Method and device for assessing postoperative working memory impairment based on functional magnetic resonance imaging

Abstract

The application provides a postoperative working memory impairment evaluation method and device based on functional magnetic resonance imaging. Functional magnetic resonance imaging data is collected during the execution of a working memory task before and after surgery to obtain a preoperative and postoperative time series signal set; the signals are divided into multiple brain regions according to a brain region division scheme, and the correlation degree between the brain regions is determined based on the signal correlation degree between the brain regions; the brain region correlation degree is screened under multiple threshold conditions, the consistent degree of the brain region belonging to different sets under different thresholds is determined to obtain a target brain region group, and a group characteristic parameter is extracted; the difference between the preoperative and postoperative characteristic parameters is calculated to form a difference parameter set, and the weight is determined based on healthy reference data for weighted fusion to obtain a working memory impairment quantitative index; an impairment degree score is obtained, the change contribution value is determined according to the difference in the connection distribution and the difference in the information interaction ability of each brain region, and a damaged brain region positioning result is generated. The application realizes quantitative evaluation and accurate positioning of postoperative working memory impairment.

Description

Technical Field

[0001] This invention relates to the field of medical image data processing and brain function assessment technology, and in particular to a method and device for assessing postoperative working memory impairment based on functional magnetic resonance imaging. Background Technology

[0002] With the development of neurosurgical procedures and related treatments, postoperative cognitive function assessment has gradually become a key focus in clinical practice. Working memory, as an important higher cognitive function, is significantly affected by brain damage, impacting patients' quality of life. Current techniques for assessing postoperative working memory impairment primarily rely on neuropsychological scales, behavioral tests, and some imaging analysis methods. For example, functional magnetic resonance imaging (fMRI) is used to obtain information on brain region activation or simple connectivity, enabling qualitative or semi-quantitative analysis of the functional state of relevant brain regions to assist clinicians in assessing changes in patients' cognitive function.

[0003] However, existing technologies generally suffer from problems such as highly subjective assessment results, insufficient characterization of complex interactions between brain regions, lack of stable and reliable quantitative indicators, and difficulty in accurately locating damaged brain regions. In particular, they lack systematic analysis methods for changes in brain region relationships and group structure stability under different threshold conditions, resulting in poor repeatability and consistency of assessment results, which makes it difficult to meet the clinical demand for refined and objective assessments.

[0004] Therefore, it is necessary to provide a new method for assessing postoperative working memory impairment. Summary of the Invention

[0005] This application provides a method and device for assessing postoperative working memory impairment based on functional magnetic resonance imaging, so as to improve the objective quantitative accuracy of postoperative working memory impairment assessment and the accuracy of localization of damaged brain regions.

[0006] This application provides a method for assessing postoperative working memory impairment based on functional magnetic resonance imaging, including: In the same subject, functional magnetic resonance imaging data were collected and preprocessed during the execution of a pre-set working memory task before and after the operation to obtain the pre-operation time sequence signal set and the post-operation time sequence signal set. Based on the pre-defined brain region division scheme, the preoperative and postoperative temporal signal sets were divided into multiple brain regions, and the brain region correlation degree was determined based on the correlation degree of signal changes between brain regions, resulting in the preoperative brain region correlation degree set and the postoperative brain region correlation degree set, respectively. Brain region sets were formed by screening the preoperative and postoperative brain region association sets under multiple threshold conditions. Target brain region groups were determined based on the consistency of the sets to which each brain region belonged under different threshold conditions. Feature parameters characterizing the information interaction ability, structural concentration, and inter-group connection strength of each target brain region group were extracted to obtain the preoperative feature parameter set and the postoperative feature parameter set. The difference between the preoperative and postoperative characteristic parameter sets is calculated to form a differential parameter set. The weights of each parameter in the pre-established health reference data are determined, and the differential parameter set is weighted and fused to obtain a quantitative index of working memory impairment. The degree of damage was scored based on the quantitative indicators of working memory impairment. For each brain region, the differences in connectivity distribution and information interaction ability of the brain region relative to different target brain region groups were determined before and after the operation. The contribution value of the changes in the brain region was determined based on the differences in connectivity distribution and information interaction ability. The localization results of the damaged brain region were generated based on the contribution value of the changes in each brain region.

[0007] This application provides a postoperative working memory impairment assessment device based on functional magnetic resonance imaging, comprising: The acquisition unit is used to acquire functional magnetic resonance imaging data and preprocess the data during the execution of a preset working memory task in the same subject before and after the operation to obtain the preoperative time-series signal set and the postoperative time-series signal set. The division unit is used to divide the preoperative and postoperative temporal signal sets into multiple brain regions according to the preset brain region division scheme, and to determine the brain region correlation degree based on the correlation degree of signal changes between brain regions, so as to obtain the preoperative brain region correlation degree set and the postoperative brain region correlation degree set respectively. The screening unit is used to screen the preoperative brain region association set and the postoperative brain region association set under multiple threshold conditions to form a brain region set. The target brain region group is determined according to the consistency of the set to which each brain region belongs under different threshold conditions. Feature parameters that characterize the information interaction ability, structural concentration and inter-group connection strength of each target brain region group are extracted to obtain the preoperative feature parameter set and the postoperative feature parameter set. The unit is used to calculate the difference between the preoperative feature parameter set and the postoperative feature parameter set to form a difference parameter set. The weights of each parameter in the pre-established health reference data are determined, and the difference parameter set is weighted and fused to obtain a quantitative index of working memory impairment. The generation unit is used to obtain a score of the degree of damage based on the quantitative index of working memory impairment, and to determine the differences in connectivity distribution and information interaction ability of each brain region relative to different target brain region groups before and after surgery. Based on the differences in connectivity distribution and information interaction ability, the change contribution value of the brain region is determined, and the localization result of the damaged brain region is generated based on the change contribution value of each brain region.

[0008] The beneficial effects of this application include: (1) By acquiring functional magnetic resonance imaging data from the same subjects before and after surgery and conducting comparative analysis, an objective quantitative assessment of changes in working memory function can be achieved, avoiding the assessment bias caused by relying solely on subjective scales and improving the accuracy and reliability of the assessment results. (2) By screening brain region associations under multiple threshold conditions and determining target brain region groups based on consistency, the unstable impact of single threshold selection can be effectively reduced, thereby improving the robustness and consistency of brain region group division. (3) By constructing a set of differential parameters and combining the weights determined by the degree of parameter dispersion in healthy reference data for weighted fusion, key change characteristics can be highlighted, individual differences and noise interference can be suppressed, thereby improving the discriminative ability of quantitative indicators of working memory impairment. (4) By simultaneously considering the differences in brain region connectivity distribution and differences in information interaction ability to determine the contribution value of changes, a refined localization of damaged brain regions can be achieved, providing clear information on the distribution of functional impairment for clinical practice and helping to formulate subsequent rehabilitation intervention strategies. Attached Figure Description

[0009] Figure 1 This is a flowchart of a postoperative working memory impairment assessment method based on functional magnetic resonance imaging provided in the first embodiment of this application.

[0010] Figure 2 This is a schematic diagram of a postoperative working memory impairment assessment device based on functional magnetic resonance imaging provided in the second embodiment of this application. Detailed Implementation

[0011] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0012] The first embodiment of this application provides a method for assessing postoperative working memory impairment based on functional magnetic resonance imaging. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a detailed description of a method for quantitatively assessing postoperative working memory impairment based on functional magnetic resonance imaging.

[0013] Step S101: In the same subject, functional magnetic resonance imaging data are collected and preprocessed during the execution of a preset working memory task before and after the operation to obtain the preoperative time sequence signal set and the postoperative time sequence signal set.

[0014] Step S101 is the data acquisition and standardization step of the entire method. Its implementation must be completed under strict control of consistent experimental conditions to ensure the comparability of preoperative and postoperative data. This step first requires identifying the subjects, generally patients undergoing neurosurgery or related treatments. Functional magnetic resonance imaging (fMRI) data should be acquired within one week before surgery and again during the stable recovery period after surgery (e.g., 1 to 4 weeks postoperatively, depending on the specific clinical protocol). To ensure consistency in the two acquisitions, the same MRI scanner should be used, preferably a 3.0T MRI scanner, with identical scanning parameters, including a repetition time (TR) of approximately 2000 ms, an echo time (TE) of approximately 30 ms, a flip angle of approximately 90°, a slice thickness of 3-4 mm, an interslice spacing of 0-1 mm, an acquisition matrix of 64×64 or higher, and a field of view covering the entire brain.

[0015] During data acquisition, subjects are required to perform a pre-defined working memory task. This task can employ a standard N-back task paradigm, such as a 2-back or 3-back task. Specifically, during the scanning process, a sequence of stimuli (such as letters or numbers) is presented visually or auditorily. Each stimulus lasts approximately 500ms, with a stimulus interval of approximately 1500ms. Subjects must determine whether the current stimulus is the same as the Nth stimulus previously and respond via a button. The entire task can employ a block design or an event-related design; for example, each task block lasts 20-30 seconds, with a 10-20 second rest block, and the total acquisition time is controlled to 5-10 minutes to obtain stable BOLD signal changes.

[0016] The acquired raw functional magnetic resonance imaging (fMRI) data is typically a four-dimensional data sequence (three spatial dimensions plus a temporal dimension), requiring further preprocessing to eliminate interference factors not related to neural activity. First, temporal alignment is performed, i.e., the slice data acquired at each time point are corrected for temporal sequence to align signals at the same time point in the temporal dimension. Then, head motion correction is performed, using rigid body registration to register the images at each time point to a reference image (usually the first or intermediate time point), correcting for minor head movements of the subject during the scan. Typically, translation should not exceed 2 mm and rotation should not exceed 2°. If these limits are exceeded, the data at that time point can be discarded or marked.

[0017] Next, spatial normalization is performed, which maps individual brain imaging data to a standard brain space (such as the MNI standard space). Nonlinear registration methods are used to unify individual structural differences, so that data from different subjects and at different time points have a unified spatial reference. Subsequently, spatial smoothing can be performed, for example, by using a 6mm full-width half-height (FWHM) Gaussian kernel for convolution, to improve the signal-to-noise ratio and enhance the consistency of adjacent voxel signals.

[0018] In the time dimension, detrending and filtering processing are also required. Detrending is used to remove low-frequency drift in the signal, typically employing linear or quadratic trend term regression. Filtering can use bandpass filtering, for example, retaining signal components in the 0.01Hz to 0.1Hz range to remove high-frequency noise and extremely low-frequency drift. Simultaneously, regression analysis can be performed on cerebrospinal fluid signals, white matter signals, and head movement parameters to further reduce the influence of non-neuronal signals.

