A method and device for analyzing individual magnetic resonance image abnormalities

Abstract

The application relates to the technical field of magnetic resonance image analysis, and discloses an individual magnetic resonance image anomaly analysis method and device, which comprises the following steps: for each analysis object, target voxel information of the analysis object is acquired; according to the target voxel information and a preset voxel difference calculation model, a target difference score of the analysis object is calculated, the target difference score is used for representing the deviation degree between the target voxel information and normal voxel distribution range information of the analysis object; it is judged whether the target difference score is greater than or equal to a preset difference score threshold value, and when the judgment result is yes, the analysis object is determined as an abnormal object; according to the target voxel information of all abnormal objects, a target difference matrix is determined; according to a non-negative matrix decomposition algorithm and the target difference matrix, a target difference factor and a target weight matrix are determined. It can be seen that the application can improve the efficiency of individual magnetic resonance image anomaly analysis, and is beneficial to timely guaranteeing the physical health of individuals.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging analysis technology, and in particular to a method and apparatus for analyzing individual magnetic resonance imaging anomalies. Background Technology

[0002] Magnetic resonance imaging (MRI) can obtain relatively rich tissue comparison information through non-invasive detection and has been widely used in clinical medical imaging examinations.

[0003] In existing technologies, due to the heterogeneity between different individuals, even under the same mental state, different individuals' magnetic resonance imaging (MRI) images show different characteristics representing individual structural abnormalities. Researchers often need to develop a one-to-one MRI image analysis plan for each individual. However, this analysis method is relatively inefficient and not conducive to timely protection of an individual's physical health.

[0004] Therefore, it is particularly important to propose a technical solution to improve the efficiency of individual magnetic resonance imaging anomaly analysis. Summary of the Invention

[0005] This invention provides a method and apparatus for analyzing anomalies in individual magnetic resonance imaging (MRI) images, which can help improve the efficiency of anomaly analysis in individual MRI images.

[0006] To address the aforementioned technical problems, the first aspect of this invention discloses a method for analyzing anomalies in individual magnetic resonance imaging, the method comprising:

[0007] For each analysis object, the target voxel information of the analysis object is obtained; based on the target voxel information and a preset voxel difference calculation model, the target difference score of the analysis object is calculated, whereby the target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object; it is determined whether the target difference score is greater than or equal to a preset difference score threshold; if it is determined that the target difference score is greater than or equal to the preset difference score threshold, then the analysis object is identified as an abnormal object.

[0008] Based on the target voxel information of all the aforementioned abnormal objects, determine the target difference matrix;

[0009] Based on the nonnegative matrix factorization algorithm and the target difference matrix, a target difference factor and a target weight matrix are determined. The target difference factor is used to represent the common features of all the abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each abnormal object.

[0010] As an optional implementation, in the first aspect of the invention, before calculating the target difference score of the analysis object based on the target voxel information and a preset voxel difference calculation model, the method further includes:

[0011] Obtain the physiological information of the analysis subject, wherein the physiological information includes at least one of age information, gender information, height information, skin color information, and body fluid information;

[0012] Based on the physiological information and the pre-trained Gaussian process regression model, the predicted voxel information of the analysis object is calculated, and the operation of calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model is triggered. The predicted voxel information is the normal voxel distribution range information of the analysis object.

[0013] As an optional implementation, in the first aspect of the present invention, the preset voxel difference calculation model is:

[0014]

[0015] Where Z represents the target difference score, VBM predicted VBM is used to represent the predicted voxel information. true Used to represent the target voxel information. Used to represent the preset expected variance. This is used to represent the prediction variance, which is calculated based on the prediction voxel information of all the analyzed objects.

[0016] As an optional implementation, in the first aspect of the present invention, the nonnegative matrix factorization algorithm is:

[0017] D = W * F + ε, W, F > 0

[0018] Wherein, D represents the target difference matrix, W represents the target weight matrix, F represents the target difference factor, and ε represents the preset bias.

