Brain abnormity network positioning method and system based on function connection network mapping

Abstract

The invention relates to a brain anomaly network positioning method and system based on functional connection network mapping in the technical field of neural image data processing. The brain abnormal network positioning method comprises the following steps: calculating a whole brain function network diagram connected with each abnormal site by using resting state function connection data of large-scale health subjects based on a plurality of dispersed abnormal sites reported in previous literatures; then superposing the function network diagrams to obtain a network probability graph; and finally, filtering the probability graph through a threshold value to obtain a final core anomaly network graph. According to the method, the inherent functional connection architecture of the brain is used as a reference system, abnormal sites which seem to be uncorrelated in different researches are successfully traced and unified to a common and stable functional network, and compared with a single brain region marker, the generated network-level biomarker integrates more source evidences, so that the network-level biomarker has the advantages that the network-level biomarker can be widely applied to the field of biomarkers of the brain region. Therefore, the problems of result heterogeneity and inconsistency in brain abnormality discovery are solved.

Description

Technical Field

[0001] This invention relates to the field of neuroimaging data processing technology, and in particular to a method and system for locating brain abnormalities based on functional connectivity network mapping. Background Technology

[0002] Mental and neurological disorders, such as schizophrenia and depression, have complex neuropathological bases that are both a focus and a challenge in modern medical research. Over the past few decades, neuroimaging techniques, particularly magnetic resonance imaging (MRI), have become core tools for exploring the structural and functional abnormalities of the brain in these diseases. By comparing brain differences between patient populations and healthy controls, researchers worldwide have accumulated and published a vast amount of research, reporting tens of thousands of three-dimensional coordinates of abnormal brain regions associated with specific diseases or symptoms. However, current technology faces several key and interrelated technical bottlenecks in utilizing this wealth of research findings and translating them into clinical applications: The high heterogeneity and low reproducibility of results: For the same disease, or even the same symptom, different studies often report abnormal brain regions with inconsistent or even contradictory anatomical locations. For example, in schizophrenia research, some studies found prefrontal cortex gray matter atrophy, while others may have located it in the temporal lobe, hippocampus, or thalamus. Similarly, brain dysfunction sites associated with suicidal behavior are scattered across multiple different regions, including the ventral prefrontal cortex, dorsal prefrontal cortex, anterior cingulate cortex, and insula. Current neuroimaging studies report a wide anatomically distributed range of abnormal brain sites associated with specific diseases (such as schizophrenia) or symptoms (such as auditory hallucinations), with conflicting conclusions. This high heterogeneity makes it difficult to independently replicate and verify the findings of any single study, resulting in poor stability of the discovered biomarkers, a lack of universally accepted clinical value, and severely hindering the translation of neuroimaging findings into clinical practice.

[0003] The fundamental limitation of the "regional localization" paradigm: A fundamental reason for the aforementioned heterogeneity lies in the fact that most existing technologies are limited to the analytical paradigm of "regional localization." Even coordinate-based meta-analysis, used to integrate multiple studies, essentially remains at the level of spatial probability aggregation of the coordinates of isolated, discrete abnormal brain regions, failing to reveal the intrinsic connections between these discrete points. However, the general consensus in neuroscience is that complex mental illnesses are not caused by isolated lesions in a single brain region, but rather stem from the overall dysfunction of a functionally interconnected distributed brain network composed of multiple brain regions. Therefore, attributing the root cause of disease to a single "hotspot" region is an oversimplification of its complex pathophysiological mechanisms.

[0004] Lack of a unified integrative analytical framework: Due to the aforementioned limitations, the field has long lacked a technical framework capable of effectively integrating these spatially seemingly random and heterogeneous brain abnormalities and systematically mapping them into a unified brain functional network with clear biological significance. In other words, there is a lack of a methodology to answer the question: Are the temporal lobe abnormalities found in study A and the prefrontal lobe abnormalities found in study B essentially different members of the same dysfunctional "criminal network"? The lack of precise therapeutic targets limits the clinical efficacy of neuromodulation techniques: While the development of non-invasive neuromodulation techniques such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS) has provided new avenues for the treatment of mental illnesses, the efficacy of these techniques largely depends on the precise selection of stimulation targets. Due to the aforementioned limitations in "regional localization" and the heterogeneity of outcomes, existing techniques have failed to provide unified, robust, neural circuit-based therapeutic targets for specific symptoms (such as auditory hallucinations and core depressive symptoms). In clinical practice, target selection often relies on broad anatomical regions (such as the left dorsolateral prefrontal cortex) or experience, leading to significant individual differences in efficacy and limiting the potential and widespread application of these advanced therapeutic techniques. Summary of the Invention

[0005] To address the technical problems of heterogeneity, inconsistency, and limitations of "regional localization" in existing brain abnormality detection research, this invention provides a brain abnormality network localization method and system based on functional connectivity network mapping.