[0019] After the above processing, a time-varying signal sequence can be obtained for each voxel. These voxel-level signals are then preserved as a standardized temporal data set according to the needs of subsequent steps. The data acquired and processed preoperatively constitutes the preoperative temporal signal set, and the data acquired and processed postoperatively constitutes the postoperative temporal signal set. Both temporal signal sets maintain consistency in spatial coordinates, temporal length, and processing flow, thus providing a reliable and comparable data foundation for subsequent brain region segmentation, correlation calculation, and differential analysis.

[0020] Furthermore, in the same subject, functional magnetic resonance imaging data are collected and preprocessed during the execution of a preset working memory task before and after surgery to obtain a preoperative time-series signal set and a postoperative time-series signal set, including: While acquiring functional magnetic resonance imaging (fMRI) data, the time series of stimulus presentation and behavioral response of subjects in working memory tasks were recorded. Based on the stimulus presentation time series, the acquired fMRI data were time-synchronized and labeled. The brain signal data corresponding to each stimulus presentation were aligned to obtain the original time-series signal sequence that corresponds one-to-one with the task events. Based on the original time-series signal sequence, the brain signal data at each time point is screened for task response consistency. Signal data corresponding to time periods in which the subjects did not make correct responses or whose reaction times exceeded the preset range are removed. The retained signal data are then reconstructed into a continuous time-series signal in chronological order to obtain the time-series signal sequence after task response correction. For the time-series signal sequences, a consistent preprocessing procedure is performed on the preoperative and postoperative data, including spatial registration of brain signal data at each time point to eliminate the influence of head movement of the subjects, and frequency range filtering of the registered signal data to retain signal components within the preset frequency band, thereby obtaining a preoperative time-series signal set and a postoperative time-series signal set with a unified time and spatial reference.

[0021] First, a unified clock reference needs to be established between the task presentation system and the MRI scanning system before the functional magnetic resonance imaging (fMRI) scan begins. The "stimulus presentation time series" refers to the ordered time record of each visual or auditory stimulus presented to the subject during the working memory task. For visual working memory tasks, such as N-back tasks using letters, numbers, shapes, or positional sequences, the start time, duration, disappearance time, and interval between adjacent stimuli for each stimulus on the display screen must be recorded. For auditory working memory tasks, the start time, end time, and time interval between adjacent sound stimuli for each sound stimulus must also be recorded. This time record should use a unified timestamp format compatible with the MRI host, for example, setting the scan start time as zero and recording the occurrence time of each subsequent stimulus presentation as a millisecond-level time value relative to that zero. Simultaneously, the "behavioral response time series" also needs to be recorded. This "behavioral response time series" refers to the time record of the subject's key presses, trigger inputs, or other explicit feedback actions in response to each stimulus during the task. This sequence not only records whether a response is made, but also the exact time of the response and which stimulus event the response corresponds to. Preferably, the stimulus presentation software automatically generates an event number each time a stimulus occurs, and binds and stores this number with the response time when the subject makes a response, thereby forming a traceable stimulus-response correspondence.

[0022] While acquiring functional magnetic resonance imaging (fMRI) data, it is necessary to synchronize each stimulus presentation with the timeline of the MRI scan. The "time synchronization marker" referred to here is a marker information that establishes a precise correspondence between the task event time and the scan sampling time. Since fMRI data is typically acquired frame-by-frame in time-series format, with each frame corresponding to a sampling moment, it is necessary to determine which frame of data each stimulus presentation occurs within, and which frames in several consecutive sampling frames after the stimulus primarily reflect the brain activity induced by the task event. Preferably, when the scanning system outputs the sampling trigger signal for each frame, the trigger signal is simultaneously sent to the task presentation system, which then directly maps the stimulus time to the frame number. If the scan repetition time is two seconds, then one frame of whole-brain volume data is obtained every two seconds. Assuming the stimulus is presented at 8.4 seconds after the start of the scan, the stimulus start time is within the fifth frame sampling period. In this embodiment, not only is the starting frame into which the stimulus falls recorded, but also, preferably, multiple frames of signals within a preset time window after the stimulus occur are associated and marked based on the hysteresis characteristics of the fMRI BOLD signal. For example, using three to eight seconds after stimulus presentation as the primary response time window, the corresponding two to three frames of functional magnetic resonance imaging (fMRI) data can be labeled as data frames related to that stimulus. After the above processing, a "raw temporal signal sequence corresponding one-to-one with task events" can be obtained. The "raw temporal signal sequence corresponding one-to-one with task events" refers to fMRI scan data that is no longer simply arranged in natural chronological order, but rather clearly indicates which stimulus event and its subsequent response window each segment of brain signal corresponds to. This one-to-one correspondence can be expressed as follows: the j-th task event corresponds to brain signal data from frame m to frame n, and the (j+1)-th task event corresponds to brain signal data from frame p to frame q.

[0023] After obtaining the original time-series signal sequence, a "task response consistency screening" is required. This screening refers to judging the validity of synchronized brain signal segments based on the subjects' actual behavioral performance in each task event, retaining only those corresponding to valid task execution. "Valid task execution" must meet at least two conditions: first, the subject's response to the stimulus is correct according to the task rules; second, the reaction time falls within a preset reasonable range. The "preset range" refers to the allowable reaction time window pre-set based on task difficulty, subject group characteristics, and device timing conditions. For example, for the visual 2-back task, a correct button press response between 300 and 1500 milliseconds after the stimulus appears can be considered a valid response; responses earlier than 300 milliseconds may be due to prediction or mis-triggering, and responses later than 1500 milliseconds may indicate cognitive processing delay or a lagging response to the previous stimulus; neither of these situations should be considered valid behavior for that task event. If a task requires the subject to press a button when the current stimulus is the same as the previous two stimuli, and the subject fails to press a button, or presses a button when they should not, then the behavioral response to that stimulus is judged as an incorrect response. For the time period corresponding to the incorrect response, the brain signal segments related to that stimulus should be removed; the time periods corresponding to no response or overdue response should also be removed.

[0024] The elimination here is not simply deleting a single time point, but rather deleting the entire brain signal window corresponding to an invalid task event. For example, if the twelfth stimulus event corresponds to frames 10 to 12 of functional magnetic resonance imaging (fMRI) data, and the subject responded incorrectly to that stimulus, then the data segment corresponding to frames 10 to 12 is marked as invalid and will not be included in the analysis set when constructing the temporal sequence signal. For instance, assuming a total of 60 stimuli were performed in the preoperative scan, with 48 valid responses, 6 incorrect responses, and 6 no response or timed-out responses, then the brain signal segments corresponding to these 48 valid events should be retained, and the brain signal segments corresponding to the remaining 12 events should be deleted. If each valid event corresponds to three frames of primary response data, then the final set of valid brain signal data consists of 144 frames. These retained data need to be rearranged according to the chronological order of the events to form a "task response corrected temporal sequence signal." The term "reconstructing a continuous temporal sequence signal" here does not require forcibly filling in the eliminated time period in physical time, but rather refers to sequentially splicing all valid segments according to the original task occurrence order to form a continuous analysis sequence containing only valid task response data. If a uniform length is required, invalid markers can be filled at the locations of the removed fragments, and these marker locations can be ignored in subsequent processing. However, it is preferable to retain only valid fragments to reduce the interference of invalid data on subsequent correlation analysis.

[0025] After obtaining the time-series signal sequence after task response correction, a consistent preprocessing procedure needs to be performed on both preoperative and postoperative data to eliminate interference factors unrelated to actual brain function changes. First, spatial registration is performed. "Spatial registration" refers to adjusting brain images from different sampling times, scanning stages, and even different time points to a consistent position in three-dimensional space to eliminate spatial offsets caused by minor head movements, posture changes, and differences in preoperative and postoperative scanning positioning. Specifically, a reference image can be selected as the registration benchmark, preferably a frame with high signal quality at an intermediate time point. Then, rigid body transformation estimation is performed between each brain signal image and the reference frame, obtaining three translation parameters and three rotation parameters to achieve optimal spatial overlap between the image and the reference frame. "Optimal overlap" can be achieved by maximizing image similarity metrics, such as maximizing normalized mutual information or minimizing the sum of squared gray-level differences. This implementation does not limit the specific similarity measurement method, but requires the same method to be used preoperatively and postoperatively. If the translation of a frame relative to the reference frame exceeds a preset threshold, such as two millimeters, or the rotation exceeds a preset threshold, such as two degrees, then the frame can be marked as a high-motion-interference frame and removed or interpolated as needed. To ensure temporal integrity, it is preferable to perform a smooth transition by combining adjacent valid frames after removing high-motion frames. However, if high-motion frames appear consecutively and are numerous, the corresponding task event can also be directly determined as invalid.

[0026] After spatial registration, frequency range filtering is required. This "frequency range filtering" refers to decomposing brain signals into different frequency components along the time dimension, retaining only signal components within a preset frequency band, and suppressing very low-frequency drift and high-frequency noise. In task-related analyses of functional magnetic resonance imaging (fMRI) BOLD signals, effective variations are typically concentrated in the lower frequency range. Therefore, it is preferable to perform bandpass filtering on each voxel or subsequent time-series signal used for analysis. For example, frequency components between 0.01 Hz and 0.1 Hz can be retained, filtering out slow drift below 0.01 Hz and high-frequency noise above 0.1 Hz. Of course, if the task uses shorter event intervals, the upper limit frequency can be adjusted appropriately according to the task design, but the same frequency filtering interval must be used pre- and post-operatively. In practice, finite impulse response (FIR) filters or infinite impulse response (IR) filters can be used to process the time series point by point, ensuring that the output signal retains only variations within the preset frequency band. For example, if the original signal of a voxel at one hundred consecutive time points contains both a slow drift trend and a rapid random fluctuation, then after frequency range filtering, what is retained is the low-to-medium frequency stable change part related to task execution, thereby improving the reliability of subsequent brain region correlation calculations.

