Brain disease target positioning system and method based on functional magnetic resonance imaging

Abstract

The invention relates to the technical field of brain disease target positioning, in particular to a brain disease target positioning system and method based on functional magnetic resonance imaging. The method comprises the following steps: calculating a whole brain function connection map of an individual patient based on 4D rs-fMRI data, calculating a Z score of each voxel function connection strength in the whole brain function connection map of the individual patient, and screening abnormal voxels based on a preset threshold value; aggregating the abnormal voxels through a clustering algorithm so as to determine a cluster; the therapeutic effect scores of the multiple clusters can be predicted through the machine learning model, the clusters and the candidate therapeutic targets can be screened based on the therapeutic effect scores, the positions of the candidate targets can be accurately determined, and multiple selections of the candidate targets can be provided; the target spot is converted into an accurate target spot suitable for an individual through a nonlinear registration algorithm; and determining a target positioning state according to the deviation between the accurate target and the actual treatment target, and adjusting corresponding parameters based on the target positioning state. The invention provides a plurality of targets and effectively determines the positions of the targets.

Description

Technical Field

[0001] This invention relates to the field of brain disease target localization technology, and in particular to a brain disease target localization system and method based on functional magnetic resonance imaging. Background Technology

[0002] Alzheimer's disease (AD) is a neurodegenerative disease characterized by β-amyloid deposition and neurofibrillary tangles. Mild cognitive impairment (MCI) is its early stage, and its pathological process is accompanied by impaired brain functional connectivity. Reduced functional connectivity (FC), measured by resting-state functional magnetic resonance imaging (rs-fMRI), indicates synaptic dysfunction and can provide evidence of specific neuronal networks involved in cognitive processes. Therefore, non-invasive brain stimulation target localization based on fMRI is needed to modulate FC in MCI patients to enhance cognitive abilities. While rs-fMRI confirms impaired brain functional connectivity in MCI, existing transcranial electrical stimulation (TCS) techniques, although capable of modulating neuronal activity, still have limitations. Firstly, target selection is often limited, with current methods targeting fixed brain regions without considering individual functional connectivity characteristics, leading to significant differences in efficacy. Secondly, data-driven approaches are insufficient; current techniques do not fully utilize past modulation data to optimize stimulation strategies, resulting in a relatively low level of intelligence.

[0003] Chinese Patent Publication No. CN109480841A discloses a method for precise localization and correction of abnormal brain regions based on functional magnetic resonance imaging. The method of the present invention includes using resting-state functional magnetic resonance imaging technology to accurately detect and calculate the activity correlation between any brain regions in the whole brain of the subject, that is, the whole brain functional connectivity.

[0004] It is evident that existing technologies have the following problems: the target is singular and the target location cannot be effectively determined. Summary of the Invention

[0005] Therefore, the present invention provides a brain disease target localization system and method based on functional magnetic resonance imaging, which overcomes the problems of single target and inability to effectively determine the target location in the prior art.

[0006] To achieve the above objectives, the present invention provides a brain disease target localization system based on functional magnetic resonance imaging, comprising: The acquisition unit is used to acquire 4D rs-fMRI data in real time; The computing unit, which is connected to the acquisition unit, is used to calculate the correlation coefficient between voxels based on 4D rs-fMRI data, starting from a preset seed point region covering the key networks of the whole brain, so as to generate a whole brain functional connectivity map of the individual patient. A screening unit, connected to the calculation unit, is used to calculate the Z-score of the functional connectivity strength of each voxel in the whole-brain functional connectivity map using statistical test methods, and to screen abnormal voxels based on a preset threshold, wherein the Z-score is the standard deviation multiple of the individual voxel's functional connectivity strength from the average value of the normal group. A determining unit, connected to the screening unit, is used to extract functional features from multiple clusters formed by aggregating the abnormal voxels using a clustering algorithm, predict the efficacy scores of multiple clusters based on the functional features using a machine learning model trained with historical patient data, and determine candidate treatment targets for the clusters selected based on the efficacy scores. The historical patient data includes rs-fMRI data, treatment target coordinates, and clinical efficacy scores. A target coordinate optimization unit, which is connected to the determining unit, is used to transform multiple candidate therapeutic targets into multiple precise targets based on individual anatomical structures using a nonlinear registration algorithm. An analysis unit, connected to the target coordinate optimization unit, is used to calculate the deviations between multiple precise target points and the patient's actual treatment target points, determine the target positioning status based on the deviations, adjust the preset threshold based on the positioning status, and adjust the iteration number of the nonlinear registration algorithm based on the target positioning status after adjusting the preset threshold.