[0006] In a first aspect, this invention proposes a method for locating brain abnormalities based on functional connectivity network mapping, which locates brain abnormalities by constructing a core abnormality network graph; the method for constructing the core abnormality network graph includes: For the m-th type of brain abnormality, three-dimensional coordinates of the peak values ​​in brain regions where differences exist between the patient group and the healthy control group are collected, and spheres with a set radius are generated with each three-dimensional coordinate point as the center. These spheres are then merged to form a seed mask. fMRI datasets from N healthy subjects are collected. The functional connectivity strength between the BOLD time series of each voxel in the n-th fMRI dataset and the average BOLD time series of the seed mask is calculated, and the n-th functional connectivity map is generated. P f ; Traverse N fMRI datasets to obtain N P f , n∈[1,N]; for N terms P fA group-level one-sample t-test was performed on the voxels to obtain the t-value map; after correction and binarization of the t-value map, the m-th network mask map was obtained. P m .

[0007] By traversing M types of brain abnormalities, M results are obtained. P m , m∈[1,M]; M items P m The network probability map is obtained by superimposing it onto a standard brain template. P p ;reserve P p Voxels with a probability greater than the overlap threshold are used to obtain the core anomaly network diagram.

[0008] Secondly, this invention also proposes a brain abnormality network localization system based on functional connectivity network mapping, which uses the brain abnormality network localization method based on functional connectivity network mapping from the first aspect. The brain abnormality network localization system includes a construction module and a network mask generation module.

[0009] The construction module is used to build the core abnormality network graph, including: processing the m-th type of brain abnormality to obtain the m-th network mask graph. P m ; Traversing M types of brain abnormalities yields M results P m , m∈[1,M]; M items P m The network probability map is obtained by superimposing it onto a standard brain template. P p ;reserve P p Voxels with a probability greater than the overlap threshold are used to obtain the core abnormality network map. The network mask generation module collects the three-dimensional coordinates of peak brain regions where differences exist between the patient group and the healthy control group, and generates spheres with a set radius centered at each three-dimensional coordinate point; these spheres are merged to form a seed mask; fMRI datasets from N healthy subjects are collected; the functional connectivity strength between the BOLD time series of each voxel in the nth fMRI dataset and the average BOLD time series of the seed mask is calculated, and the nth functional connectivity map is generated. P f ; Traverse N fMRI datasets to obtain N P f , n∈[1,N]; for N terms P f One-sample t-tests were performed on the voxels at the group level to obtain t-plots; the t-plots were then corrected and binarized to obtain... P m .

[0010] In a second aspect, the present invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, enables the implementation of the steps of the brain abnormality network localization method based on functional connectivity network mapping as described in the first aspect.

[0011] The beneficial effects of this invention are as follows: 1. This invention utilizes the brain's inherent functional connectivity architecture as a reference system to successfully trace and unify seemingly unrelated abnormal sites in different studies (such as gray matter atrophy points in the frontal, temporal, and parietal lobes reported in different studies of schizophrenia) onto a common and robust functional network. The resulting network-level biomarkers integrate more source evidence compared to single brain region biomarkers, thus exhibiting higher stability and cross-sample reproducibility, thereby solving the problems of heterogeneity and inconsistency in the discovery of brain abnormalities.

[0012] 2. This invention provides a scheme for precise mapping from "anatomical location" to "functional network". This scheme uses resting-state functional connectivity data from a large number of healthy people as a "standard functional map" to reveal the functionally tightly coupled distributed brain network basis that is commonly pointed to behind discrete abnormal sites found in different studies. Furthermore, by quantitatively comparing it with classic functional networks such as default mode networks and attention networks, this abnormal network is given clear and interpretable neurocognitive functional attributes, thereby overcoming the limitations of "regional localization" in traditional methods.

[0013] 3. Compared with single studies or traditional meta-analysis results, this invention integrates data and evidence from multiple studies, exhibiting higher statistical robustness and cross-sample reproducibility, providing objective and reliable neuroanatomical basis for the auxiliary diagnosis, staging, prognosis, and efficacy evaluation of diseases.