[0027] After the aforementioned time synchronization marking, task response consistency screening, spatial registration, and frequency range screening, the preoperative and postoperative temporal signal sets can be obtained respectively. The term "unified time reference" refers to the complete consistency between preoperative and postoperative temporal signals in event alignment rules, effective response judgment rules, time window truncation methods, and temporal sequence construction methods; the term "unified spatial reference" refers to the complete consistency between preoperative and postoperative images in registration reference, spatial coordinate system, and voxel position correspondence. In other words, the preoperative and postoperative signals do not simply originate from the same subject, but are two directly comparable data sets formed under the same task logic, the same behavioral screening conditions, the same spatial registration standard, and the same frequency screening rules. The preoperative temporal signal set can be represented as a continuous time series set containing only brain signals corresponding to effective preoperative task events, while the postoperative temporal signal set is the postoperative corresponding set constructed according to the exact same rules.

[0028] Step S102: Based on the preset brain region division scheme, the preoperative temporal signal set and the postoperative temporal signal set are divided into multiple brain regions, and the brain region correlation degree is determined based on the correlation degree of signal changes between brain regions, so as to obtain the preoperative brain region correlation degree set and the postoperative brain region correlation degree set respectively.

[0029] Step S102 involves structurally dividing the whole-brain signals and establishing relationships between brain regions based on the preoperative and postoperative temporal signal sets obtained in step S101. Its implementation requires ensuring spatial consistency and consistency in computational methods to ensure comparability between preoperative and postoperative results. First, a pre-defined brain region division scheme needs to be determined. This scheme can use publicly available standard brain region templates, such as multi-brain region atlases constructed based on standard brain space. Specifically, a three-dimensional labeled image containing several brain regions can be used. Each voxel in this image is assigned a unique brain region number and is located in the same spatial coordinate system as the functional magnetic resonance imaging (fMRI) data after spatial normalization in step S101.

[0030] In practice, the brain region segmentation template is loaded into the same spatial resolution and coordinate system as the temporal signal set. For each voxel in the preoperative temporal signal set, the corresponding brain region number is found based on its spatial coordinates, and voxels with the same brain region number are grouped into the same brain region. For each brain region, the signal values ​​of all voxels within that region at each time point are summarized, preferably using an arithmetic mean to obtain the representative signal value of that brain region at that time point, thereby generating a time-varying signal sequence for each brain region. By repeating the above process for all brain regions, the preoperative temporal signal set can be transformed into a set of signal change sequences for multiple brain regions. The postoperative temporal signal set is also processed in the same way, ensuring that the two sets are completely consistent in the number, numbering, and spatial location of brain regions.

[0031] After obtaining the signal change sequences of each brain region, it is necessary to further determine the degree of correlation between brain regions. Specifically, for any two different brain regions, their corresponding signal change sequences are taken, and their correlation is calculated over time. The Pearson correlation coefficient is preferably used as a measure of correlation; that is, after removing the mean from the signal sequences of the two brain regions at each time point, the degree of their coordinated changes is calculated. For any pair of brain regions, a value between -1 and 1 can be obtained to characterize the consistency of their signal changes. The closer this value is to 1, the more consistent the signal change trends of the two brain regions during the task; the closer it is to -1, the opposite the trends; and the closer it is to 0, the weaker the correlation.

[0032] Repeating the above calculation process for each pair of brain regions yields correlation results covering all brain region combinations. To ensure the comparability of preoperative and postoperative results, the same calculation procedures must be performed independently on both the preoperative and postoperative time-series signal sets, including the same brain region segmentation template, the same signal extraction method, and the same correlation calculation method. Ultimately, for the preoperative data, a set containing the pairwise correlations of all brain regions is obtained, recording the degree of correlation between each pair of brain regions; for the postoperative data, a similarly structured set of correlations is obtained.

[0033] In the actual storage and implementation process, the aforementioned brain region correlation set can adopt any data structure that can represent the correspondence between brain region pairs. For example, it can be recorded using brain region number pairs as indices and correlation values ​​as corresponding values, or it can be stored in a two-dimensional array according to the brain region order. Regardless of the storage form used, the brain region numbering order must be kept consistent so that the correlation between preoperative and postoperative values ​​can be matched and compared one by one in subsequent steps. Through the above processing, the preoperative brain region correlation set and the postoperative brain region correlation set can be obtained respectively, providing basic data for subsequent threshold-based screening, group division, and feature extraction.

[0034] Furthermore, based on a preset brain region division scheme, the preoperative and postoperative temporal signal sets are divided into multiple brain regions, and the brain region correlation is determined based on the correlation between signal changes between brain regions, resulting in a preoperative brain region correlation set and a postoperative brain region correlation set, including: Based on the preoperative and postoperative temporal signal sets, the whole brain is divided into multiple pre-defined brain regions according to a unified spatial coordinate system. The signal values ​​of all voxels in each brain region at each time point are weighted and summarized. The weighting coefficient is determined according to the distance between the spatial position of each voxel in the brain region and the center position of the brain region, thereby obtaining the representative temporal signal sequence corresponding to each brain region. For the representative temporal signal sequence of each brain region, time delay correction processing is performed. By gradually shifting the signal sequence within a preset time offset range and calculating the degree of matching with the original sequence, the time offset that makes the signal fluctuation characteristics most consistent is determined. Based on the time offset, the brain region temporal signal is corrected to obtain the time-aligned brain region temporal signal sequence. Based on the time-aligned brain region temporal signal sequence, the segmented correlation of signal changes between any two brain regions is calculated. The entire time series is divided into multiple continuous time periods, and the correlation between the signal sequences of two brain regions is calculated in each time period. Then, the correlation of each time period is weighted and averaged to obtain the brain region correlation value that reflects local temporal consistency, thus forming a brain region correlation set.

[0035] First, the entire brain needs to be divided into regions using a unified spatial coordinate system. This "unified spatial coordinate system" refers to preoperative and postoperative functional magnetic resonance imaging (fMRI) data that have been spatially normalized and mapped to the same standard space, such as a standard template space, ensuring that the same spatial location has consistent anatomical significance before and after surgery. Based on this, a pre-defined brain region division scheme is used to divide the entire brain volume into multiple non-overlapping regions. Each brain region consists of several voxels, which spatially belong to the same anatomical or functional area.

[0036] After dividing the brain regions, the voxel signals within each region need to be aggregated to obtain a representative temporal signal sequence for that region. The "representative temporal signal sequence" refers to a time series that describes the overall signal changes of the brain region. To improve the stability of this representative signal, this implementation does not directly use simple averaging, but instead introduces a weighted aggregation method based on spatial location. Specifically, the center location of each brain region needs to be determined first. This center location can be obtained by calculating the average of the spatial coordinates of all voxels within the region. For example, if a brain region contains several voxels with spatial coordinates (x1, y1, z1), (x2, y2, z2), etc., the center coordinates of the brain region can be obtained by averaging each coordinate component. Subsequently, for each voxel within the brain region, the spatial distance between it and the center of the brain region is calculated. This distance can be calculated using three-dimensional Euclidean distance, i.e., taking the square root of the sum of the squares of the coordinate differences in the three directions.

[0037] After obtaining the distances between each voxel and the brain region center, weighting coefficients need to be determined based on these distances. Preferably, a method of increasing weight for closer voxels can be used, such as setting the weights as the reciprocal of the distance or a distance-decreasing function. For example, if the distance from the center of a brain region to voxel A is 1 unit and the distance to voxel B is 2 units, then voxel A can be assigned a weight of 1 / 1 and voxel B a weight of 1 / 2, respectively. All weights are then normalized so that the sum of the weights is 1. Subsequently, at each time point, the signal values ​​of all voxels within the brain region are multiplied by their corresponding weights and summed to obtain the representative signal value of the brain region at that time point. Repeating this process for all time points yields the representative temporal signal sequence of the brain region. The same processing method is used for both preoperative and postoperative data to obtain the corresponding brain region-level temporal signal sequence set.

[0038] After obtaining representative temporal signal sequences of brain regions, time delay correction is required. "Time delay correction" refers to eliminating the time lag differences in neural activity responses between different brain regions, aligning signal changes in each region as closely as possible in time. Specifically, for each brain region's temporal signal sequence, a translation operation is performed within a preset time offset range. This "time offset range" can be set according to the scan repetition time and task characteristics; for example, it can be set to a range of several sampling points forward or backward. Under each offset condition, the original temporal signal sequence is shifted forward or backward by the corresponding number of time points, and the degree of matching between the translated sequence and the original sequence is calculated. The "degree of matching" refers to the consistency in the trend of change between the two time sequences, which can be achieved by calculating the correlation between the two sequences.

[0039] To illustrate, if the original signal sequence of a brain region is a set of numerical sequences that change over time, we can calculate the correlation between the shifted sequence and the original sequence in three cases: shifting the sequence forward by one time point, shifting it backward by one time point, and not shifting it at all. Assuming the correlation is highest when shifting backward by one time point, we consider that there is a time delay in this brain region, and the signal sequence of that region should be time-corrected accordingly. This means shifting the entire sequence forward by one time point to better align it with other brain regions in time. Repeating this process for all brain regions yields the time-aligned temporal signal sequence of the brain region.

[0040] After time delay correction, it is necessary to calculate the correlation between brain regions. Unlike the traditional method of directly calculating the correlation of the entire time series, this implementation adopts a segmented correlation calculation method. Specifically, the time-aligned brain region temporal signal sequence is divided into multiple consecutive time segments according to chronological order. Each time segment contains several consecutive time points, which can be divided, for example, based on task structure or a fixed length. For any two brain regions, the correlation between their temporal signals is calculated within each time segment, thereby obtaining correlation values ​​corresponding to multiple time segments. The correlation here can employ conventional correlation calculation methods, such as comparing the signal change trends within two time segments to obtain a value reflecting consistency.

[0041] After obtaining the correlation levels for each time period, these correlation values ​​need to be weighted and averaged to obtain the final brain region correlation value. The weighted average can be set according to the importance of each time period; for example, different weights can be assigned based on the magnitude of signal changes or task relevance within the time period. If the weights for each time period are equal, the correlation values ​​for each time period can be directly averaged. For example, if the correlation levels of two brain regions in the three time periods are 0.6, 0.8, and 0.4 respectively, their average can be approximately 0.6, which can be used as the correlation between these two brain regions. This segmented calculation and summarization method avoids the excessive influence of abnormal fluctuations in a particular time period on the overall result, thus obtaining a more stable correlation estimate.

[0042] By performing the above segmented correlation calculation and weighted averaging on each pair of all brain regions, a complete set of brain region correlations can be formed. This set is calculated independently before and after surgery, and is completely consistent in terms of brain region division, time correction method, and correlation calculation rules, thus ensuring direct comparability between the two.