[0007] Furthermore, the analysis unit is also used to calculate the Euclidean distance between each of the precise target points and the actual treatment target points, and to determine the deviation based on the average of multiple Euclidean distances; the analysis unit is also used to calculate the variance of multiple historical deviations when the target point positioning status is unqualified; the analysis unit is also used to adjust the preset threshold based on the ratio of the preset variance to the variance when the variance is less than the preset variance; wherein, when the deviation is greater than the preset deviation, the target point positioning status is determined to be unqualified.

[0008] Furthermore, the analysis unit is also used to increase the preset threshold based on the ratio of the preset variance to the preset variance, and the increase in the preset threshold is proportional to the ratio.

[0009] Furthermore, the analysis unit is also used to increase the preset threshold based on the ratio of the preset number of abnormal voxels to the number of voxels, and the increase of the preset threshold is proportional to the ratio.

[0010] Furthermore, the target coordinate optimization unit also includes a registration module, which is used to display the registered individual brain image and the standard template image side by side; the analysis unit is also used to repeatedly adjust the preset threshold at least once if the target localization status is unqualified after adjusting the preset threshold, until the number of adjustments is less than the preset number and the target localization status is qualified, or the number of adjustments is equal to the preset number and the adjustment stops; the analysis unit is also used to calculate the difference between the deviation and the preset deviation if the target localization status is unqualified after stopping the adjustment; the analysis unit is also used to start the registration module if the difference is greater than the preset difference, and if the individual brain image generated by the registration module is not aligned with the standard template image, the iteration number of the nonlinear registration algorithm is adjusted based on the ratio of the difference to the preset difference; wherein, when the deviation is less than or equal to the preset deviation, the target localization status is determined to be qualified.

[0011] Furthermore, the analysis unit is also used to increase the number of iterations of the nonlinear registration algorithm based on the ratio of the difference to the preset difference, and the increase in the number of iterations of the nonlinear registration algorithm is proportional to the ratio.

[0012] Furthermore, the analysis unit is also used to plot a time-historical deviation curve based on the acquired historical deviation when the target localization status is unqualified after adjusting the number of iterations of the nonlinear registration algorithm; if the integral of the curve is greater than the preset integral, the amount of historical patient data is adjusted based on the ratio of the integral of the curve to the preset integral.

[0013] Furthermore, the analysis unit is also used to increase the amount of historical patient data based on the ratio of the integral of the curve to a preset integral, and the increase in the amount of historical patient data is proportional to the ratio.

[0014] Furthermore, the analysis unit is also used to calculate the average slope of the time-historical deviation curve when the target localization status is unqualified after adjusting the amount of historical patient data; the analysis unit is also used to reduce the verification cycle of the acquisition unit's acquisition device based on the ratio of the average value to the preset average value when the average value is greater than the preset average value, and the reduction in the verification cycle of the acquisition unit's acquisition device is proportional to the ratio.