[0014] 4. This invention can not only identify the entire dysfunctional network, but also identify key nodes in the network, thereby providing new, network-based, and more promising intervention targets for neuromodulation therapy, in order to improve the accuracy and individualization of treatment effects.

[0015] 5. This invention provides validated and objectively effective intervention targets for neuromodulation therapy, successfully bridging basic research and clinical practice. The core abnormal network it locates is not a purely theoretical model, and its clinical guiding value has been directly verified. For example, in the study of auditory hallucination symptoms in schizophrenia, the "auditory hallucination state dysfunction network" located by this method has been confirmed.

[0016] 6. This invention precisely covers the TMS and tDCS stimulation targets that have been proven clinically effective in multiple independent studies. This discovery, for the first time, unifies the pathophysiology of disease with the intervention targets for treatment at the network level, strongly demonstrating that the network identified in this invention is a "clinically effective network." This not only provides an objective basis for optimizing existing treatment plans, but also paves the way for developing novel, personalized, and precise treatment strategies targeting specific network nodes, thereby efficiently translating basic research findings into fruitful clinical applications, possessing significant social and economic value. Attached Figure Description

[0017] Figure 1 This is a flowchart of the brain abnormality network localization method based on functional connectivity network mapping in the embodiment; Figure 2 This is the network probability graph in the embodiment; Figure 3 This is a schematic diagram showing the proportion of explanations from abnormal brain networks and classical networks in the embodiments; Figure 4 This is a flowchart in the embodiment verifying whether the results of the brain abnormality network localization method based on functional connectivity network mapping are correct. Detailed Implementation

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "or / and" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] This invention utilizes large-scale resting-state functional magnetic resonance imaging (fMRI) data as a standard functional connectivity atlas to systematically integrate and precisely locate anatomically discrete brain abnormalities (including gray matter structural abnormalities, task-oriented functional activation abnormalities, and resting-state activity abnormalities) associated with specific neuropsychiatric disorders (such as schizophrenia) or specific clinical symptoms (such as auditory hallucinations) from multiple independent neuroimaging studies. Furthermore, it performs spatial correlation analysis between the located abnormal networks and recognized classical brain functional networks, thereby providing quantitative functional characterization and interpretation of the located brain abnormal networks. Based on this, this embodiment proposes a brain abnormality network localization method based on functional connectivity network mapping. The core of this method lies in how to construct a unified and robust core abnormality network map from scattered and heterogeneous literature evidence. How to construct the core abnormality network map is the focus of this localization method. The method for constructing the core abnormality network map includes: processing the m-th type of brain abnormality to obtain the m-th network mask map. P m ; Traversing M types of brain abnormalities yields M results P m , m∈[1,M]; M items P m The network probability map is obtained by superimposing it onto a standard brain template. P p ;reserve P p Voxels with a probability greater than the overlap threshold are used to obtain the core anomaly network diagram. P m The method includes: acquiring the three-dimensional coordinates of peak brain regions where differences exist between the patient group and the healthy control group, and generating spheres with a set radius centered on each three-dimensional coordinate point; merging each sphere to form a seed mask; acquiring fMRI datasets from N healthy subjects; calculating the functional connectivity strength between the BOLD time series of each voxel in the nth fMRI dataset and the average BOLD time series of the seed mask, and generating the nth functional connectivity map. P f ; Traverse N fMRI datasets to obtain N P f , n∈[1,N]; for N terms P f One-sample t-tests were performed on the voxels at the group level to obtain t-plots; the t-plots were then corrected and binarized to obtain... P m Please refer to Figure 1 The following section details the steps involved in the brain abnormality network localization method.

[0021] S1. Data collection and parameter setting.

[0022] The field of neuroimaging has accumulated a vast amount of research data over decades. However, for the same disease (such as schizophrenia), different research reports show a wide variety of "abnormal brain regions," leading to contradictory conclusions. This is akin to having many reports on crime activity in a city, but each report identifies different "crime hotspots," making it impossible to pinpoint the true "core area of ​​the crime network." The first step in this embodiment is to systematically collect and structure these scattered "clues" and prepare a "standard atlas" for analysis. The collected data includes the following: 1. Published neuroimaging studies related to M types of brain abnormalities (such as "gray matter atrophy in schizophrenia" and "abnormal resting-state brain activity in patients with depression"). For example, in this embodiment, a study aimed at locating the gray matter atrophy network in schizophrenia collected data from 109 independent VBM (voxel-based morphometry) studies. This is the data source of this invention, aiming to structure and standardize existing, scattered research results in the field, serving as the original input for subsequent integrated analysis. It can objectively and systematically incorporate evidence from different laboratories around the world, maximizing the aggregation of existing knowledge and avoiding the limitations of single studies due to small sample sizes and differences in research design, thus laying a solid data foundation for the subsequent discovery of general patterns.