[0043] Step S103: Screen the preoperative brain region association set and the postoperative brain region association set under multiple threshold conditions to form a brain region set. Determine the target brain region group based on the consistency of the set to which each brain region belongs under different threshold conditions. Extract feature parameters that characterize the information interaction ability, structural concentration degree and inter-group connection strength of each target brain region group to obtain the preoperative feature parameter set and the postoperative feature parameter set.

[0044] Step S103, based on the preoperative and postoperative brain region correlation sets obtained in step S102, involves extracting stable structures and quantifying features of the correlations between brain regions. Its core lies in identifying brain region groups with stable tissue structures through consistency analysis under multiple threshold conditions, and further extracting feature parameters that reflect the functional states within and between these groups. This step requires that preoperative and postoperative data undergo completely consistent processing independently to ensure the reliability of subsequent difference analysis.

[0045] In practice, multiple threshold conditions are first introduced for the set of brain region connectivity. These thresholds are used to filter the connectivity relationships between brain regions. Preferably, several thresholds are selected that are evenly distributed within a preset range; for example, several increasing thresholds can be selected within the connectivity value range, so that each threshold corresponds to a different filtering intensity. For each threshold, all brain region pairs in the brain region connectivity set with connectivity greater than or equal to that threshold are retained, while those with connectivity less than that threshold are removed, thus obtaining a subset of brain region connectivity relationships under that threshold condition. Based on this subset, all brain regions can be connected according to their retained connectivity relationships, ensuring that any two brain regions in the same set have a connected path through the retained connectivity relationships, thereby forming several brain region sets under that threshold condition.

[0046] By applying the above processing to each preset threshold, brain region set partitioning results under multiple threshold conditions can be obtained. Since different thresholds can cause changes in the partitioning structure of brain region sets, statistical analysis of the assignment of the same brain region under different threshold conditions is necessary. Specifically, for each brain region, the identification information of the brain region set it belongs to under each threshold condition is recorded, and other brain regions belonging to the same set under different threshold conditions are statistically analyzed. If certain brain regions appear in the same set as this brain region under multiple threshold conditions, it indicates that these brain regions have high structural stability. By statistically analyzing this co-occurrence relationship, a stability description result can be constructed for each brain region. For example, the number of times this brain region co-occurs with other brain regions in the same set under different threshold conditions can be counted, and this number can be normalized to obtain an index reflecting the degree of stable co-occurrence between brain regions.

[0047] After obtaining the above stability description, target brain region groups can be further identified. Specifically, brain regions can be aggregated according to their degree of stable co-occurrence, and a group of brain regions whose stable co-occurrence exceeds a preset criterion can be grouped into the same target brain region group. The preset criterion can be set empirically or determined through analysis of stability distribution; for example, selecting a combination of brain regions with a high level of stable co-occurrence ensures that the obtained target brain region groups have high consistency under different threshold conditions. Through this process, several structurally stable brain region groups can be obtained, reflecting a set of brain regions that maintain stable connectivity under different screening conditions.

[0048] After identifying the target brain region groups, it is necessary to further extract parameters that reflect their functional characteristics for each group. Specifically, firstly, for each target brain region group, based on the brain region correlation set obtained in step S102, the correlation relationships between brain regions within the group are extracted, and these correlation relationships are summarized to obtain parameters reflecting the information interaction capability within the group. For example, the correlation degrees of all brain region pairs within the group can be averaged or weighted to obtain the overall information interaction level within the group. Secondly, parameters reflecting the degree of structural concentration can be calculated based on the distribution of correlation relationships between brain regions within the group. For example, the tightness of the group structure can be characterized by analyzing the density or evenness of the correlation relationships within the group. Thirdly, for different target brain region groups, cross-group brain region correlation relationships can be extracted, and these cross-group correlation relationships are summarized to obtain parameters reflecting the strength of connections between groups. These parameters are used to characterize the degree of information transmission or functional coupling between different groups.

[0049] The extraction of the aforementioned feature parameters needs to be performed independently on both the preoperative and postoperative brain region correlation sets, using completely consistent calculation methods and parameter settings. Ultimately, for preoperative data, a preoperative feature parameter set consisting of feature parameters corresponding to multiple target brain region groups can be obtained; for postoperative data, a postoperative feature parameter set corresponding to the structures can be obtained. The two feature parameter sets maintain consistency in parameter type, group correspondence, and calculation methods, thus providing a foundation for the quantitative analysis of preoperative and postoperative functional differences in subsequent steps.

[0050] To further illustrate the specific implementation of this step, a concrete example is provided below. Assume that in step S102, the whole brain has been divided into several brain regions based on a standard brain region segmentation scheme, and preoperative and postoperative brain region correlation sets have been obtained. Taking the preoperative brain region correlation set as an example, multiple threshold conditions are first selected, such as several thresholds that gradually increase from low to high, to filter the correlation between brain regions. Under the first lower threshold condition, all brain region pairs with correlation higher than that threshold are retained. At this time, due to the relatively lenient screening conditions, the resulting brain region set is usually large, and multiple brain regions may be classified into the same set. Under the higher threshold condition, only brain region pairs with strong correlation are retained. At this time, the brain region set is divided more finely, the number of sets increases, and the number of brain regions contained in each set decreases.

[0051] Under each threshold condition, based on the preserved brain region connections, brain regions with connectivity are divided into several brain region sets. For example, under a certain threshold, if brain region A has a preserved connection with brain region B, and brain region B also has a preserved connection with brain region C, then brain regions A, B, and C are classified into the same brain region set; while other brain regions without connectivity are classified into different sets. Repeating the above process for all threshold conditions yields multiple brain region set classification results.

[0052] Subsequently, the set affiliation of each brain region under different threshold conditions is statistically analyzed. For example, for brain region A, the instances where it co-occurs with other brain regions under various threshold conditions are recorded. The number of threshold conditions under which brain region A and brain region B are classified into the same set is counted, and the statistical results are then normalized to obtain the stable co-occurrence degree between brain regions A and B. Performing the same operation on all brain region pairs yields descriptive results reflecting the stable relationships between brain regions.

[0053] Based on this, brain regions can be grouped according to their degree of stable co-occurrence. For example, a group of brain regions with a high degree of stable co-occurrence can be classified into the same target brain region group. If brain regions A, B, and C are classified into the same set under most threshold conditions, then they are identified as one target brain region group; while brain regions D, E, and F form another stable set under most threshold conditions, and are thus classified as another target brain region group. In this way, several structurally stable brain region groups can be obtained.

[0054] After the target brain region groups are divided, specific feature parameters need to be extracted for each group, ensuring that the extraction process has clear calculation rules and a unified data source. In practice, for any target brain region group, the correlation data between all brain regions within that group is first extracted from the brain region correlation set obtained in step S102. For example, if a target brain region group includes brain regions A, B, and C, the correlation values ​​between brain regions A and B, A and C, and B and C are extracted. These correlation values ​​are then aggregated, preferably using an arithmetic mean, to obtain the average correlation value within the group. This average value is used to characterize the information interaction capability of the group. Based on this, the distribution of these correlation values ​​can be further calculated, such as the degree of deviation of each correlation value from the average or the degree of uniformity of distribution, thereby obtaining a parameter reflecting the degree of structural concentration of the group. This parameter is used to characterize whether the connections within the group are balanced or dominated by a few strong connections.

[0055] Next, it is necessary to calculate the connection strength between groups. In practice, for any two different target brain region groups, the correlation values ​​between all brain regions in one group and all brain regions in the other group are extracted. For example, for brain regions A and B in group 1 and brain regions D and E in group 2, the correlation values ​​between A and D, A and E, B and D, and B and E are extracted, and these correlation values ​​are summarized, preferably using an average or weighted average method, to obtain the connection strength parameter between the two groups. Repeating the above operation for all group combinations yields a complete set of inter-group connection strengths, which reflects the information interaction between different functional groups.

[0056] In practical implementation, all the above-mentioned feature parameters should be stored in a uniform order. For example, the information interaction ability parameters, structural concentration parameters, and connection strength parameters with other groups should be arranged sequentially according to the target brain region group number, thereby constructing a feature parameter set with consistent structure. For the preoperative brain region correlation set and the postoperative brain region correlation set, the same brain region group division results and completely consistent calculation methods must be strictly used for feature extraction to ensure that the preoperative feature parameter set and the postoperative feature parameter set correspond completely in terms of parameter type, order, and quantity.

[0057] Furthermore, the preoperative and postoperative brain region association sets are screened under multiple threshold conditions to form brain region sets. Target brain region groups are determined based on the consistency of the sets to which each brain region belongs under different threshold conditions. Feature parameters characterizing the information interaction ability, structural concentration, and inter-group connection strength of each target brain region group are extracted to obtain preoperative and postoperative feature parameter sets, including: Based on the brain region association set, multiple increasing threshold sequences are generated according to the preset association value range. Under each threshold condition, brain region association pairs that satisfy the association degree greater than or equal to the threshold are selected. Based on the selected brain region association pairs, brain region connectivity is constructed and the corresponding brain region set under the threshold is obtained, forming a multi-threshold brain region set sequence. For a multi-threshold brain region set sequence, the number of the brain region set to which each brain region belongs under each threshold condition is recorded. Based on the ratio of the number of times brain regions jointly belong to the same brain region set under different threshold conditions to the total number of thresholds, the stable co-occurrence coefficient between any two brain regions is calculated, forming a set of stable co-occurrence relationships between brain regions. Based on the set of stable co-occurrence relationships of brain regions, brain regions with stable co-occurrence coefficients greater than a preset stability threshold are aggregated to generate multiple candidate brain region groups. The stable co-occurrence coefficients of all brain region pairs within each candidate brain region group are averaged to obtain a group stability index. Candidate brain region groups whose group stability index meets preset conditions are selected as target brain region groups. For each target brain region group, the correlation between brain regions within the group is extracted based on the brain region correlation set, and the average value is calculated to obtain the information interaction capability parameter. The structural concentration parameter is calculated based on the dispersion of the correlation within the group, and the inter-group connection strength parameter is calculated based on the average value of the brain region correlation between different target brain region groups, thus forming the corresponding feature parameter set.

[0058] In this embodiment, the steps involve repeatedly constructing brain region connectivity structures under different screening intensities based on the already obtained set of brain region associations, and further utilizing cross-condition stability to determine brain region groups with reliable functional significance. This avoids the randomness caused by a single screening condition and provides a stable structural basis for subsequent feature extraction.