[0019] As an optional implementation, in the first aspect of the present invention, the target difference factor has a corresponding preset optimal number, and the preset optimal number is determined by means of:

[0020] Divide the preset set of exception objects into at least two preset subsets of exception objects;

[0021] For each of the preset abnormal object subsets, a preset difference matrix is ​​obtained for the preset abnormal object subset. The preset difference matrix is ​​determined by the preset voxel information of each preset abnormal object in the preset abnormal object subset.

[0022] For each of the pre-optimal numbers in the pre-optimal number set, the pre-optimal difference matrix of the pre-optimal abnormal object subset is decomposed according to the pre-optimal number and the non-negative matrix factorization algorithm to obtain the first weight matrix and the first difference factor of the pre-optimal abnormal object subset, wherein the first weight matrix is ​​the product of the pre-optimal number and the number of the pre-optimal abnormal objects in the pre-optimal abnormal object subset.

[0023] Based on the preset difference matrix, the first weight matrix, and the first difference factor of the preset subset of abnormal objects, a first decomposition error is calculated, and the first decomposition error is used to represent the decomposition error of the preset difference matrix.

[0024] For each other preset abnormal object subset outside the preset abnormal object subset, the preset difference matrix of the other preset abnormal object subset is decomposed according to the pre-optimal number, the non-negative matrix factorization algorithm and the first difference factor to obtain the second weight matrix of the other preset abnormal object subset;

[0025] A second decomposition error is calculated based on the preset difference matrix, the second weight matrix, and the first difference factor of the other preset abnormal object subset. The second decomposition error is used to represent the decomposition error of the preset difference matrix of the other preset abnormal object subset when the first difference factor is applied.

[0026] Based on all the first decomposition errors and all the second decomposition errors, calculate the target calculated value of the pre-optimal number, which is used to represent the degree of deviation of the decomposition matrix corresponding to the pre-optimal number;

[0027] The preset optimal number is determined based on the target calculated value.

[0028] As an optional implementation, in the first aspect of the present invention, before determining the preset optimal number based on the target calculated value, the method for determining the preset optimal number further includes:

[0029] Determine whether the number of times the preset abnormal object set is divided is greater than or equal to a preset number threshold. If it is determined that the number of times the preset abnormal object set is divided is greater than or equal to the preset number threshold, then trigger the operation of determining the preset optimal number based on the target calculated value.

[0030] When it is determined that the number of times the preset abnormal object set is divided is less than the preset number threshold, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered.

[0031] As an optional implementation, in the first aspect of the present invention, determining the preset optimal number based on the target calculated value includes:

[0032] The average target calculated value for each of the same pre-optimal numbers is obtained by averaging all the target calculated values ​​of the same pre-optimal number.

[0033] Among all the average target calculation values, the preliminary optimal number of the smallest average target calculation value is determined as the preset optimal number.

[0034] A second aspect of the present invention discloses an individual magnetic resonance imaging anomaly analysis device, the device comprising:

[0035] The acquisition module is used to acquire the target voxel information of each analysis object;

[0036] The first calculation module is used to calculate the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model. The target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object.

[0037] The judgment module is used to determine whether the target difference score is greater than or equal to a preset difference score threshold.

[0038] The determination module is used to determine the analysis object as an abnormal object when the judgment module determines that the target difference score is greater than or equal to the preset difference score threshold.

[0039] The determining module is further configured to determine a target difference matrix based on the target voxel information of all the abnormal objects;

[0040] The determining module is further configured to determine a target difference factor and a target weight matrix based on a non-negative matrix factorization algorithm and the target difference matrix. The target difference factor is used to represent the common features of all the abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each of the abnormal objects.

[0041] As an optional implementation, in a second aspect of the present invention, the acquisition module is further configured to acquire physiological information of the analysis object before the first calculation module calculates the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model. The physiological information includes at least one of age information, gender information, height information, skin color information, and body fluid information.

[0042] The device also includes:

[0043] The second calculation module is used to calculate the predicted voxel information of the analysis object based on the physiological information and the pre-trained Gaussian process regression model, and to trigger the first calculation module to perform the operation of calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model, wherein the predicted voxel information is the normal voxel distribution range information of the analysis object.