[0023] 2. Resting-state fMRI dataset from N healthy subjects. This fMRI dataset requires preprocessing before use. Preprocessing methods include: removal of initial time points, slice time correction, head movement correction (e.g., assessing and excluding subjects with excessive head movements), spatial normalization to MNI standard space, removal of various interfering signals (covariate regression, such as Friston-24 head movement parameters, whole-brain average signal, white matter signal, cerebrospinal fluid signal, etc.), bandpass filtering (e.g., 0.01-0.1 Hz), and spatial smoothing (e.g., using a 6 mm full-width Gaussian kernel). This collected fMRI dataset can serve as a "standard version of a brain functional navigation atlas." This "brain functional navigation atlas" does not depict anatomical structures, but rather depicts the "functional highway network" of information flow between different brain regions through the temporal correlation of spontaneous low-frequency oscillations (0.01-0.1 Hz) of blood oxygen level-dependent (BOLD) signals.

[0024] 3. Standard Brain Template. This standard brain template is a standardized whole-brain gray matter mask file, which can be used to set the effective computational boundaries of the "navigation map," ensuring that all subsequent functional connectivity analyses are confined to the gray matter regions responsible for higher cognitive functions, excluding interference from white matter, cerebrospinal fluid, and extracranial tissues. The collected "brain functional navigation map atlas" is the reference system and core benchmark for network "querying" and "mapping" in this invention. The "computational boundaries" can significantly improve the accuracy, efficiency, and signal-to-noise ratio of the calculations.

[0025] 4. Classical Functional Network Template Set. This can adopt the 7-network or 17-network template published by Yeo et al. in 2011, or the 14-network template published by Shirer et al. in 2012. This network template divides the brain's gray matter into several functionally independent networks. For example, the network template divides the brain into the default mode network (DMN), sensorimotor network (SMN), visual network (VN), attention network (including dorsal DAN and ventral VAN), frontoparietal control network (FPN), limbic network (LN), and subcortical network, etc.

[0026] For the computational process involved in brain abnormality network localization methods, core analysis parameters can be pre-defined. These parameters include: 1. Seed point radius: Defines the radius of the sphere generated around the 3D coordinate point. In this embodiment, the radius is set to 4mm.

[0027] 2. Statistical correction method: In this embodiment, multiple comparison correction method is used in the correction process of T-value plot, and the error detection rate, phylogenetic error and significance level are preset.

[0028] 3. Overlap threshold: Set the probability threshold used to define the core anomalous network graph during final integration, such as 50% or 60%.

[0029] S2. Construct the network mask diagram P m The data collected in S1 consists only of a set of discrete coordinate points, which is insufficient for network analysis. These abstract "points" need to be transformed into a biologically meaningful and analyzable "network." This is one of the core steps of this invention and a key differentiator from traditional meta-analysis methods. All M types of brain abnormalities must be processed to obtain M [data points / data]. P m The following explanation will use one type of brain abnormality, "Study 1," as an example.

[0030] Based on "Research 1", construct P m The method includes the following steps: S21. Collect brain abnormality data from "Study 1" and identify the three-dimensional coordinates of the peak values ​​in brain regions where differences exist between the patient group and the healthy control group. Generate spheres in a three-dimensional standard space with the three-dimensional coordinates as the center, using a pre-set seed point radius (4mm). Merge all spheres to form a single, irregularly shaped seed mask. This step transforms all discrete abnormal points found in "Study 1" from spatial coordinates without a physical entity (only indicating 'where') into a unified analytical object with anatomical volume (clarifying 'what'). Existing technologies (such as traditional CBMA) only treat these discrete coordinate points as independent statistical events. This invention is the first to propose "packaging" all abnormal points from the same comparative experiment into a whole, treating it as a unified functional probe. This achieves the transformation from scattered "points" to a unified "surface," which is the logical starting point and necessary prerequisite for subsequent network analysis, ensuring that subsequent analysis can fully represent all the findings of this study.