[0015] To achieve the above objectives, the present invention provides a method for locating brain disease targets based on functional magnetic resonance imaging (fMRI), which is applied to the aforementioned system for locating brain disease targets based on fMRI, comprising: 4D rs-fMRI data are acquired in real time through the acquisition unit; The computing unit connected to the acquisition unit calculates the correlation coefficients between voxels based on 4D rs-fMRI data, starting from a preset seed point region covering key networks of the whole brain, in order to generate a whole brain functional connectivity map for the individual patient. The screening unit connected to the computing unit uses statistical testing methods to calculate the Z score of the functional connectivity strength of each voxel in the whole brain functional connectivity map, and filters abnormal voxels based on a preset threshold. The Z score is the standard deviation multiple of the functional connectivity strength of an individual voxel from the average value of the normal group. The determining unit connected to the screening unit extracts functional features from multiple clusters formed by aggregating the abnormal voxels using a clustering algorithm. Based on the functional features, a machine learning model trained using historical patient data predicts the efficacy scores of multiple clusters. Based on the efficacy scores, candidate treatment targets are determined for each cluster. The historical patient data includes rs-fMRI data, treatment target coordinates, and clinical efficacy scores. The target coordinate optimization unit connected to the determination unit uses a nonlinear registration algorithm to transform multiple candidate therapeutic targets into multiple precise targets based on individual anatomical structures. The analysis unit connected to the target coordinate optimization unit calculates the deviations between multiple precise target points and the patient's actual treatment target points, determines the target point positioning status based on the deviations, adjusts the preset threshold based on the positioning status, and adjusts the number of iterations of the nonlinear registration algorithm based on the target point positioning status after adjusting the preset threshold.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention calculates the whole-brain functional connectivity map of an individual patient based on 4D rs-fMRI data, calculates the Z-score of the functional connectivity strength of each voxel in the whole-brain functional connectivity map, and filters abnormal voxels based on a preset threshold, enabling more accurate identification of abnormal seed points in the brain; it aggregates abnormal voxels using a clustering algorithm to determine clusters, integrating them into multiple clusters based on the characteristics of each abnormal voxel, thus enabling more effective screening of candidate therapeutic targets; a machine learning model can predict the efficacy scores of multiple clusters, and clusters and candidate therapeutic targets are screened based on the efficacy scores, enabling more accurate determination of the location of candidate targets and providing multiple choices of candidate targets, thereby improving the effectiveness of the targets; a nonlinear registration algorithm is used to transform them into precise targets suitable for the individual; the deviation between the precise target and the actual therapeutic target is used to determine the target location status, and corresponding parameters are adjusted based on the target location status. This invention provides multiple targets and effectively determines their locations.

[0017] Furthermore, this invention determines whether to adjust the preset threshold based on the variance of multiple historical deviations. This allows for a more accurate determination of the reasons for unqualified target positioning based on the historical patterns of the deviations, and a more effective adjustment of the corresponding parameters, thereby making the target positioning more accurate.

[0018] Furthermore, the present invention adjusts the preset threshold based on the ratio of the preset variance to the preset threshold, which can adjust the preset threshold more accurately. It can obtain more candidate seed points by relaxing the criteria, thereby obtaining more choices in the screening process and further improving the target positioning accuracy.

[0019] Furthermore, the present invention adjusts the preset threshold based on the ratio of the preset number of abnormal voxels to the number of voxels. This allows for more accurate adjustment of the preset threshold during the adjustment process, thereby more effectively screening candidate seed points, increasing their number, providing more choices, and further improving the target location accuracy.

[0020] Furthermore, this invention determines whether to adjust the iteration number of the nonlinear registration algorithm based on the difference between the deviation and the preset deviation and the comparison results between the individual brain image generated by the registration module and the standard template image. This allows for more effective adjustment based on more accurate reasons, thereby enabling more accurate screening of candidate therapeutic targets and further improving the target localization accuracy.

[0021] Furthermore, this invention adjusts the number of iterations of the nonlinear registration algorithm based on the ratio of the difference to a preset difference, which can more accurately adjust the number of iterations of the nonlinear registration algorithm, thereby further improving the accuracy of screening candidate therapeutic targets and thus further improving the target localization accuracy.

[0022] Furthermore, the present invention determines whether to adjust the amount of historical patient data based on the integral of the time-history deviation curve, which can more effectively adjust the corresponding parameters based on more accurate reasons, thereby further improving the accuracy of the machine learning model and thus further improving the target localization accuracy.

[0023] Furthermore, the present invention adjusts the amount of historical patient data based on the ratio of the integral of the curve to the preset integral, which can more accurately adjust the amount of historical patient data, thereby making the machine learning model more effective and further improving the target localization accuracy.