[0044] As an optional implementation, in the second aspect of the present invention, the preset voxel difference calculation model is:

[0045]

[0046] Where Z represents the target difference score, VBM predicted VBM is used to represent the predicted voxel information. true Used to represent the target voxel information. Used to represent the preset expected variance. This is used to represent the prediction variance, which is calculated based on the prediction voxel information of all the analyzed objects.

[0047] As an optional implementation, in the second aspect of the present invention, the nonnegative matrix factorization algorithm is:

[0048] D = W * F + ε, W, F > 0

[0049] Where D represents the target difference matrix, W represents the target weight matrix, F represents the target difference factor, and ε represents the preset bias.

[0050] As an optional implementation, in the second aspect of the present invention, the target difference factor has a corresponding preset optimal number, and the preset optimal number is determined by means of:

[0051] Divide the preset set of exception objects into at least two preset subsets of exception objects;

[0052] For each of the preset abnormal object subsets, a preset difference matrix is ​​obtained for the preset abnormal object subset. The preset difference matrix is ​​determined by the preset voxel information of each preset abnormal object in the preset abnormal object subset.

[0053] For each of the pre-optimal numbers in the pre-optimal number set, the pre-optimal difference matrix of the pre-optimal abnormal object subset is decomposed according to the pre-optimal number and the non-negative matrix factorization algorithm to obtain the first weight matrix and the first difference factor of the pre-optimal abnormal object subset, wherein the first weight matrix is ​​the product of the pre-optimal number and the number of the pre-optimal abnormal objects in the pre-optimal abnormal object subset.

[0054] Based on the preset difference matrix, the first weight matrix, and the first difference factor of the preset subset of abnormal objects, a first decomposition error is calculated, and the first decomposition error is used to represent the decomposition error of the preset difference matrix.

[0055] For each other preset abnormal object subset outside the preset abnormal object subset, the preset difference matrix of the other preset abnormal object subset is decomposed according to the pre-optimal number, the non-negative matrix factorization algorithm and the first difference factor to obtain the second weight matrix of the other preset abnormal object subset;

[0056] A second decomposition error is calculated based on the preset difference matrix, the second weight matrix, and the first difference factor of the other preset abnormal object subset. The second decomposition error is used to represent the decomposition error of the preset difference matrix of the other preset abnormal object subset when the first difference factor is applied.

[0057] Based on all the first decomposition errors and all the second decomposition errors, calculate the target calculated value of the pre-optimal number, which is used to represent the degree of deviation of the decomposition matrix corresponding to the pre-optimal number;

[0058] The preset optimal number is determined based on the target calculated value.

[0059] As an optional implementation, in a second aspect of the present invention, before determining the preset optimal number based on the target calculated value, the method for determining the preset optimal number further includes:

[0060] Determine whether the number of times the preset abnormal object set is divided is greater than or equal to a preset number threshold. If it is determined that the number of times the preset abnormal object set is divided is greater than or equal to the preset number threshold, then trigger the operation of determining the preset optimal number based on the target calculated value.

[0061] When it is determined that the number of times the preset abnormal object set is divided is less than the preset number threshold, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered.

[0062] As an optional implementation, in a second aspect of the present invention, determining the preset optimal number based on the target calculated value includes:

[0063] The average target calculated value for each of the same pre-optimal numbers is obtained by averaging all the target calculated values ​​of the same pre-optimal number.

[0064] Among all the average target calculation values, the preliminary optimal number of the smallest average target calculation value is determined as the preset optimal number.

[0065] A third aspect of the present invention discloses another device for analyzing individual magnetic resonance imaging anomalies, the device comprising:

[0066] Memory containing executable program code;

[0067] A processor coupled to the memory;

[0068] The processor calls the executable program code stored in the memory to execute the individual magnetic resonance imaging anomaly analysis method disclosed in the first aspect of the present invention.

[0069] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute the individual magnetic resonance imaging anomaly analysis method disclosed in the first aspect of the present invention.