[0031] S22. Calculate the functional connectivity strength between the BOLD time series of each voxel in the nth fMRI dataset and the average BOLD time series of the seed mask of "Study 1". In this embodiment, the Pearson correlation coefficient is used to calculate the functional connectivity strength to obtain the correlation coefficient value. Subsequently, the correlation coefficient value is subjected to Fisher-Z transform to improve normality and generate an individualized functional connectivity map for "Study 1". P f This step utilizes a "functional connectivity map" of the healthy brain to identify all brain regions that exhibit closely coordinated activity with the aberration origins defined in "Study 1." This is equivalent to taking the "business card" (seed mask) from "Study 1" and asking each healthy individual: "In your brain, which areas are functionally 'in the same circle' as the areas on this business card?" Through this step, anatomical abnormalities (such as gray matter atrophy) or functional abnormalities (such as task activation) can be linked to the brain's intrinsic functional organization. It uses standardized functional connectivity data to "decode" the network properties behind structural aberration sites, surpassing single-modality data analysis and serving as a crucial bridge for cross-modal and cross-research integration.

[0032] S23. Each of the N healthy participants generated a corresponding functional connectivity diagram for "Study 1". P f For N P fA group-level one-sample t-test was performed on the voxels to obtain a t-plot, which is a highly robust plot representing the population and functionally significantly connected to the anomalous sites in "Study 1". This is equivalent to extracting stable, universally applicable connection patterns across individuals from hundreds of individualized "friendship" maps, excluding individual-specific "random friends" or noise. This step only considers positive connections because the biological significance of negative connections remains controversial; this ensures a clearer and more robust interpretation of the results.

[0033] S24. Perform rigorous statistical correction on the T-value plot to control the false positive rate. A pre-defined statistical correction method (multiple comparison correction) is used. Significant regions that pass the correction are assigned a value of 1, and insignificant regions are assigned a value of 0. Specifically, significant regions are those where the p-value is less than the significance level (in this embodiment, the significance level can be set to 0.05) in the multiple comparison correction method; insignificant regions are those where the p-value is greater than or equal to the significance level. Finally, a binary network mask is generated, showing only black and white values. P m By binarizing, the analytical results from different studies (which may differ in research methods, scanner parameters, and participant populations) are forcibly converted into a unified and standardized data format. This transforms continuous statistical value maps into a clear and anatomically defined "network map," greatly simplifying subsequent integration calculations and solving the previous problem of being unable to directly compare and superimpose research results due to differences in standards and effect sizes.

[0034] Through processes S21 to S24, the research data of all brain abnormalities, such as "Research 2", "Research 3", ..., "Research M", are calculated in parallel, and corresponding network mask diagrams are generated for each. P m .

[0035] S3. Constructing the core abnormality network diagram is essentially an integration of multiple research findings. In this embodiment, 109 brain abnormality data points were collected to obtain 109 network mask diagrams. P m However, differences may still exist among them. It is necessary to find the core, most consistent "common network" from these diagrams to ultimately resolve the initial problem of "heterogeneity in research findings." This specifically includes the following steps: S31. Initialize a standard brain template, transforming it into a three-dimensional all-zero matrix, and then... (The sentence is incomplete and requires more context to translate accurately.) P m One by one, they are superimposed onto this standard brain template to obtain the network probability map. P p .like Figure 2 As shown, it is a network probability diagram corresponding to the network impairment of auditory hallucinations in schizophrenia. Figure 2 Different colors (from green to yellow) represent the probability that different voxels belong to the core anomalous network. P p Dividing the value of each voxel by M yields the probability of that voxel. Suppose a location in the brain, in 109 studies, 80 generated... P m If all values ​​at this point are 1 (i.e., belong to the network), then the summed value at that point is 80. Dividing this by 109 gives a final probability value of 0.734. This means that there is a 73.4% probability that this location belongs to the core anomalous network being searched for. This step is essentially a "spatial voting" process. Each location (voxel) in the brain is being counted based on the number of independent studies that generated this value. P m It was covered. The final result... P p It visually demonstrates the probability of different parts of the brain becoming members of the core network, which is itself a highly valuable intermediate result.

[0036] S32, to P p Apply a preset overlap threshold so that only [the element] is retained. P p Voxels with a probability greater than an overlap threshold are selected, while the values ​​of the remaining voxels are set to 0. This overlap threshold filters out connection regions that only occasionally appear in a few studies, ultimately identifying the most robust and core abnormal network graph that is common in the vast majority of studies. This step ensures the high stability and reproducibility of the final output, fundamentally solving the problem of research heterogeneity and finding the "greatest common denominator" behind the heterogeneity findings that has plagued the field for many years—that unified core abnormal network graph. Based on this core abnormal network graph, the corresponding abnormal brain network can be quickly located for a specific mental illness. For example, when it is necessary to understand the abnormal brain network involved in auditory hallucinations, the corresponding brain lesion area can be quickly located by querying the core abnormal network graph.