[0059] In practice, multiple incremental thresholds are first generated based on a set of brain region correlation scores. Each pair of brain regions in this set corresponds to a numerical value representing the degree of consistency in their signal changes. To enable systematic analysis of connectivity structures under different screening intensities, multiple thresholds are selected within this numerical range and arranged in ascending order to form a threshold sequence. Each threshold in this sequence corresponds to a screening condition; under this condition, only brain region correlation pairs with a correlation score greater than or equal to that threshold are retained, and other correlations are not included in the current structure construction.

[0060] Under each threshold condition, the connectivity relationships between brain regions are constructed based on the selected brain region association pairs. Here, "brain region connectivity relationship" refers to the degree of association between two brain regions that satisfies the threshold condition; that is, a connection is considered to exist between them. Based on this, all brain regions are divided according to connectivity relationships, such that any two brain regions that can be reached from each other through one or more connectivity paths belong to the same brain region set. Thus, a set of brain region set divisions can be obtained under each threshold condition. As the threshold gradually increases, the number of retained connectivity relationships decreases, and the division of brain region sets gradually changes from a relatively aggregated overall structure to a more refined structure. The brain region sets obtained under all threshold conditions are recorded sequentially, forming a multi-threshold brain region set sequence.

[0061] After obtaining the multi-threshold brain region set sequence, it is necessary to systematically record the assignment of each brain region under different threshold conditions. In practice, a recording sequence can be established for each brain region, with a length equal to the number of thresholds. Each position records the identifier information of the brain region set to which it belongs under the corresponding threshold condition. This "identifier information" is only used to distinguish different brain region sets under the same threshold condition, and does not require that the identifiers across different threshold conditions have consistent meanings. For example, under one threshold condition, brain region A and brain region B may be assigned to set one, while under another threshold condition, they may be assigned to set two. However, there is no direct correspondence between the set numbers; they are only used for distinction under the current threshold.

[0062] Building upon this, a "stable co-occurrence coefficient" is introduced to characterize the stability of two brain regions co-occurring within the same set of brain regions under different threshold conditions. The "stable co-occurrence coefficient" is defined as follows: for any two brain regions, determine whether they belong to the same set of brain regions under all threshold conditions, count the number of times they simultaneously belong to the same set, and divide this number by the total number of thresholds to obtain a value between zero and one. The closer this value is to one, the more stable the connection relationship is, indicating that the two brain regions consistently remain in the same structure under different screening intensities; the closer the value is to zero, the less likely they are to co-occur in the same set under different threshold conditions, indicating an unstable relationship. For example, if there are five threshold conditions, and brain regions A and B are classified into the same set of brain regions under four of these conditions, then their stable co-occurrence coefficient is four divided by five, i.e., 0.8. By calculating the above coefficient pairwise for all brain regions, a complete set of stable co-occurrence relationships between brain regions can be obtained.

[0063] After obtaining a set of stable co-occurrence relationships among brain regions, it is necessary to further aggregate these regions based on their stability. Specifically, a stability threshold is first set, for example, 0.7 or 0.8. Brain region pairs with a stable co-occurrence coefficient greater than or equal to this threshold are considered to have stable connections. Based on these stable connections, all brain regions are aggregated, such that any two brain regions that are interconnected through one or more stable connections are grouped into the same candidate brain region group. Multiple candidate brain region groups can be obtained in this way. Subsequently, to further screen for groups with more reliable structures, a group stability index needs to be calculated for each candidate brain region group. The "group stability index" is defined as the average of the stable co-occurrence coefficients between all pairs of brain regions within the group. Specifically, for a candidate group, the stable co-occurrence coefficients between all pairs of brain regions within the group are extracted, and these values ​​are averaged to obtain the group's stability index. If this index is greater than a preset condition, the group is considered structurally stable and retained as a target brain region group; otherwise, it is discarded.

[0064] After identifying the target brain region groups, functional characteristic parameters need to be extracted for each group. First, for each target brain region group, the correlation values ​​between all brain regions within that group are extracted from the original brain region correlation set, and these values ​​are averaged to obtain the information interaction capability parameter of the group. This parameter reflects the overall information transmission level within the group. Second, the degree of structural concentration within the group needs to be calculated. The "degree of structural concentration" refers to the uniformity of the correlation distribution between brain regions within the group, which can be achieved by calculating the deviation of each correlation from the average value. For example, first, the average correlation value within the group is obtained, then the difference between each correlation and the average value is calculated, and the absolute value of these differences is processed and averaged to obtain the dispersion value. The smaller the value, the more concentrated the structure. Third, the connection strength between different target brain region groups needs to be calculated. In practice, for any two target brain region groups, the correlation values ​​between all brain regions in one group and all brain regions in the other group are extracted, and these values ​​are averaged to obtain the connection strength parameter between the two groups.

[0065] By employing the above method, parameters related to information interaction ability, structural concentration, and inter-group connection strength are extracted from each target brain region group and arranged in a uniform order to form a complete set of feature parameters. This set of feature parameters is obtained using completely consistent calculation methods before and after surgery, thus ensuring that subsequent differential analyses have clear correspondences and a reliable basis for comparison.

[0066] Step S104: Calculate the difference between the preoperative feature parameter set and the postoperative feature parameter set to form a difference parameter set, and determine the weights of each parameter based on the dispersion of the pre-established health reference data. Perform weighted fusion on the difference parameter set to obtain a quantitative index of working memory impairment.

[0067] Step S104 involves quantifying the differences between the preoperative and postoperative feature parameter sets obtained in step S103 and forming a unified evaluation index. In practice, the correspondence between the preoperative and postoperative feature parameter sets must first be confirmed. Since step S103 ensured complete consistency in target brain region group division, parameter types, and arrangement order, each parameter can be matched one-to-one according to the same index order. For each target brain region group, parameters related to information interaction ability, structural concentration, and inter-group connection strength are established, creating a one-to-one correspondence between preoperative and postoperative parameters.

[0068] After establishing the correspondence, the difference between each pair of corresponding parameters is calculated to reflect the postoperative changes relative to the preoperative changes. In practice, the difference can be obtained by subtracting the preoperative parameter value from the postoperative parameter value. The same calculation is performed for each parameter to obtain a set of difference parameters. If there are parameters with different dimensions or significantly different value ranges, a uniform normalization process can be performed on each parameter before or after the difference calculation. For example, linear scaling can be used to convert each parameter to a uniform range to avoid excessive influence of a particular parameter's large value range on the subsequent fusion results. All parameter differences are arranged in their original order to form a set of difference parameters. Each element in this set corresponds to a specific group or the degree of change in the relationship between groups.

[0069] To further improve the discriminative power of differential parameters in assessment, weights need to be introduced for each parameter. These weights are derived from pre-established health reference data. In practice, a certain number of healthy subjects can be selected, and their characteristic parameters can be obtained under the same task conditions and data processing procedures. The values ​​of each parameter category within the healthy population can be statistically analyzed to obtain the distribution of that parameter in a healthy state. Based on this, the dispersion of each parameter category is calculated to reflect the range of fluctuation among healthy individuals. The dispersion can be obtained by statistically analyzing the magnitude of change of the parameter within the healthy population, for example, by calculating the deviation of the parameter's average across all healthy samples from individual samples, thus obtaining a value characterizing the parameter's stability. Generally, the smaller the dispersion, the more stable the parameter is among healthy individuals, and its changes are more likely to reflect actual functional changes; parameters with larger dispersion indicate greater fluctuations, and their changes have relatively lower reliability for damage assessment.

[0070] Based on the aforementioned degree of dispersion, a corresponding weight can be assigned to each parameter. In practice, the weight can be set according to the principle that the degree of dispersion is inversely proportional to the weight; that is, parameters with lower dispersion are assigned larger weights, and parameters with higher dispersion are assigned smaller weights. To ensure the consistency of all weights overall, the initial weights can be normalized so that the sum of all parameter weights is a preset constant, thus facilitating subsequent weighted fusion calculations.

[0071] After obtaining the set of differential parameters and their corresponding weights, a weighted fusion process is performed on the differential parameter set. Specifically, each differential parameter is multiplied by its corresponding weight to obtain a weighted parameter value. All weighted parameter values ​​are then summed, preferably using a summation method, to obtain a comprehensive value. This comprehensive value is the quantitative indicator of working memory impairment. This indicator reflects the overall degree of change in multiple functional dimensions after surgery compared to before surgery, and its magnitude directly characterizes the severity of working memory function impairment.

[0072] To further clarify the specific methods for obtaining quantitative indicators of working memory impairment, a concrete calculation process can be used as an example. Assume that the preoperative and postoperative feature parameter sets have been obtained in step S103, and the differences between the corresponding parameters are calculated in step S104, thus forming a set of difference parameters. For example, if a subject's difference parameters across the three feature dimensions are 0.20, -0.10, and 0.05, these difference parameters can be uniformly processed first. For instance, the original differences can be used directly, or their absolute values ​​can be taken as needed to characterize the magnitude of change, thereby obtaining a set of parameter values ​​that can be used for fusion.

[0073] Meanwhile, in the health reference data, the dispersion of the three feature dimensions mentioned above was statistically calculated, for example, the dispersion values ​​are 0.05, 0.20, and 0.10. Based on the principle that dispersion is inversely proportional to weight, the initial weights of each parameter can be set to the reciprocal of the dispersion, i.e., 20, 5, and 10 respectively. To ensure the consistency of the weights overall, the weights can be normalized. For example, the sum of all weights can be used as the normalization benchmark, resulting in normalized weights of 20 / (20+5+10), 5 / (20+5+10), and 10 / (20+5+10).

[0074] After obtaining the normalized weights, each difference parameter is multiplied by its corresponding weight to obtain a weighted parameter value. All weighted parameter values ​​are then summed to obtain a comprehensive numerical value. For example, multiplying the difference parameters 0.20, 0.10, and 0.05 by their respective weights and then summing the results yields the final quantitative index of working memory impairment. This quantitative index is a single numerical value, and its magnitude directly reflects the degree of comprehensive change in multiple functional dimensions after surgery compared to before surgery; a larger value indicates a more significant degree of working memory impairment.