[0070] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0071] In this embodiment of the invention, for each analysis object, the target voxel information of the analysis object is obtained; based on the target voxel information and a preset voxel difference calculation model, the target difference score of the analysis object is calculated, whereby the target difference score represents the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object; it is determined whether the target difference score is greater than or equal to a preset difference score threshold, and if it is determined that the target difference score is greater than or equal to the preset difference score threshold, then the analysis object is identified as an abnormal object; based on the target voxel information of all abnormal objects, a target difference matrix is ​​determined; based on the non-negative matrix factorization algorithm and the target difference matrix, a target difference factor and a target weight matrix are determined, whereby the target difference factor represents the common characteristics of all abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each abnormal object. As can be seen, implementing this invention can calculate the target difference score of the analyzed object based on the obtained target voxel information and the preset voxel difference calculation model, thereby determining whether the analyzed object is an abnormal object. At the same time, since the target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of each analyzed object, it can improve the efficiency of abnormal object identification while improving the accuracy of abnormal object analysis. By determining the target difference matrix based on the target voxel information of all abnormal objects, and expanding and decomposing it according to the non-negative matrix factorization algorithm to obtain the target difference factor representing the common characteristics of all abnormal objects and the target weight matrix including the abnormal weight corresponding to each abnormal object, it can help improve the efficiency of individual magnetic resonance imaging abnormality analysis and help protect the individual's health in a timely manner. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a flowchart illustrating an individual magnetic resonance imaging anomaly analysis method disclosed in an embodiment of the present invention;

[0074] Figure 2 This is a schematic diagram of the structure of an individual magnetic resonance imaging anomaly analysis device disclosed in an embodiment of the present invention;

[0075] Figure 3 This is a schematic diagram of another individual magnetic resonance imaging anomaly analysis device disclosed in an embodiment of the present invention;

[0076] Figure 4 This is a schematic diagram of the structure of another individual magnetic resonance imaging anomaly analysis device disclosed in an embodiment of the present invention. Detailed Implementation

[0077] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0078] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0079] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0080] This invention discloses a method and apparatus for individual magnetic resonance imaging (MRI) anomaly analysis. Based on the acquired target voxel information and a preset voxel difference calculation model, it calculates the target difference score of the analyzed object, thereby determining whether the object is an anomaly. Furthermore, since the target difference score represents the deviation between the target voxel information and the normal voxel distribution range information of each analyzed object, it improves both the accuracy and efficiency of anomaly identification. By determining the target difference matrix based on the target voxel information of all anomalies, and using a non-negative matrix factorization algorithm, it expands and decomposes the matrix to obtain target difference factors representing the common characteristics of all anomalies and a target weight matrix including the anomaly weight corresponding to each anomaly. This improves the efficiency of individual MRI anomaly analysis and helps to promptly protect individual health. Detailed descriptions follow.

[0081] Example 1

[0082] Please see Figure 1 , Figure 1This is a flowchart illustrating an individual magnetic resonance imaging anomaly analysis method disclosed in an embodiment of the present invention. Figure 1 The described method for analyzing individual magnetic resonance imaging anomalies can be applied to magnetic resonance imaging (MRI) devices, as well as to intelligent devices related to MRI devices. These intelligent devices include, but are not limited to, one or more of medical devices, cloud devices, edge computing devices, relay devices, base station devices, urban management devices, and intelligent connected devices. This invention does not limit the scope of such applications. Figure 1 As shown, this method for analyzing anomalies in individual magnetic resonance imaging may include the following operations:

[0083] 101. For each analysis object, obtain the target voxel information of that analysis object.

[0084] In this embodiment of the invention, optionally, the target voxel information is determined based on the magnetic resonance brain structural images of each analysis object.

[0085] 102. Based on the target voxel information and the preset voxel difference calculation model, calculate the target difference score of the analysis object. The target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object.

[0086] In this embodiment of the invention, as an optional implementation, before calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model, the method may further include the following operations:

[0087] Obtain the physiological information of the subject of analysis, which includes at least one of the following: age, gender, height, skin color, and body fluids.

[0088] Based on physiological information and a pre-trained Gaussian process regression model, the predicted voxel information of the analysis object is calculated, and the operation of calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model is triggered. The predicted voxel information is the normal voxel distribution range information of the analysis object.