[0037] In terms of specific applications after obtaining the core abnormal network map, it can quickly locate the abnormal brain networks for a patient's symptoms / diseases and provide treatment plans. For example, a patient may experience auditory hallucinations. Based on the characteristic symptoms of auditory hallucinations, the damaged brain networks corresponding to these symptoms can be identified. Then, electrical stimulation (tDCS) or magnetic stimulation (TMS) can be used to modulate the corresponding brain networks to alleviate the patient's symptoms. Similarly, it can be applied to other brain diseases, such as depression and anhedonia. This core abnormal network map is used to find the brain networks corresponding to the disease. That is, because this brain network is damaged, the corresponding brain disease occurs. After finding the relevant abnormal brain networks, modulation and treatment can be carried out to alleviate or treat the patient's symptoms.

[0038] S4. Although the core abnormal network map obtained in the above steps can locate networks based on the patient's symptoms or condition, it is only a spatial distribution map. It does not know "what this located network represents in terms of function." This step assigns a functional "label" to the located network by comparing it with known brain functional systems. Specifically, it includes the following steps: The core abnormality network graph is loaded onto a set of classic functional network templates, and the correlation between each classic network in the classic functional network template set and the core abnormality network graph is calculated. This yields the explanatory percentage of the brain abnormality network and the classic network, such as... Figure 3 As shown. In this embodiment, the degree of correlation between the classic network and the core anomalous network graph can be measured by the metric of "overlap ratio". There are two ways to calculate the "overlap ratio": Method 1 (Explaining the composition of the anomalous network): Calculate the number of voxels in the overlapping portion of the core anomalous network graph and the classic network, and then divide by the total number of voxels in the core anomalous network graph. For example, a result of 0.25 indicates that 25% of the core anomalous network graph falls within a specific classic network.

[0039] Method 2 (Interpreting the extent of impact on the classical network): Calculate the number of voxels in the overlapping area between the core anomalous network graph and the classical network, and then divide by the total number of voxels in the classical network. For example, a result of 0.40 indicates that 40% of the region of this particular classical network is covered by the core anomalous network graph, i.e., the degree of its impact.

[0040] This step, through quantitative spatial comparison, allows for an objective assessment of which known functional systems the newly discovered anomalous networks are primarily associated with. It moves beyond simply presenting the analysis as an "unknown mind map," providing direct, clear, and interpretable functional implications by linking the results to recognized knowledge systems (classical networks). For example, the analysis could directly tell the user: "The gray matter atrophy network in schizophrenia is primarily a complex network with ventral attention network (24.3%), sensorimotor network (20.0%), and default mode network (18.8%) as its core components." This significantly enhances the scientific and clinical communication value of the results.

[0041] In this embodiment, the brain abnormality network localization method based on functional connectivity network mapping ultimately outputs a thresholded core abnormality network map, which can be stored as a standard .nii file. This file can be read and analyzed by all mainstream neuroimaging software, integrating the findings of all heterogeneous studies and locating the core brain networks most relevant to specific diseases or symptoms. Simultaneously, it can also generate an analysis report containing visualizations of analysis parameters, intermediate processes, and final results.

[0042] To verify the correctness of the results of the brain abnormality network localization method based on functional connectivity network mapping in this invention, a permutation test was used to verify the results. The permutation test is a classic nonparametric statistical method whose core idea is to "create a set of 'fake data' with the same structure as the 'real data' but random content, and calculate the corresponding results to highlight how special the calculation results of the 'real data' are." Figure 4 As shown, its inspection process includes: Within the whole-brain gray matter mask, a random coordinate set is randomly generated, identical to the original number of studies and the number of coordinate points for each brain abnormality study. Processes S2 and S3 are then executed on this random coordinate set to generate an empty network graph based on random input. This process of generating the random coordinate set and corresponding empty network graph is repeated 1000 times, resulting in a null distribution of 1000 different shaped empty network graphs. Then, the spatial similarity between the core abnormality network graph obtained from real data and the 1000 empty network graphs is calculated (using the Dice coefficient). A p-value is calculated by comparing the similarity score of the core abnormality network graph within the null distribution. The results obtained by this invention, based on this testing method, show that the core abnormality network graph based on real data is significantly different from the empty network graph (the Dice coefficient is significantly lower than the average level of the empty network graph). This statistically proves that the disease-related network located by this invention has high biological specificity and is not formed by chance. This provides strong statistical evidence for the effectiveness of the brain abnormality network localization method based on functional connectivity network mapping provided by this invention, which is independent of data distribution assumptions and greatly enhances the scientific rigor and persuasiveness of the conclusions of this invention.