[0024] Furthermore, this invention adjusts the verification cycle of the acquisition device in the acquisition unit based on the average slope of the time-history deviation curve, which can more timely verify when the acquisition device in the acquisition unit has a slight displacement or drift, thereby acquiring 4D rs-fMRI data more accurately and further improving the target localization accuracy. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the brain disease target localization system based on functional magnetic resonance imaging according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the steps of the brain disease target localization method based on functional magnetic resonance imaging in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the steps for determining the comparison result between the deviation and the preset deviation in an embodiment of the present invention. Figure 4 This is a flowchart illustrating the steps of determining the target location status based on adjusting a preset threshold, as described in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0027] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0028] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0029] Please see Figure 1 As shown, it is a schematic diagram of the brain disease target localization system based on functional magnetic resonance imaging according to an embodiment of the present invention.

[0030] The system includes an acquisition unit, a calculation unit, a filtering unit, a determination unit, a target coordinate optimization unit, and an analysis unit.

[0031] The acquisition unit is used to acquire 4D rs-fMRI data in real time; The computing unit is connected to the acquisition unit and is used to calculate the correlation coefficient between voxels based on 4D rs-fMRI data, starting from a preset seed point region covering the key networks of the whole brain, so as to generate a whole brain functional connectivity map for the individual patient. The screening unit is connected to the calculation unit and is used to calculate the Z score of the functional connectivity strength of each voxel in the whole brain functional connectivity map using statistical test methods, and to screen abnormal voxels based on a preset threshold. The Z score is the standard deviation multiple of the individual voxel's functional connectivity strength from the average value of the normal group. The determining unit is connected to the screening unit and is used to extract functional features from multiple clusters formed by aggregating the abnormal voxels using a clustering algorithm, predict the efficacy scores of multiple clusters based on the functional features using a machine learning model trained with historical patient data, and determine candidate treatment targets for the clusters selected based on the efficacy scores. The historical patient data includes rs-fMRI data, treatment target coordinates, and clinical efficacy scores. The target coordinate optimization unit is connected to the determination unit, and it is used to transform multiple candidate therapeutic targets into multiple precise targets based on individual anatomical structures using a nonlinear registration algorithm. The analysis unit is connected to the target coordinate optimization unit. It is used to calculate the deviation between multiple precise target points and the actual treatment target points of the patient, determine the target positioning status based on the deviation, adjust the preset threshold based on the positioning status, and adjust the iteration number of the nonlinear registration algorithm based on the target positioning status after adjusting the preset threshold.

[0032] Specifically, 4D rs-fMRI data refers to a single rs-fMRI scan that produces a four-dimensional (4D) data volume comprising three spatial dimensions (X, Y, Z) plus a temporal dimension. The entire brain is divided into thousands of tiny three-dimensional units called voxels. Each voxel generates a time-series signal during the scan, representing the changes in activity intensity of that small piece of brain tissue over several minutes. Functional connectivity is determined by the time-series signal; that is, if the signals from two different brain regions fluctuate synchronously over time (i.e., are highly correlated), it indicates that the two brain regions are functionally connected.

[0033] Specifically, the computational unit first starts from a pre-defined seed point region of the key whole-brain network (such as the posterior cingulate cortex), extracts the time-series signals of all voxels within that region, and averages them to obtain a representative seed point time series. Then, the unit calculates the Pearson correlation coefficient *r* between the time series of each voxel in the whole brain and the seed point time series, generating a preliminary functional connectivity map. To meet the normality requirement for subsequent statistical tests, a Fisher Z-transform is performed on each correlation coefficient *r* value in the map, ultimately yielding an individual patient-specific whole-brain functional connectivity map with Z-values ​​representing functional connectivity strength.

[0034] Specifically, the screening unit compares the patient's functional connectivity map with an existing standard database of healthy individuals. This database contains the mean (μ) and standard deviation (σ) of the corresponding functional connectivity Z-scores from a large number of healthy subjects. For each voxel in the map, the unit calculates its Z-score using the following formula: ; in, The patient's Z-score, This is the average Z-score connecting the normal population. This is the standard Z-score for the normal population. The Z-score represents the degree to which the connectivity strength at that location in the patient deviates from the average level of the normal population. After calculation, the unit filters out all significantly abnormal voxels according to a preset threshold; these voxels indicate brain regions with significantly weakened functional connectivity.