[0089] In this optional embodiment, the aforementioned physiological information may also include at least one of blood information, DNA information, gene information, pupil information, and fingerprint information.

[0090] As can be seen, implementing this optional embodiment can calculate the predicted voxel information of each analysis object based on the above-mentioned physiological information and the pre-trained Gaussian process regression model, which is beneficial to further calculate the target difference score of the analysis object based on the above-mentioned target voxel information and the preset voxel difference calculation model, thereby improving the accuracy and efficiency of identifying abnormal objects.

[0091] In this optional embodiment, the training process of the Gaussian process regression model described above can be obtained by training a pre-collected discovery set and a preset standard Gaussian process regression equation.

[0092] Further optionally, the aforementioned discovery set may include collecting images of mentally abnormal subjects and normal magnetic resonance brain structures of each mentally abnormal subject, wherein the mentally abnormal subjects exhibit at least one of the following phenomena: depression, schizophrenia, obsessive-compulsive disorder, etc.

[0093] Additional magnetic resonance brain structural images of other mentally ill individuals were downloaded as a validation set for the aforementioned Gaussian process regression model.

[0094] Alternatively, the Gaussian process regression model described above can also use magnetic resonance brain structural images of normal individuals as part of the discovery set and validation set.

[0095] Alternatively, the validation set may also include magnetic resonance brain images of subjects with other mental disorders such as schizophrenia, bipolar disorder, and attention deficit hyperactivity disorder, thereby improving the predictive robustness and model robustness of the Gaussian process regression model.

[0096] In this embodiment of the invention, as another optional implementation, the aforementioned preset voxel difference calculation model is as follows:

[0097]

[0098] Where Z represents the target difference score, VBM predicted VBM is used to represent prediction voxel information. true Used to represent target voxel information Used to represent the preset expected variance. This is used to represent the prediction variance, which is calculated based on the prediction voxel information of all analyzed objects.

[0099] As can be seen, implementing this optional embodiment can calculate the target difference score of the analyzed object through a specific preset voxel difference calculation model, thereby determining whether the analyzed object is an anomalous object, which can further improve the accuracy and efficiency of anomalous object determination.

[0100] 103. Determine whether the target difference score is greater than or equal to the preset difference score threshold.

[0101] 104. When the target difference score is determined to be greater than or equal to the preset difference score threshold, the analysis object is identified as an abnormal object.

[0102] 105. Determine the target difference matrix based on the target voxel information of all abnormal objects.

[0103] 106. Based on the nonnegative matrix factorization algorithm and the target difference matrix, determine the target difference factor and the target weight matrix. The target difference factor is used to represent the common characteristics of all abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each abnormal object.

[0104] In this embodiment of the invention, optionally, after obtaining the target difference factor and target weight matrix, this method can be directly applied to the individual magnetic resonance imaging anomaly analysis of new anomalous objects. Based on the target voxel information of the new anomalous object, its anomalous features can be directly matched. Specifically, the anomalous features may include its anomalous weights. Further optionally, the preset voxel difference calculation model may include at least one clustering sub-model, specifically a K-means clustering sub-model. Using the target difference score of the anomalous object and the target voxel information of the anomalous object, the anomalous subtype of the anomalous object can also be calculated. Specifically, the anomalous subtype is obtained based on the above-mentioned anomalous weight analysis before the above-mentioned anomalous weight output.

[0105] As can be seen, implementing the embodiments of the present invention can calculate the target difference score of the analyzed object based on the obtained target voxel information and the preset voxel difference calculation model, thereby determining whether the analyzed object is an abnormal object. At the same time, since the target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of each analyzed object, it can improve the efficiency of abnormal object identification while improving the accuracy of abnormal object analysis. By determining the target difference matrix based on the target voxel information of all abnormal objects, and expanding and decomposing the target difference factor representing the common characteristics of all abnormal objects and the target weight matrix including the abnormal weight corresponding to each abnormal object according to the non-negative matrix factorization algorithm, it is beneficial to improve the efficiency of individual magnetic resonance imaging abnormality analysis and to timely protect the individual's health.

