the present invention 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.

[0081] 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 hemodynamic-based method for preoperative imaging localization of brain functional areas in neurosurgery, characterized in that, The method includes: In the human brain medical database, a whole-brain standardized voxel grid of the target brain is obtained; each voxel of the whole-brain standardized voxel grid includes blood oxygenation time-series signal and diffusion tensor. Based on the temporal phase shift of the blood oxygenation time-series signal of each voxel, the corresponding vascular response hysteresis value is determined; based on the spatial variation trend of the vascular response hysteresis value in the neighborhood of each voxel, the corresponding gradient vector is determined; component analysis is performed on the diffusion tensor of each voxel to determine the corresponding anatomical vector; based on the mutual deviation between the gradient vector and the anatomical vector at each voxel, the corresponding flow structure deviation index is determined. The flow structure deviation index at each voxel is fused with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability value; based on the local permeability value at each voxel and the diffusion tensor at the corresponding voxel, the corresponding whole brain perfusion value is determined; based on the whole brain perfusion value at each voxel and the vascular response hysteresis value at the corresponding voxel, the corresponding blood oxygen correction signal is determined. Based on the relative baseline offset of the blood oxygen correction signal at each voxel, the confidence level of the corresponding brain functional area is determined; based on the magnitude of the confidence level of the brain functional area, the brain functional areas are divided.

2. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 1, characterized in that, The method for obtaining the vascular response hysteresis value includes: The blood oxygen time-series signal at each voxel is used as input, and the corresponding complex blood oxygen signal is output using Hilbert transform. The complex blood oxygen signal is then standardized to determine the corresponding standard complex blood oxygen signal. The sum of all principal values ​​of the standard complex blood oxygen signal at each voxel is averaged to determine the vascular response hysteresis value at each voxel.

3. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics as described in claim 1, characterized in that, The method for obtaining the gradient vector includes: The gradient vector is determined by calculating the gradient of all vascular response hysteresis values ​​within the preset neighborhood window of each voxel using the central difference method.

4. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics as described in claim 1, characterized in that, The method for obtaining the anatomical vector includes: The diffusion tensor at each voxel is inversely transformed to determine the guiding reference tensor; the guiding reference tensor is then standardized to determine the guiding tensor at each voxel. The guiding tensor at each voxel is decomposed into eigenvalues ​​to determine the minimum eigenvector corresponding to the minimum eigenvalue at each voxel; the minimum eigenvector is then used as the anatomical vector at each voxel.

5. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 1, characterized in that, The method for obtaining the flow direction structure deviation index includes: The absolute value of the dot product between the gradient vector at each voxel and the anatomical vector at the corresponding voxel is taken as the flow structure deviation value; the product between the magnitude of the gradient vector at each voxel and the magnitude of the anatomical vector at the corresponding voxel is taken as the flow structure deviation magnitude value; the ratio between the flow structure deviation value and the flow structure deviation magnitude value is negatively correlated to determine the corresponding flow structure deviation index.

6. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 1, characterized in that, The method for obtaining the local permeability value includes: The flow direction deviation index at each voxel is positively correlated and mapped to determine the corresponding mapping deviation index; the mapping deviation index at each voxel is weighted and fused with the vascular response hysteresis value at the corresponding voxel to determine the corresponding local permeability reference value; the local permeability reference value is negatively correlated and mapped with the sum of the preset permeability constants to determine the corresponding local permeability value.

7. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 4, characterized in that, The method for obtaining the whole brain perfusion values ​​includes: A cost accumulation equation is constructed based on the local permeation value and the guidance tensor at each voxel; the calculation result of the cost accumulation equation is the cumulative cost at each voxel; the local permeation value is negatively correlated with the cumulative cost; the guidance tensor is positively correlated with the cumulative cost. The cumulative cost at each voxel is standardized to determine the whole-brain perfusion value at each voxel.

8. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 1, characterized in that, The method for obtaining the blood oxygen correction signal includes: The whole brain perfusion value at each voxel is positively correlated to determine the corresponding amplitude compensation factor; the vascular response lag value at each voxel is positively correlated to determine the corresponding phase compensation factor; the corresponding blood oxygen compensation factor is determined based on the product between the amplitude compensation factor and the phase compensation factor; and the blood oxygen correction signal at each voxel is determined based on the product between the blood oxygen compensation factor and the blood oxygen time-series signal.

9. The method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics according to claim 1, characterized in that, The methods for obtaining the confidence scores of the brain functional areas include: The Pearson correlation coefficient between the blood oxygen correction signal and the preset baseline signal at each voxel was used as the confidence level of the brain functional area at each voxel.

10. A method for preoperative imaging localization of brain functional areas in neurosurgery based on hemodynamics, as described in claim 1, characterized in that, The methods for dividing brain functional regions include: If the confidence level of the brain functional area is greater than the preset confidence threshold, then the corresponding voxel is classified as a brain functional area. If the confidence level of the brain functional area is less than or equal to the preset confidence threshold, the corresponding voxel will be classified as an abnormal area.

Images