[0035] Specifically, for each treatment record in the historical patient database, the functional features of the stimulated target cluster are extracted from the patient's baseline fMRI data. This dataset is then labeled with the corresponding efficacy score for that treatment, which serves as the model's prediction target (label).

[0036] The training process uses regression models (such as random forest regression, gradient boosting regression tree GBRT, or neural networks) to learn the complex mapping relationship between these "target features" and "therapeutic outcomes." The trained model predicts the potential effects of stimulating a target based on its functional state. For clusters of defined units (such as clusters A, B, and C), the aforementioned functional features of each cluster are extracted. The features of each cluster are input into a pre-trained machine learning model, which outputs a predicted efficacy score for each cluster. The clusters are then ranked from highest to lowest based on the predicted efficacy scores, and the top 3 clusters are recommended to determine the final optimal individualized stimulation target. Among the top three clusters, the voxel coordinates with the most significant statistical anomalies, i.e., the voxels with the lowest functional connectivity Z-values, are selected as candidate therapeutic targets. Since there are individual differences in the size and shape of each patient's brain, these standard coordinates need to be transformed into the patient's individual brain space. The target coordinate optimization unit uses the patient's high-resolution T1 structural image and a nonlinear registration algorithm to reverse the standard spatial coordinates to the patient's individual space, obtaining a precise anatomical location coordinate visible on the patient's brain scan image. This process is existing technology and will not be described in detail here.

[0037] Please see Figure 2 The diagram shown is a flowchart of the steps of the brain disease target localization method based on functional magnetic resonance imaging in an embodiment of the present invention.

[0038] S1 acquires 4D rs-fMRI data in real time through the acquisition unit; S2, the computing unit connected to the acquisition unit calculates the correlation coefficient between each voxel based on 4D rs-fMRI data, starting from a preset seed point region covering the key networks of the whole brain, in order to generate a whole brain functional connectivity map of the individual patient. S3, the Z score of the functional connectivity strength of each voxel in the whole brain functional connectivity map is calculated by the screening unit connected to the computing unit using statistical test methods, and abnormal voxels are screened based on a preset threshold, wherein the Z score is the standard deviation multiple of the functional connectivity strength of an individual voxel from the average value of the normal group. S4, the determining unit connected to the screening unit extracts functional features from the multiple clusters formed by aggregating the abnormal voxels using a clustering algorithm, predicts the efficacy scores of the multiple clusters based on the functional features and the machine learning model trained using historical patient data, and determines candidate treatment targets for the clusters selected based on the efficacy scores. The historical patient data includes rs-fMRI data, treatment target coordinates and clinical efficacy scores. S5, the target coordinate optimization unit connected to the determining unit uses a nonlinear registration algorithm to transform multiple candidate therapeutic targets into multiple precise targets based on individual anatomical structures. S6, the analysis unit connected to the target coordinate optimization unit calculates the deviation between multiple precise target points and the patient's actual treatment target points, determines the target point positioning status based on the deviation, adjusts the preset threshold based on the positioning status, and adjusts the iteration number of the nonlinear registration algorithm based on the target point positioning status after adjusting the preset threshold.

[0039] Please see Figure 3 The diagram shows a flowchart illustrating the steps of determining the target location status based on a comparison between the deviation and a preset deviation, according to an embodiment of the present invention. The analysis unit in this embodiment is further configured to calculate the variance of multiple historical deviations when the target location status is unqualified; the analysis unit is also configured to adjust the preset threshold based on the ratio of the preset variance to the original variance when the variance is less than the preset variance; wherein, when the deviation is greater than the preset deviation, the target location status is determined to be unqualified.

[0040] Specifically, taking the detection of abnormal brain functional networks as an example, based on the hardware performance limits of the functional magnetic resonance imaging system (such as the signal-to-noise ratio and spatial resolution under 3T / 7T field strength) and the fault tolerance requirements of the actual scanning environment (such as head movement artifacts and physiological noise interference), and combined with several historical data obtained from the statistical analysis of historical cases, the corresponding preset or critical parameter values ​​are set to achieve a balance between the accuracy and robustness of target localization.