[0106] In this embodiment of the invention, as another optional implementation, the above-mentioned nonnegative matrix factorization algorithm is as follows:

[0107] D = W * F + ε, W, F > 0

[0108] Where D represents the target difference matrix, W represents the target weight matrix, F represents the target difference factor, and ε represents the preset bias.

[0109] In this optional embodiment, the target difference matrix described above can be specifically (D, n*v), where n represents the number of abnormal objects and v represents the target voxel information of the abnormal objects.

[0110] Further optionally, the target difference factor mentioned above can be specifically (F, k*v), where k is used to represent the preset optimal number described below. Further optionally, the target weight matrix mentioned above can be specifically (W, n*k).

[0111] It is evident that implementing this optional embodiment can improve the accuracy and feasibility of the extended decomposition of the target difference matrix through a specific non-negative matrix factorization algorithm. By improving the accuracy of the target difference factors representing the common characteristics of all abnormal objects and the target weight matrix including the abnormal weights corresponding to each abnormal object, the extended decomposition can improve the efficiency of individual magnetic resonance imaging anomaly analysis and timely protect the individual's health. At the same time, it can improve the accuracy of grasping the common characteristics of abnormal objects and the abnormal weights of each abnormal object, and further improve the extended applicability of individual magnetic resonance imaging anomaly analysis.

[0112] In an optional embodiment, the aforementioned target difference factors have a corresponding preset optimal number, and the method for determining the preset optimal number may include the following operations:

[0113] Divide the preset set of exception objects into at least two preset subsets of exception objects.

[0114] In this optional embodiment, the specific method of dividing the preset abnormal object set into at least two preset abnormal object subsets can be random division.

[0115] For each preset subset of abnormal objects, obtain the preset difference matrix of that preset subset of abnormal objects. The preset difference matrix is ​​determined by the preset voxel information of each preset abnormal object in the preset subset of abnormal objects.

[0116] For each of the pre-optimal numbers in the pre-optimal number set, the pre-difference matrix of the pre-defined abnormal object subset is decomposed according to the pre-optimal number and the non-negative matrix factorization algorithm to obtain the first weight matrix and the first difference factor of the pre-defined abnormal object subset. The first weight matrix is ​​the product of the pre-optimal number and the number of pre-defined abnormal objects in the pre-defined abnormal object subset.

[0117] Based on the preset difference matrix, the first weight matrix, and the first difference factor of the preset subset of abnormal objects, the first decomposition error is calculated, and the first decomposition error is used to represent the decomposition error of the preset difference matrix.

[0118] In this optional embodiment, the formula for calculating the first decomposition error is optionally:

[0119] e11=D1-W11*F11

[0120] Wherein, e11 represents the first decomposition error, D1 represents the preset difference matrix of the preset subset of abnormal objects, W11 represents the first weight matrix, and F11 represents the first difference factor.

[0121] For each other preset abnormal object subset outside the preset abnormal object subset, the preset difference matrix of the other preset abnormal object subset is decomposed according to the pre-optimal number, the non-negative matrix factorization algorithm and the first difference factor, to obtain the second weight matrix of the other preset abnormal object subset.

[0122] The second decomposition error is calculated based on the preset difference matrix, the second weight matrix, and the first difference factor of the other preset abnormal object subset. The second decomposition error is used to represent the decomposition error of the preset difference matrix of the other preset abnormal object subset when the first difference factor is applied.

[0123] In this optional embodiment, the formula for calculating the second decomposition error is further optionally:

[0124] e12 = D12 - W12 * F11

[0125] Wherein, e12 represents the second decomposition error, D12 represents the preset difference matrix of the other preset abnormal object subset, W12 represents the second weight matrix, and F11 represents the first difference factor.

[0126] Based on all first decomposition errors and all second decomposition errors, calculate the target value of the pre-optimal number. The target value is used to represent the degree of deviation of the decomposition matrix corresponding to the pre-optimal number.

[0127] Based on the target calculated value, determine the preset optimal number.