[0075] Furthermore, the difference between the preoperative and postoperative feature parameter sets forms a difference parameter set. Weights are determined based on the dispersion of each parameter in pre-established health reference data. This difference parameter set is then weighted and fused to obtain a quantitative index of working memory impairment, including: Based on the preoperative and postoperative feature parameter sets, a one-to-one matching is performed according to the parameter type and its corresponding target brain region group relationship. For each pair of corresponding parameters, the difference between the postoperative value and the preoperative value is calculated, and the obtained difference is normalized according to a preset uniform scale to obtain a standardized difference parameter sequence. For the standardized differential parameter sequence, based on the pre-established health reference data, the value set of the corresponding parameter in the healthy subject group is extracted respectively, and the mean absolute deviation of the mean value set of each parameter from the mean value set of each sample is calculated to obtain the dispersion index sequence of the corresponding parameter. Based on the dispersion index sequence, the initial weight values ​​of each parameter are calculated according to the principle that the dispersion is inversely proportional to the weight. All initial weight values ​​are then normalized so that the sum of all weights equals a preset constant, thus obtaining a normalized weight sequence. The standardized difference parameter sequence and the normalized weight sequence are multiplied item by item, and all product results are accumulated to obtain a comprehensive change value. Based on a preset mapping rule, the comprehensive change value is converted into a quantitative indicator of working memory impairment.

[0076] In this embodiment, the steps involve calculating the differences between the preoperative and postoperative characteristic parameter sets based on the already obtained preoperative and postoperative characteristic parameter sets, and objectively assigning weights to the importance of each parameter in conjunction with health reference data, ultimately forming a unified quantitative index for working memory impairment.

[0077] First, parameter matching is required between the preoperative and postoperative feature parameter sets. Since the previous steps ensured complete consistency in parameter type, corresponding brain region groups, and arrangement order, a correspondence can be established one-to-one according to the same index order. For each parameter, such as the information interaction ability parameter, structural concentration parameter, or inter-group connection strength parameter of a target brain region group, there is a one-to-one correspondence between preoperative and postoperative values. Based on this, the difference between the postoperative and preoperative values ​​is calculated for each pair of corresponding parameters, thus obtaining the original difference value reflecting the degree of change in that parameter.

[0078] Since the numerical ranges of different types of parameters may vary, it is necessary to standardize the scaling of these differences to avoid any one type of parameter unduly amplifying the results. In practice, this can be achieved by first calculating the maximum absolute value of the differences among all parameters and using this value as a normalization benchmark. Each difference value is then divided by this benchmark, thus compressing all the differences into a uniform range, such as between -1 and 1. The resulting parameter sequence is the standardized difference parameter sequence, where each element is a dimensionless value, enabling fair comparisons in subsequent weighted calculations.

[0079] Subsequently, the dispersion of each parameter needs to be determined based on health reference data. Health reference data refers to the set of corresponding characteristic parameters obtained from a number of healthy subjects under the same task conditions and data processing procedures. For each type of parameter in the standardized difference parameter sequence, the set of values ​​for that parameter across all healthy subjects is extracted from the health reference data. For example, for the information interaction ability parameter of a certain group, a set of values ​​corresponding to multiple healthy individuals can be obtained. Based on this, the average of this set is first calculated, then the absolute difference between each sample value and the average is calculated, and the average of all absolute differences is taken to obtain the mean absolute deviation of the parameter. This mean absolute deviation is defined as the dispersion index of the parameter. This index reflects the range of fluctuation of the parameter among healthy individuals; the smaller the value, the more stable the parameter is in a healthy state.

[0080] After obtaining the dispersion indices of all parameters, the weights of each parameter need to be determined based on the dispersion. In practice, the initial weight of each parameter can be set to the reciprocal of its dispersion. For example, if the dispersion of a parameter is 0.05, its initial weight is 1 divided by 0.05, which is 20; if the dispersion of another parameter is 0.20, its initial weight is 5. This method allows parameters that are more stable in a healthy state to receive greater weights. To ensure the consistency of the weights overall, all initial weights need to be normalized. Specifically, all initial weights are summed as a normalization factor, and then each initial weight is divided by this sum to obtain a normalized weight sequence, such that the sum of all weights equals a preset constant, for example, 1.

[0081] After obtaining the standardized difference parameter sequence and the normalized weight sequence, a term-by-term product operation is performed on them, that is, each standardized difference parameter is multiplied by its corresponding weight to obtain a weighted parameter value. Then, all weighted parameter values ​​are summed to obtain a comprehensive change value. This comprehensive change value is a single numerical value that reflects the degree of comprehensive change in multiple functional characteristic dimensions after surgery compared to before surgery.

[0082] To ensure the comprehensive change value has clear clinical interpretable significance, it needs to be further converted into a quantitative indicator of working memory impairment. In practice, a mapping rule can be pre-defined, for example, linearly mapping the comprehensive change value to a pre-defined scoring interval based on the distribution range of the comprehensive change value in healthy reference data. For instance, the average comprehensive change value of healthy individuals can be used as a benchmark, with larger changes corresponding to higher levels of impairment, thus converting the comprehensive change value into a quantitative indicator with clear dimensions. Through this process, a unique quantitative indicator of working memory impairment can be obtained. This indicator reflects overall functional changes and has a uniform numerical scale, making it directly usable for subsequent assessment of the degree of impairment.

[0083] Step S105: Obtain the degree of damage score based on the quantitative index of working memory impairment, and determine the differences in connectivity distribution and information interaction ability of each brain region relative to different target brain region groups before and after surgery. Determine the contribution value of the changes in the brain region based on the differences in connectivity distribution and information interaction ability, and generate the localization result of the damaged brain region based on the contribution value of the changes in each brain region.

[0084] Step S105, based on the quantitative indicators of working memory impairment obtained in step S104, expresses the overall degree of impairment and further realizes the spatial localization of the damaged brain regions. Specifically, the degree of impairment score is first determined based on the quantitative indicators of working memory impairment obtained in step S104. A mapping relationship between the quantitative indicators and the scores can be established in advance. For example, based on the distribution range of the quantitative indicators in health reference data, the quantitative indicators can be divided into several intervals, and a scoring range can be set for each interval. Alternatively, a linear mapping method can be used to convert the quantitative indicators proportionally into a preset scoring scale. For example, the minimum and maximum values ​​of the quantitative indicators can be mapped to the upper and lower limits of the scoring range, respectively, thus obtaining a continuously changing degree of impairment score. Through these methods, the quantitative indicators can be converted into scoring results that are easy to understand and compare clinically.

[0085] After obtaining the overall score, further analysis of local changes is performed on each brain region. Specifically, for any given brain region, the connectivity between that region and different target brain region groups needs to be determined based on both the preoperative and postoperative brain region connectivity sets. Taking a specific brain region as an example, firstly, the connectivity values ​​between that region and all brain regions belonging to each target brain region group are extracted from the preoperative data, and then categorized according to the groups. For example, the average connectivity between that brain region and all brain regions within a target brain region group is calculated as the preoperative connectivity strength of that brain region relative to that group. This process is repeated for all target brain region groups to obtain the preoperative connectivity distribution of that brain region relative to each group. The same method is used to process the postoperative data to obtain the postoperative connectivity distribution of that brain region relative to each group.

[0086] After obtaining the preoperative and postoperative connectivity distributions, the two can be compared to determine the differences in connectivity distribution. In practice, for the connectivity strength of a brain region in a certain target brain region group, the difference between the postoperative value and the preoperative value is calculated, thereby obtaining the degree of connectivity change of the brain region in that group; repeating this process for all groups yields a set of connectivity distribution differences of the brain region in different target brain region groups, which reflects the changes in connectivity relationships of the brain region between different functional groups.

[0087] Simultaneously, it is necessary to determine the differences in information interaction capabilities within this brain region. Information interaction capability can be measured based on the overall correlation between this brain region and other brain regions. Specifically, the correlation scores between this brain region and all other brain regions can be extracted, and these correlation scores can be aggregated, for example, by calculating an average or weighted average, to obtain the preoperative information interaction capability of this brain region. The same method can be applied to postoperative data to obtain the postoperative information interaction capability. Subsequently, the postoperative information interaction capability is compared with the preoperative information interaction capability, and the difference between the two is calculated to obtain the difference in information interaction capability of this brain region.

[0088] After obtaining the differences in connectivity distribution and information interaction ability, these two factors need to be combined to determine the contribution value of the brain region's changes. Specifically, the differences in connectivity distribution across various target brain region groups can be summarized, for example, by averaging or summing the absolute values ​​to obtain a value reflecting the overall degree of change in connectivity distribution. This value is then combined with the difference in information interaction ability of the brain region, preferably using a product method, to obtain the contribution value of the brain region's changes. This contribution value considers both the changes in connectivity patterns across different groups and the changes in its overall information interaction ability, thus comprehensively reflecting the extent to which the brain region plays a role in postoperative functional changes.

[0089] Repeating the above calculation process for all brain regions yields a set of change contribution values ​​for each region. Finally, the localization results of the damaged brain regions are generated based on the change contribution values ​​of each region. In practice, all brain regions can be sorted according to the magnitude of their change contribution values, with regions having larger change contribution values ​​being prioritized as damaged regions; alternatively, regions with change contribution values ​​exceeding a preset threshold can be selected as damaged areas and marked in standard brain space to form intuitive localization results.

[0090] To further illustrate the specific calculation method of the change contribution value and its application in the localization of damaged brain regions, a concrete example is provided below. Assume that several target brain region groups have been obtained in the preceding steps, and the preoperative and postoperative brain region connectivity has been calculated. Taking a certain brain region as an example, suppose that this brain region had connectivity strength values ​​of 0.60, 0.45, and 0.30 relative to the three target brain region groups before surgery, and the corresponding connectivity strength values ​​after surgery are 0.40, 0.50, and 0.20, respectively. Then, the differences in connectivity distribution of this brain region across the three groups can be obtained by calculating the differences between the postoperative and preoperative values, which are -0.20, 0.05, and -0.10, respectively. To reflect the overall degree of change in connectivity distribution, the absolute values ​​of the above differences can be taken and averaged, for example, (0.20 + 0.05 + 0.10) divided by 3, thus obtaining the degree of change in connectivity distribution of this brain region as approximately 0.12.

[0091] Meanwhile, regarding the information interaction ability of this brain region, assuming that the preoperative average value of its correlation with other brain regions was 0.50, while the postoperative value was 0.35, the difference in information interaction ability is the postoperative value minus the preoperative value, i.e., -0.15. For subsequent calculations, the absolute value of this difference can be taken, resulting in 0.15. Subsequently, the degree of change in the connectivity distribution of this brain region is combined with the degree of change in information interaction ability, for example, by multiplying by 0.12 to calculate the contribution value of the change, i.e., 0.12 multiplied by 0.15, resulting in a contribution value of 0.018 for this brain region.