[0041] Specifically, if the preset deviation L0 between multiple precise target points and the patient's actual treatment target point is 5mm, the comparison process between the deviation L0 between the multiple precise target points and the patient's actual treatment target point and the preset deviation L0 is as follows: If the deviation L between multiple precise target points and the patient's actual treatment target point is less than or equal to the preset deviation L0, then the target point positioning status is determined to be qualified. If the deviation L between multiple precise target points and the patient's actual treatment target point is greater than the preset deviation L0, then the target point positioning status is determined to be unqualified.

[0042] Specifically, when the target location status is unqualified, multiple historical deviations are acquired, and the variance of these historical deviations is calculated. If the variance is less than a preset variance, it indicates that the deviation remains at a large value. Therefore, it is necessary to relax the criteria to obtain more candidate seed points, thereby obtaining more choices during the screening process and improving the target location accuracy. A preset threshold is adjusted based on the ratio of the preset variance to the preset variance, where the preset ratio P0 = 1.2. The comparison process between the preset ratio P0 and the preset ratio P is as follows: Since this invention primarily identifies brain regions with significantly reduced functional connectivity (such as the hippocampus and the middle of the cingulate gyrus), and since the connectivity within the default mode network in Alzheimer's patients exhibits a significant negative Z-score, the original preset threshold is set to -2.0, and voxels with Z-scores less than the preset threshold are selected and identified as abnormal voxels.

[0043] If the ratio P of the preset variance to the original variance is less than or equal to the preset ratio P0, the preset threshold will be adjusted to 0.9 times the original preset threshold. If the ratio P of the preset variance to the original variance is greater than the preset ratio P0, then the preset threshold will be adjusted to 0.85 times the original preset threshold.

[0044] Specifically, a preset threshold is adjusted based on the preset ratio of the number of abnormal voxels to the total number of voxels. Where the preset ratio of the preset number of abnormal voxels to the total number of voxels is Q0 = 1.15, the comparison process between the preset ratio Q0 and the preset ratio Q0 is as follows: If the ratio Q of the preset number of abnormal voxels to the number of voxels is less than or equal to the preset ratio Q0, the preset threshold will be adjusted to 0.92 times the preset threshold after variance adjustment. If the ratio Q of the preset number of abnormal voxels to the number of voxels is greater than the preset ratio Q0, the preset threshold will be adjusted to 0.83 times the preset threshold after variance adjustment.

[0045] Please see Figure 4 The diagram shown is a flowchart illustrating the steps of determining the target location status based on adjusting a preset threshold, according to an embodiment of the present invention.

[0046] Specifically, the target localization status is re-checked. If the target localization status is still unqualified, the preset threshold is adjusted at least once until the number of adjustments is less than the preset number and the target localization status is qualified, or the number of adjustments is equal to the preset number, at which point the adjustment stops. If the target localization status is still unqualified after the adjustment stops, the difference between the deviation and the preset deviation is calculated. If the difference is greater than the preset difference, the registration module is started. If the individual brain image generated by the registration module is not aligned with the standard template image, it indicates that the number of iterations of the nonlinear registration algorithm is insufficient, resulting in inaccurate determination of the precise target. The number of iterations of the nonlinear registration algorithm is then adjusted based on the ratio of the difference to the preset difference. The preset ratio R0 = 1.1. The comparison process based on the ratio R0 of the difference to the preset difference is as follows: If the ratio R of the difference to the preset difference is less than or equal to the preset ratio R0, the number of iterations of the nonlinear registration algorithm is adjusted to 1.7 times the original number of iterations, where the adjusted number of iterations is rounded up. If the ratio R of the difference to the preset difference is greater than the preset ratio R0, the number of iterations of the nonlinear registration algorithm will be adjusted to 2.9 times the original number of iterations, where the adjusted number of iterations is rounded up.