[0128] In this optional embodiment, e11 and e12, e22 and e21 can be obtained similarly, where e22 is used to represent the first decomposition error of the preset difference matrix of the preset difference matrix of the preset abnormal object subset that is different from e11 in at least two preset abnormal object subsets, and e21 is used to represent the decomposition error of the preset difference matrix of the preset abnormal object subset of e22 when the first difference factor is applied to the preset difference matrix of the other preset abnormal object subsets.

[0129] In this optional embodiment, the specific calculation formula for calculating the target calculated value of the pre-optimal number based on all first decomposition errors and all second decomposition errors can be:

[0130] GE = e12 - e11 + e21 - e22

[0131] Here, GE is used to represent the target calculated value mentioned above.

[0132] It is evident that implementing this optional embodiment can improve the accuracy of determining the preset optimal number, which is beneficial to improving the accuracy of determining the target difference factor, improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, and further improving the scalability of individual magnetic resonance anomaly analysis.

[0133] In this optional embodiment, as an optional implementation method, the method for determining the preset optimal number before determining the preset optimal number based on the target calculated value may further include the following operations:

[0134] Determine whether the number of partitions of the preset abnormal object set is greater than or equal to the preset number threshold. If it is determined that the number of partitions of the preset abnormal object set is greater than or equal to the preset number threshold, then trigger the operation of determining the preset optimal number based on the target calculated value.

[0135] When it is determined that the number of times the preset abnormal object set is divided is less than the preset number of times threshold, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered.

[0136] In this optional embodiment, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered by the above-mentioned method of determining the preset optimal number, which involves dividing the preset abnormal object set into at least two preset abnormal object subsets and subsequent steps, until it is determined that the number of times the preset abnormal object set is divided is greater than or equal to the preset number threshold. Then, the operation of determining the preset optimal number based on the target calculated value is triggered.

[0137] It is evident that implementing this optional embodiment can provide a method for multiple verifications, thereby further improving the accuracy of determining the target calculated value, which is conducive to further improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, and is conducive to further improving the scalability of individual magnetic resonance anomaly analysis.

[0138] In this optional embodiment, as another optional implementation, determining the preset optimal number based on the target calculated value may include the following operations:

[0139] The average of all calculated target values ​​for the same preliminary optimal number is obtained by averaging all calculated target values ​​for each preliminary optimal number.

[0140] Among all the average target calculation values, the minimum number of the preliminary optimal values ​​is determined as the preset optimal number.

[0141] It is evident that implementing this optional embodiment can provide a specific method for determining the preset optimal number, thereby further improving the accuracy of determining the preset optimal number. This is beneficial for further improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, further improving the scalability of individual magnetic resonance anomaly analysis, improving the efficiency of individual magnetic resonance image anomaly analysis, and protecting the individual's health.

[0142] Example 2

[0143] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an individual magnetic resonance imaging anomaly analysis device disclosed in an embodiment of the present invention. Figure 2 The described individual magnetic resonance imaging anomaly analysis device can be applied to magnetic resonance imaging equipment, or to intelligent devices related to magnetic resonance imaging equipment. These intelligent devices include, but are not limited to, one or more of medical devices, cloud devices, edge computing devices, relay devices, base station devices, urban management devices, and intelligent connected devices. This invention does not limit the scope of the application. Figure 2 As shown, this is the device for analyzing individual magnetic resonance imaging anomalies:

[0144] The acquisition module 201 is used to acquire the target voxel information of each analysis object.

[0145] The first calculation module 202 is used to calculate the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model. The target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object.

[0146] The judgment module 203 is used to determine whether the target difference score is greater than or equal to the preset difference score threshold.

[0147] The determination module 204 is used to determine the analysis object as an abnormal object when the judgment module 203 determines that the target difference score is greater than or equal to the preset difference score threshold.

[0148] The determination module 204 is also used to determine the target difference matrix based on the target voxel information of all abnormal objects.

[0149] The determination module 204 is also used to determine the target difference factor and the target weight matrix based on the non-negative matrix factorization algorithm and the target difference matrix. The target difference factor is used to represent the common features of all abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each abnormal object.