[0092] Repeating the above calculation process for all brain regions yields a set of change contribution values ​​corresponding to each region. For example, if the change contribution value for another brain region is 0.005, while that for the third brain region is 0.020, it indicates that the functional change in the third brain region is the most significant. Based on this result, all brain regions can be sorted from largest to smallest change contribution value, and the brain regions with the highest change contribution values ​​can be selected as the main damaged brain regions; alternatively, a preset threshold can be set, such as selecting brain regions with change contribution values ​​greater than 0.010 as damaged areas. Subsequently, the selected brain regions are marked in standard brain space, for example, by highlighting the corresponding brain region locations on a brain structure image, thus forming an intuitive result for locating the damaged brain regions.

[0093] Furthermore, the method involves obtaining a damage severity score based on working memory impairment quantification indicators, determining the differences in connectivity distribution and information interaction capabilities of each brain region relative to different target brain region groups before and after surgery, determining the contribution value of changes in brain regions based on these differences, and generating the localization results of damaged brain regions based on the contribution values ​​of changes in each brain region, including: Based on the quantitative index of working memory impairment, a piecewise continuous mapping function is constructed according to the statistical distribution interval of the comprehensive change value in the pre-established health reference data. The quantitative index of working memory impairment is substituted into the mapping function to obtain the degree of impairment score. The mapping function is determined by the mean and upper and lower deviation range of the comprehensive change value in the healthy subject group. For each brain region, based on the preoperative brain region correlation set and the postoperative brain region correlation set, the correlation between the brain region and all brain regions in each target brain region group is extracted and the average correlation of the group is calculated. The connection distribution vector of the brain region relative to each target brain region group is obtained before and after surgery. The connection distribution difference vector is constructed based on the difference between the corresponding elements of the postoperative connection distribution vector and the preoperative connection distribution vector. Based on the connectivity distribution difference vector, the absolute values ​​of each difference value are processed and summed to obtain the connectivity distribution change amplitude parameter. At the same time, the information interaction ability is calculated based on the average value of the correlation between this brain region and other brain regions in the whole brain before and after surgery. The information interaction ability difference parameter is obtained by subtracting the preoperative value from the postoperative value. The change contribution value of the brain region is obtained by multiplying the parameter of the change amplitude of the connection distribution with the parameter of the difference in information interaction ability. A brain region contribution sequence is constructed based on the change contribution value. All brain regions are sorted in descending order according to the change contribution value. Brain regions with change contribution values ​​higher than a preset threshold are selected as damaged brain regions, thereby generating the localization result of the damaged brain region.

[0094] In this embodiment, the steps are based on the obtained quantitative indicators of working memory impairment and the aforementioned brain region correlation set and target brain region group division results, to quantitatively express the overall degree of impairment, and further to accurately locate the local functional changes in each brain region.

[0095] First, a score for the degree of impairment needs to be obtained based on quantitative indicators of working memory impairment. To this end, a statistical distribution of the comprehensive change value needs to be established beforehand based on healthy reference data. In practice, a set of comprehensive change values ​​can be obtained from several healthy subjects following the same steps as described in this method, and the average value and upper and lower deviation ranges of this set can be calculated. The upper and lower deviation ranges can be obtained by statistically analyzing the differences between each sample value and the average value. For example, the intervals formed by adding a certain multiple of the deviation to the average value and subtracting a certain multiple of the deviation from the average value can be determined to obtain the range of change under healthy conditions. Based on this, a piecewise continuous mapping function is constructed. A piecewise continuous mapping function refers to a function that uses different linear relationships within different numerical intervals but maintains continuity at the interval connections. For example, when the comprehensive change value is close to the healthy average value, the score changes relatively slowly; when the comprehensive change value deviates significantly from the healthy range, the score changes more rapidly. In practice, the healthy average value can be used as an intermediate benchmark, corresponding to the median score, and the slope of different intervals can be determined based on the upper and lower deviation ranges, thus forming a continuously changing mapping curve. By substituting the quantitative indicators of working memory impairment of the subject to be evaluated into the mapping function, the corresponding impairment score can be obtained. This score is a single value used to represent the overall degree of impairment.

[0096] After obtaining the overall score, a local analysis of each brain region is required. Specifically, for any given brain region, the correlation scores between that region and all brain regions within each target brain region group are first extracted based on the preoperative brain region correlation score set, and then summarized according to the groups. For example, for a target brain region group, the average correlation score between that brain region and all brain regions within that group is calculated as the preoperative connectivity strength of that brain region relative to that group. This process is repeated for all target brain region groups, resulting in a vector composed of multiple values, where each element corresponds to the connectivity strength between that brain region and a target brain region group. This vector is defined as the preoperative connectivity distribution vector of that brain region. The same method is used to process the postoperative brain region correlation score set to obtain the postoperative connectivity distribution vector of that brain region.

[0097] After obtaining the preoperative and postoperative connectivity distribution vectors, element-wise calculations are performed on both vectors to obtain the connectivity distribution difference vector. Specifically, for each corresponding element, the postoperative value is subtracted from the preoperative value to obtain the change in connectivity of that brain region within the corresponding target brain region group. To reflect the overall magnitude of the change, the absolute values ​​of each element in the difference vector are taken and summed to obtain a single numerical value, which is defined as the connectivity distribution change magnitude parameter. For example, if the differences for a brain region in the three groups are -0.20, 0.05, and -0.10, the absolute values ​​are 0.20, 0.05, and 0.10, respectively. Summing these values ​​yields 0.35, which represents the connectivity distribution change magnitude.

[0098] Simultaneously, it is necessary to calculate the difference in information interaction ability of this brain region. In this invention, information interaction ability is defined as the average level of the correlation between this brain region and all other brain regions. Specifically, the correlation values ​​between this brain region and all other brain regions are extracted from the preoperative brain region correlation set, and their average value is calculated as the preoperative information interaction ability. Similarly, the corresponding correlation values ​​are extracted from the postoperative brain region correlation set, and their average value is calculated as the postoperative information interaction ability. Subsequently, the preoperative information interaction ability is subtracted from the postoperative information interaction ability to obtain the information interaction ability difference parameter. This parameter can be positive or negative, and its absolute value reflects the degree of change.

[0099] After obtaining the parameters for the magnitude of change in connectivity distribution and the difference in information interaction ability, the two are combined to obtain the contribution value of the change. Specifically, the absolute values ​​of the magnitude of change in connectivity distribution and the difference in information interaction ability are multiplied to obtain the contribution value of the brain region. For example, if the magnitude of change in connectivity distribution in a brain region is 0.35 and the difference in information interaction ability is -0.10, then its contribution value is 0.35 multiplied by 0.10, which is 0.035. This value is used to comprehensively characterize the degree of change in the brain region across both connectivity patterns and overall information interaction ability.

[0100] Repeating the above calculation process for all brain regions yields a set of change contribution values ​​corresponding to each region. Arranging these values ​​in order of brain region numbering forms a brain region contribution sequence. Subsequently, the values ​​in this sequence are sorted, for example, from largest to smallest, to identify brain regions with significant changes. Simultaneously, a preset threshold can be set, such as selecting a quantile value based on the distribution of change contribution values ​​across all brain regions, to identify brain regions with change contribution values ​​exceeding this threshold as damaged brain regions. Finally, these damaged brain regions are marked in standard brain space, for example, by highlighting the corresponding areas in a 3D brain map, thereby generating intuitive results for locating the damaged brain regions.

[0101] In the above embodiments, a method for assessing postoperative working memory impairment based on functional magnetic resonance imaging (fMRI) is provided. Correspondingly, this application also provides a device for assessing postoperative working memory impairment based on fMRI. Since this embodiment, i.e., the second embodiment, is basically similar to the method embodiment, it is described simply; relevant details can be found in the description of the method embodiment. The method embodiments described below are merely illustrative.

[0102] The second embodiment of this application provides a postoperative working memory impairment assessment device based on functional magnetic resonance imaging, comprising: The acquisition unit 201 is used to acquire functional magnetic resonance imaging data and preprocess the data during the execution of a preset working memory task in the same subject before and after the operation to obtain a preoperative time-series signal set and a postoperative time-series signal set. The division unit 202 is used to divide the preoperative temporal signal set and the postoperative temporal signal set into multiple brain regions according to the preset brain region division scheme, and determine the brain region correlation degree based on the correlation degree of signal changes between brain regions, so as to obtain the preoperative brain region correlation degree set and the postoperative brain region correlation degree set respectively. The screening unit 203 is used to screen the preoperative brain region association set and the postoperative brain region association set under multiple threshold conditions to form a brain region set. The target brain region group is determined according to the consistency of the set to which each brain region belongs under different threshold conditions. Feature parameters that characterize the information interaction ability, structural concentration and inter-group connection strength of each target brain region group are extracted to obtain the preoperative feature parameter set and the postoperative feature parameter set. The determination unit 204 is used to calculate the difference between the preoperative feature parameter set and the postoperative feature parameter set to form a difference parameter set, and to determine the weights based on the dispersion of each parameter in the pre-established health reference data, and to perform weighted fusion on the difference parameter set to obtain a quantitative index of working memory impairment. The generation unit 205 is used to obtain a score of the degree of damage based on the quantitative index of working memory impairment, and to determine the differences in connectivity distribution and information interaction ability of each brain region relative to different target brain region groups before and after surgery. Based on the differences in connectivity distribution and information interaction ability, the change contribution value of the brain region is determined, and the localization result of the damaged brain region is generated based on the change contribution value of each brain region.

[0103] A third embodiment of this application provides an electronic device, the electronic device comprising: processor; The memory is used to store a program, which, when read and executed by the processor, executes a postoperative working memory impairment assessment method based on functional magnetic resonance imaging provided in the first embodiment of this application.

[0104] The fourth embodiment of this application provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it performs a postoperative working memory impairment assessment method based on functional magnetic resonance imaging provided in the first embodiment of this application.