[0043] In summary, this invention first treats the heterogeneous, discrete sets of anomalous coordinates reported in each independent study as multiple "entry points" into the same potential functional network, rather than as isolated statistical results. Through the generation of a unique, study-specific seed mask, these discrete "entry points" are transformed into a unified analytical probe with anatomical volume. Subsequently, using large-scale fMRI data from healthy subjects as a "standard brain functional connectivity map," the functional connectivity range of each probe in the standard brain is queried and calculated. Finally, through group-level statistical tests and crucial binarization processing, the results from each different source and with varying standards are transformed into a clear, unambiguous, and uniformly formatted network mask. This stage, using "functional connectivity" as a universal language, achieves the first crucial transformation from "heterogeneous coordinates" to a "standardized network map," laying a solid foundation for subsequent cross-study integration. Furthermore, after obtaining multiple standardized network masks, this invention employs a "spatial voting" integration strategy. By stacking all network masks at the voxel level and calculating the frequency of network coverage at each spatial location, a whole-brain network probability map is generated. The value of each point on the network probability graph quantitatively represents the probability that the point belongs to the final core network. By applying a preset overlap threshold to the network probability graph, a stable and core anomalous network that is common in most studies is extracted from the probability distribution. This stage cleverly utilizes probabilistic integration to elevate the "consensus" from a discrete point space to a continuous network space, perfectly solving the problem of inconsistent conclusions between different studies that existing technologies cannot resolve. Finally, this invention also performs quantitative spatial overlap analysis with the extracted core anomalous network and the recognized classic functional network maps in the field. By calculating correlation indicators, this invention assigns clear functional attributes to this newly discovered core anomalous network, clearly answering the core scientific question of "which functional systems of the brain are mainly affected by this core anomalous network." In summary, the working principle of this invention is a layer-by-layer information extraction and evidence integration process from "heterogeneous evidence" to "standardized network intermediates," then to "probabilistic consensus," and finally to "statistically validated robust core networks." Furthermore, by creatively using data related to functional connectivity groups as a bridge and reference system, it successfully mapped and unified scattered, multimodal anatomical / functional findings onto a common functional network framework, achieving a paradigm shift from "region" to "network".

[0044] In another embodiment, a brain abnormality network localization system based on functional connectivity network mapping is also proposed, which includes a construction module, a localization module, and a network mask generation module.

[0045] The construction module is used to build the core abnormality network graph, including: processing the m-th type of brain abnormality to obtain the m-th network mask graph. P m By traversing M types of brain abnormalities, M results are obtained. P m , m∈[1,M]. Take M items... P m The network probability map is obtained by superimposing it onto a standard brain template. P p .reserve P p Voxels with a probability greater than the overlap threshold are used to obtain the core abnormality network map. The localization module is used to detect the type of brain abnormality in patients and locate the abnormal brain networks based on the core abnormality network map. The network mask generation module is used to collect the three-dimensional coordinates of peak values ​​in brain regions where there are differences between the patient group and the healthy control group, and generate spheres with a set radius centered at each three-dimensional coordinate point. Each sphere is merged to form a seed mask. fMRI datasets from N healthy subjects are collected. The functional connectivity strength between the BOLD time series of each voxel in the nth fMRI dataset and the average BOLD time series of the seed mask is calculated, and the nth functional connectivity map is generated. P f Traverse N fMRI datasets to obtain N... P f , n∈[1,N]. For N terms P f One-sample t-tests were performed on the voxels at the group level to obtain t-plots. The t-plots were then corrected and binarized to obtain... P m .

[0046] In another embodiment, a computer-readable storage medium storing a computer program is also proposed. When executed by a processor, this computer program is capable of implementing the steps of the brain abnormality network localization method based on functional connectivity network mapping as described in the above embodiments.