[0047] Specifically, the target localization status is re-examined after adjusting the number of iterations of the nonlinear registration algorithm. If the target localization status is unqualified, a time-historical deviation curve is plotted based on the acquired historical deviation. If the integral of the curve is greater than the preset integral, it indicates that the historical deviation has always been large, or that the limited quantity and diversity of historical patient data have resulted in poor generalization ability of the trained machine learning model. Furthermore, the model overlearns details and noise in the training data rather than underlying general patterns, thus perfectly fitting the historical data but having weak predictive ability for new data. Therefore, the quantity of historical patient data is adjusted based on the ratio of the curve integral to the preset integral. The preset ratio T0 = 1.7. The comparison process between the curve integral and the preset integral ratio T and the preset ratio T0 is as follows: If the ratio T of the integral of the curve to the preset integral is less than or equal to the preset ratio T0, the number of historical patient data will be adjusted to 1.9 times the original number, and the adjusted number of historical patient data will be rounded up. If the ratio T of the integral of the curve to the preset integral is greater than the preset ratio T0, the number of historical patient data will be adjusted to 2.7 times the original number, and the adjusted number of historical patient data will be rounded up.

[0048] Specifically, the target localization status is re-examined after adjusting the amount of historical patient data. If the target localization status is unqualified, the average slope of the time-historical deviation curve is calculated. If the average value is greater than the preset average value, it indicates that the acquisition equipment in the acquisition unit has experienced a slight displacement or drift, resulting in inaccurate 4D rs-fMRI data acquisition. The verification cycle of the acquisition equipment in the acquisition unit is then adjusted based on the ratio of the average value to the preset average value. The preset ratio U0 = 1.26. The comparison process between the average value and the preset average value U0 and the preset ratio U0 is as follows: If the ratio U of the average value to the preset average value is less than or equal to the preset ratio U0, the verification cycle of the data acquisition unit will be adjusted to 0.93 times the original verification cycle. The unit of the verification cycle is hours, and the adjusted verification cycle values ​​are all rounded up. If the ratio U of the average value to the preset average value is greater than the preset ratio U0, the verification cycle of the data acquisition unit will be adjusted to 0.81 times the original verification cycle. The unit of the verification cycle is hours, and the adjusted verification cycle values ​​are all rounded up.

[0049] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A functional magnetic resonance imaging-based brain disease target positioning system, characterized in that, The method comprises the following steps: a collecting unit is configured to collect 4D rs-fMRI data in real time; a calculating unit is connected to the collecting unit and configured to calculate correlation coefficients between voxels based on the 4D rs-fMRI data, and generate a whole-brain functional connectivity map of an individual patient, starting from a preset seed point region covering a key network of the whole brain; a screening unit is connected to the calculating unit and configured to calculate Z scores of functional connectivity strengths of voxels in the whole-brain functional connectivity map by using a statistical test method, and screen abnormal voxels based on a preset threshold, wherein the Z score is a number of standard deviations of an individual voxel functional connectivity strength from an average value of a normal group; a determining unit is connected to the screening unit and configured to extract functional features from a plurality of clusters obtained by clustering the abnormal voxels by using a clustering algorithm, predict therapeutic efficacy scores of the plurality of clusters by using a machine learning model trained based on historical patient data, and determine candidate therapeutic target points from the clusters screened based on the therapeutic efficacy scores, wherein the historical patient data comprises rs-fMRI data, target point coordinates of a therapy, and clinical therapeutic efficacy scores; a target point coordinate optimization unit is connected to the determining unit and configured to convert the plurality of candidate therapeutic target points into a plurality of accurate target points based on an individual anatomical structure by using a nonlinear registration algorithm; an analysis unit is connected to the target point coordinate optimization unit and configured to calculate deviations of the plurality of accurate target points from actual therapeutic target points of the patient, determine a target point positioning state based on the deviations, adjust the preset threshold based on the target point positioning state, and adjust an iteration number of the nonlinear registration algorithm based on the target point positioning state after the preset threshold is adjusted.

2. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 1, characterized in that, The analysis unit is further configured to calculate Euclidean distances between each of the accurate target points and the actual therapeutic target points, and determine the deviations based on an average value of the plurality of Euclidean distances; The analysis unit is further configured to calculate variances of a plurality of historical deviations in a case where the target point positioning state is unqualified; The analysis unit is further configured to adjust the preset threshold based on a ratio of a preset variance to the variances in a case where the variances are less than the preset variance; In a case where the deviations are greater than a preset deviation, the target point positioning state is determined to be unqualified.

3. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 2, characterized in that, The analysis unit is further configured to increase the preset threshold based on the ratio of the preset variance to the variances, and an increase amplitude of the preset threshold is proportional to the ratio.

4. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 3, characterized in that, The analysis unit is further configured to increase the preset threshold based on a ratio of a preset number of the abnormal voxels to a number of the abnormal voxels, and an increase amplitude of the preset threshold is proportional to the ratio.

5. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 4, characterized in that, The target point coordinate optimization unit further comprises a registration module configured to display a registered individual brain image and a standard template image side by side; The analysis unit is further configured to repeatedly adjust the preset threshold at least once in a case where the target point positioning state is unqualified after the preset threshold is adjusted, and stop adjusting when an adjustment number is less than a preset number and the target point positioning state is qualified or the adjustment number is equal to the preset number; The analysis unit is further configured to calculate a difference between the deviations and the preset deviations in a case where the target point positioning state is unqualified after the stop of the adjustment. The analysis unit is further configured to start the registration module if the difference is greater than the preset difference, and adjust the iteration number of the nonlinear registration algorithm based on a ratio of the difference to the preset difference if the individual brain image generated by the registration module is not aligned with the standard template image. The deviation is less than or equal to the preset deviation, and the target point positioning state is determined to be qualified.

6. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 5, characterized in that, The analysis unit is further configured to increase the iteration number of the nonlinear registration algorithm based on the ratio of the difference to the preset difference, and the increase amplitude of the iteration number of the nonlinear registration algorithm is proportional to the ratio.

7. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 6, characterized in that, The analysis unit is further configured to draw a time-historical deviation curve based on the obtained historical deviations if the target point positioning state is unqualified after the iteration number of the nonlinear registration algorithm is adjusted. If the integral of the curve is greater than a preset integral, the number of historical patient data is adjusted based on a ratio of the integral of the curve to the preset integral.

8. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 7, characterized in that, The analysis unit is further configured to increase the number of historical patient data based on the ratio of the integral of the curve to the preset integral, and the increase amplitude of the number of historical patient data is proportional to the ratio.

9. The functional magnetic resonance imaging-based brain disease target positioning system according to claim 8, characterized in that, The analysis unit is further configured to calculate an average value of the slope of the time-historical deviation curve if the target point positioning state is unqualified after the number of historical patient data is adjusted. The analysis unit is further configured to reduce the verification period of the acquisition device of the acquisition unit based on a ratio of the average value to a preset average value if the average value is greater than the preset average value, and the reduction amplitude of the verification period of the acquisition device of the acquisition unit is proportional to the ratio.

10. A method for locating a target of a brain disease based on functional magnetic resonance imaging, applied to the system for locating a target of a brain disease based on functional magnetic resonance imaging according to any one of claims 1-9, characterized in that, The method comprises: acquiring 4D rs-fMRI data in real time through an acquisition unit; calculating correlation coefficients between voxels based on the 4D rs-fMRI data through a calculation unit connected to the acquisition unit, taking a preset seed point region covering a key network of the whole brain as a starting point, to generate a whole brain functional connectivity map of the patient individual; calculating Z scores of functional connectivity strengths of voxels in the whole brain functional connectivity map using a statistical test method through a screening unit connected to the calculation unit, and screening abnormal voxels based on a preset threshold, wherein the Z score is a number of standard deviations of the individual voxel functional connectivity strength from the average value of the normal group; extracting functional features from a plurality of clusters obtained by clustering the abnormal voxels through a determination unit connected to the screening unit, predicting therapeutic efficacy scores of the plurality of clusters based on the functional features using a machine learning model trained based on historical patient data, and determining candidate treatment target points based on the clusters screened out based on the therapeutic efficacy scores, wherein the historical patient data includes rs-fMRI data, target point coordinates of treatment, and clinical therapeutic efficacy scores; converting a plurality of the candidate treatment target points into a plurality of accurate target points based on individual anatomical structures through a target point coordinate optimization unit connected to the determination unit; calculating deviations of the plurality of accurate target points from actual treatment target points of the patient through an analysis unit connected to the target point coordinate optimization unit, determining a target point positioning state based on the deviations, adjusting the preset threshold based on the target point positioning state, and adjusting the iteration number of the nonlinear registration algorithm based on the target point positioning state after the preset threshold is adjusted.

Images