[0105] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A method for postoperative working memory impairment assessment based on functional magnetic resonance imaging, characterized in that, include: In the same subject, functional magnetic resonance imaging data were collected and preprocessed during the execution of a pre-set working memory task before and after the operation to obtain the pre-operation time sequence signal set and the post-operation time sequence signal set. Based on the pre-defined brain region division scheme, the preoperative and postoperative temporal signal sets were divided into multiple brain regions, and the brain region correlation degree was determined based on the correlation degree of signal changes between brain regions, resulting in the preoperative brain region correlation degree set and the postoperative brain region correlation degree set, respectively. Brain region sets were formed by screening the preoperative and postoperative brain region association sets under multiple threshold conditions. Target brain region groups were determined based on the consistency of the sets to which each brain region belonged under different threshold conditions. Feature parameters characterizing the information interaction ability, structural concentration, and inter-group connection strength of each target brain region group were extracted to obtain the preoperative feature parameter set and the postoperative feature parameter set. The difference between the preoperative and postoperative characteristic parameter sets is calculated to form a differential parameter set. The weights of each parameter in the pre-established health reference data are determined, and the differential parameter set is weighted and fused to obtain a quantitative index of working memory impairment. The degree of damage was scored based on the quantitative indicators of working memory impairment. For each brain region, the differences in connectivity distribution and information interaction ability of the brain region relative to different target brain region groups were determined before and after the operation. The contribution value of the changes in the brain region was determined based on the differences in connectivity distribution and information interaction ability. The localization results of the damaged brain region were generated based on the contribution value of the changes in each brain region. 2.The postoperative working memory impairment evaluation method based on functional magnetic resonance imaging according to claim 1, wherein, The process involves filtering the preoperative and postoperative brain region association sets under multiple threshold conditions to form brain region sets. Target brain region groups are determined based on the consistency of the sets to which each brain region belongs under different threshold conditions. Feature parameters characterizing the information interaction ability, structural concentration, and inter-group connection strength of each target brain region group are extracted, resulting in a preoperative feature parameter set and a postoperative feature parameter set, including: Based on the brain region association set, multiple increasing threshold sequences are generated according to the preset association value range. Under each threshold condition, brain region association pairs that satisfy the association degree greater than or equal to the threshold are selected. Based on the selected brain region association pairs, brain region connectivity is constructed and the corresponding brain region set under the threshold is obtained, forming a multi-threshold brain region set sequence. For a multi-threshold brain region set sequence, the number of the brain region set to which each brain region belongs under each threshold condition is recorded. Based on the ratio of the number of times brain regions jointly belong to the same brain region set under different threshold conditions to the total number of thresholds, the stable co-occurrence coefficient between any two brain regions is calculated, forming a set of stable co-occurrence relationships between brain regions. Based on the set of stable co-occurrence relationships of brain regions, brain regions with stable co-occurrence coefficients greater than a preset stability threshold are aggregated to generate multiple candidate brain region groups. The stable co-occurrence coefficients of all brain region pairs within each candidate brain region group are averaged to obtain a group stability index. Candidate brain region groups whose group stability index meets preset conditions are selected as target brain region groups. For each target brain region group, the correlation between brain regions within the group is extracted based on the brain region correlation set, and the average value is calculated to obtain the information interaction capability parameter. The structural concentration parameter is calculated based on the dispersion of the correlation within the group, and the inter-group connection strength parameter is calculated based on the average value of the brain region correlation between different target brain region groups, thus forming the corresponding feature parameter set. 3.The postoperative working memory impairment evaluation method based on functional magnetic resonance imaging according to claim 1, wherein, The difference between the preoperative and postoperative characteristic parameter sets is used to form a difference parameter set. Weights are determined based on the dispersion of each parameter in pre-established health reference data. This difference parameter set is then weighted and fused to obtain a quantitative index of working memory impairment, including: Based on the preoperative and postoperative feature parameter sets, a one-to-one matching is performed according to the parameter type and its corresponding target brain region group relationship. For each pair of corresponding parameters, the difference between the postoperative value and the preoperative value is calculated, and the obtained difference is normalized according to a preset uniform scale to obtain a standardized difference parameter sequence. For the standardized differential parameter sequence, based on the pre-established health reference data, the value set of the corresponding parameter in the healthy subject group is extracted respectively, and the mean absolute deviation of the mean value set of each parameter from the mean value set of each sample is calculated to obtain the dispersion index sequence of the corresponding parameter. Based on the dispersion index sequence, the initial weight values ​​of each parameter are calculated according to the principle that the dispersion is inversely proportional to the weight. All initial weight values ​​are then normalized so that the sum of all weights equals a preset constant, thus obtaining a normalized weight sequence. The standardized difference parameter sequence and the normalized weight sequence are multiplied item by item, and all product results are accumulated to obtain a comprehensive change value. Based on a preset mapping rule, the comprehensive change value is converted into a quantitative indicator of working memory impairment. 4.The postoperative working memory impairment evaluation method based on functional magnetic resonance imaging according to claim 1, wherein, The method involves obtaining a damage severity score based on working memory impairment quantification indicators, determining the differences in connectivity distribution and information interaction capabilities of each brain region relative to different target brain region groups before and after surgery, determining the contribution value of changes in brain regions based on the differences in connectivity distribution and information interaction capabilities, and generating the localization results of damaged brain regions based on the contribution values ​​of changes in each brain region, including: Based on the quantitative index of working memory impairment, a piecewise continuous mapping function is constructed according to the statistical distribution interval of the comprehensive change value in the pre-established health reference data. The quantitative index of working memory impairment is substituted into the mapping function to obtain the degree of impairment score. The mapping function is determined by the mean and upper and lower deviation range of the comprehensive change value in the healthy subject group. For each brain region, based on the preoperative brain region correlation set and the postoperative brain region correlation set, the correlation between the brain region and all brain regions in each target brain region group is extracted and the average correlation of the group is calculated. The connection distribution vector of the brain region relative to each target brain region group is obtained before and after surgery. The connection distribution difference vector is constructed based on the difference between the corresponding elements of the postoperative connection distribution vector and the preoperative connection distribution vector. Based on the connectivity distribution difference vector, the absolute values ​​of each difference value are processed and summed to obtain the connectivity distribution change amplitude parameter. At the same time, the information interaction ability is calculated based on the average value of the correlation between this brain region and other brain regions in the whole brain before and after surgery. The information interaction ability difference parameter is obtained by subtracting the preoperative value from the postoperative value. The change contribution value of the brain region is obtained by multiplying the parameter of the change amplitude of the connection distribution with the parameter of the difference in information interaction ability. A brain region contribution sequence is constructed based on the change contribution value. All brain regions are sorted in descending order according to the change contribution value. Brain regions with change contribution values ​​higher than a preset threshold are selected as damaged brain regions, thereby generating the localization result of the damaged brain region.

5. The method for assessing postoperative working memory impairment based on functional magnetic resonance imaging according to claim 1, characterized in that, In the same subject, functional magnetic resonance imaging (fMRI) data were collected and preprocessed during the execution of a preset working memory task before and after surgery to obtain a preoperative time-series signal set and a postoperative time-series signal set, including: While acquiring functional magnetic resonance imaging (fMRI) data, the time series of stimulus presentation and behavioral response of subjects in working memory tasks were recorded. Based on the stimulus presentation time series, the acquired fMRI data were time-synchronized and labeled. The brain signal data corresponding to each stimulus presentation were aligned to obtain the original time-series signal sequence that corresponds one-to-one with the task events. Based on the original time-series signal sequence, the brain signal data at each time point is screened for task response consistency. Signal data corresponding to time periods in which the subjects did not make correct responses or whose reaction times exceeded the preset range are removed. The retained signal data are then reconstructed into a continuous time-series signal in chronological order to obtain the time-series signal sequence after task response correction. For the time-series signal sequences, a consistent preprocessing procedure was performed on the preoperative and postoperative data, including spatial registration of brain signal data at each time point to eliminate the influence of head movement of the subjects, and frequency range filtering of the registered signal data to retain signal components within the preset frequency band, thereby obtaining a preoperative time-series signal set and a postoperative time-series signal set with a unified time and spatial reference. 6.The postoperative working memory impairment evaluation method based on functional magnetic resonance imaging according to claim 1, wherein, According to a preset brain region division scheme, the preoperative and postoperative temporal signal sets are divided into multiple brain regions, and the brain region correlation is determined based on the correlation of signal changes between brain regions, resulting in a preoperative brain region correlation set and a postoperative brain region correlation set, including: Based on the preoperative and postoperative temporal signal sets, the whole brain is divided into multiple pre-defined brain regions according to a unified spatial coordinate system. The signal values ​​of all voxels in each brain region at each time point are weighted and summarized. The weighting coefficient is determined according to the distance between the spatial position of each voxel in the brain region and the center position of the brain region, thereby obtaining the representative temporal signal sequence corresponding to each brain region. For the representative temporal signal sequence of each brain region, time delay correction processing is performed. By gradually shifting the signal sequence within a preset time offset range and calculating the degree of matching with the original sequence, the time offset that makes the signal fluctuation characteristics most consistent is determined. Based on the time offset, the brain region temporal signal is corrected to obtain the time-aligned brain region temporal signal sequence. Based on the time-aligned brain region temporal signal sequence, the segmented correlation of signal changes between any two brain regions is calculated. The entire time series is divided into multiple continuous time periods, and the correlation between the signal sequences of two brain regions is calculated in each time period. Then, the correlation of each time period is weighted and averaged to obtain the brain region correlation value that reflects local temporal consistency, thus forming a brain region correlation set.

7. An apparatus for postoperative working memory impairment assessment based on functional magnetic resonance imaging, characterized by, include: The acquisition unit is used to acquire functional magnetic resonance imaging data and preprocess the data during the execution of a preset working memory task in the same subject before and after the operation to obtain the preoperative time-series signal set and the postoperative time-series signal set. The division unit is used to divide the preoperative and postoperative temporal signal sets into multiple brain regions according to the preset brain region division scheme, and to determine the brain region correlation degree based on the correlation degree of signal changes between brain regions, so as to obtain the preoperative brain region correlation degree set and the postoperative brain region correlation degree set respectively. The screening unit is used to screen the preoperative brain region association set and the postoperative brain region association set under multiple threshold conditions to form a brain region set. The target brain region group is determined according to the consistency of the set to which each brain region belongs under different threshold conditions. Feature parameters that characterize the information interaction ability, structural concentration and inter-group connection strength of each target brain region group are extracted to obtain the preoperative feature parameter set and the postoperative feature parameter set. The unit is used to calculate the difference between the preoperative feature parameter set and the postoperative feature parameter set to form a difference parameter set. The weights of each parameter in the pre-established health reference data are determined, and the difference parameter set is weighted and fused to obtain a quantitative index of working memory impairment. The generation unit is used to obtain a score of the degree of damage based on the quantitative index of working memory impairment, and to determine the differences in connectivity distribution and information interaction ability of each brain region relative to different target brain region groups before and after surgery. Based on the differences in connectivity distribution and information interaction ability, the change contribution value of the brain region is determined, and the localization result of the damaged brain region is generated based on the change contribution value of each brain region.

Images