[0047] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0048] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A brain abnormal network localization method based on functional connectivity network mapping, characterized in that, The method locates the brain abnormal network by constructing a core abnormal network graph; The method for constructing the core abnormal network graph comprises: For the mth brain abnormality, collect the three-dimensional coordinate points of the peak values of the brain regions that are different between the patient group and the healthy control group, and generate a sphere with a set radius with each three-dimensional coordinate point as the center of the sphere; combine each sphere to form a seed mask; collect fMRI data sets of N healthy subjects; calculate the functional connection strength between the BOLD time series of each voxel in the nth fMRI data set and the average BOLD time series of the seed mask, and generate the nth functional connection map P f ; traverse the N fMRI data sets to obtain N P f , n ∈ [1, N]; perform a one-sample T test on the voxels in the N P f at the group level to obtain a T value map; and after correction and binary processing of the T value map, obtain the mth network mask map P m ; M core abnormal networks are obtained by traversing M brain abnormalities P m , m ∈ [1, M]; M core abnormal networks are obtained by traversing M brain abnormalities P m , m ∈ [1, M]; M core abnormal networks are obtained by traversing M brain abnormalities P p ; the voxels with the probability greater than the overlap threshold value are reserved P p ; the voxels with the probability greater than the overlap threshold value are reserved 2. The method of claim 1, wherein the method further comprises: Obtaining the functional connectivity map P f The method comprises: calculating the functional connection strength between the voxels in the MRI data set and the seed mask using the Pearson correlation coefficient to obtain a correlation coefficient value, and then performing Fisher-Z transformation on the correlation coefficient value to obtain the functional connectivity map.

3. The method of claim 1, wherein, Correction of the T-value graph adopts a multiple comparison correction method, and error discovery rate, family error and significance level are preset; The binarization processing of the T-value graph comprises: assigning 1 to a significant region and 0 to a non-significant region through correction; wherein the significant region is a region with a p-value less than the significance level in the multiple comparison correction method; and the non-significant region is a region with a p-value greater than or equal to the significance level.

4. The method of claim 1, wherein, Network probability map P p The value of each voxel divided by M gives the probability of the corresponding voxel. The overlap threshold is set to 50% or 60%; P p The voxels with a value greater than the overlap threshold are retained, and the values of the remaining voxels are set to 0.

5. The method of claim 1, wherein, The method for locating the brain abnormal network based on the functional connection network mapping further comprises: The core abnormal network graph is loaded on a classical functional network template set, and the association degree between each classical network in the classical functional network template set and the core abnormal network graph is calculated to obtain an explanation proportion of the brain abnormal network and the classical network.

6. The method of claim 5, wherein, The classical functional network template set comprises: a default mode network, a sensory motor network, a visual network, an attention network, a frontal-parietal control network, a limbic network and a subcortical network according to brain function system division.

7. The method of claim 5, wherein the method further comprises: The calculation method of the association degree comprises: calculating the voxel number of the overlapping part of the core abnormal network graph and the classical network, and then dividing by the total voxel number of the core abnormal network graph. Or, the calculation method of the association degree comprises: calculating the voxel number of the overlapping part of the core abnormal network graph and the classical network, and then dividing by the total voxel number of the classical network.

8. The method of claim 1, wherein, The standard brain template comprises standardized whole brain gray matter mask data, which is used to set an effective calculation boundary and limit the brain abnormal network analysis in the brain gray matter region.

9. A brain abnormal network localization system based on functional connectivity network mapping, characterized in that, The method comprises: The constructing module is configured to construct the core abnormal network graph, including: processing the mth brain abnormality to obtain an mth network mask graph P m ; traversing M brain abnormalities to obtain M P m , m ∈ [1, M]; superimposing the M P m network probability graphs on a standard brain template to obtain a network probability graph P p ; retaining P p voxels with a probability greater than an overlap threshold value to obtain the core abnormal network graph a network mask map generation module for collecting three-dimensional coordinate points of peak values of brain regions that are different between the patient group and the healthy control group, and generating a sphere with a set radius with each three-dimensional coordinate point as the center of the sphere; merging each sphere to form a seed mask; collecting fMRI data sets of N healthy subjects; calculating the functional connection strength between the BOLD time series of each voxel in the nth fMRI data set and the average BOLD time series of the seed mask, and generating the nth functional connection map P f ; traversing the N fMRI data sets to obtain N P f , n ∈ [1, N]; performing a one-sample T test on the voxels in the N P f at the group level to obtain a T value map; and performing correction and binary processing on the T value map to obtain P m .

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to realize the steps of the method for locating the brain abnormal network based on the functional connection network mapping according to any one of claims 1 to 8.

Images