[0150] As can be seen, implementing the embodiments of the present invention can calculate the target difference score of the analyzed object based on the obtained target voxel information and the preset voxel difference calculation model, thereby determining whether the analyzed object is an abnormal object. At the same time, since the target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of each analyzed object, it can improve the efficiency of abnormal object identification while improving the accuracy of abnormal object analysis. By determining the target difference matrix based on the target voxel information of all abnormal objects, and expanding and decomposing the target difference factor representing the common characteristics of all abnormal objects and the target weight matrix including the abnormal weight corresponding to each abnormal object according to the non-negative matrix factorization algorithm, it is beneficial to improve the efficiency of individual magnetic resonance imaging abnormality analysis and to timely protect the individual's health.

[0151] In this embodiment of the invention, as an optional implementation, the acquisition module 201 is further configured to acquire the physiological information of the analysis object before the first calculation module calculates the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model. The physiological information includes at least one of age information, gender information, height information, skin color information, and body fluid information.

[0152] And, such as Figure 3 As shown, the device may further include:

[0153] The second calculation module 205 is used to calculate the predicted voxel information of the analysis object based on physiological information and a pre-trained Gaussian process regression model, and to trigger the first calculation module 202 to perform the operation of calculating the target difference score of the analysis object based on the target voxel information and a preset voxel difference calculation model. The predicted voxel information is the normal voxel distribution range information of the analysis object.

[0154] As can be seen, implementing this optional embodiment can calculate the predicted voxel information of each analysis object based on the above-mentioned physiological information and the pre-trained Gaussian process regression model, which is beneficial to further calculate the target difference score of the analysis object based on the above-mentioned target voxel information and the preset voxel difference calculation model, thereby improving the accuracy and efficiency of identifying abnormal objects.

[0155] In this embodiment of the invention, as another optional implementation, the aforementioned preset voxel difference calculation model is as follows:

[0156]

[0157] Where Z represents the target difference score, VBM predicted VBM is used to represent prediction voxel information. true Used to represent target voxel information Used to represent the preset expected variance. This is used to represent the prediction variance, which is calculated based on the prediction voxel information of all analyzed objects.

[0158] As can be seen, implementing this optional embodiment can calculate the target difference score of the analyzed object through a specific preset voxel difference calculation model, thereby determining whether the analyzed object is an anomalous object, which can further improve the accuracy and efficiency of anomalous object determination.

[0159] In this embodiment of the invention, as another optional implementation, the above-mentioned nonnegative matrix factorization algorithm is as follows:

[0160] D = W * F + ε, W, F > 0

[0161] Where D represents the target difference matrix, W represents the target weight matrix, F represents the target difference factor, and ε represents the preset bias.

[0162] It is evident that implementing this optional embodiment can improve the accuracy and feasibility of the extended decomposition of the target difference matrix through a specific non-negative matrix factorization algorithm. By improving the accuracy of the target difference factors representing the common characteristics of all abnormal objects and the target weight matrix including the abnormal weights corresponding to each abnormal object, the extended decomposition can improve the efficiency of individual magnetic resonance imaging anomaly analysis and timely protect the individual's health. At the same time, it can improve the accuracy of grasping the common characteristics of abnormal objects and the abnormal weights of each abnormal object, and further improve the extended applicability of individual magnetic resonance imaging anomaly analysis.

[0163] In an optional embodiment, the aforementioned target difference factor has a corresponding preset optimal number, and the method for determining the preset optimal number may include the following operations:

[0164] Divide the preset abnormal object set into at least two preset abnormal object subsets.

[0165] For each preset subset of abnormal objects, obtain the preset difference matrix of that preset subset of abnormal objects. The preset difference matrix is ​​determined by the preset voxel information of each preset abnormal object in the preset subset of abnormal objects.

[0166] For each of the pre-optimal numbers in the pre-optimal number set, the pre-difference matrix of the pre-defined abnormal object subset is decomposed according to the pre-optimal number and the non-negative matrix factorization algorithm to obtain the first weight matrix and the first difference factor of the pre-defined abnormal object subset. The first weight matrix is ​​the product of the pre-optimal number and the number of pre-defined abnormal objects in the pre-defined abnormal object subset.

[0167] Based on the preset difference matrix, the first weight matrix, and the first difference factor of the preset subset of abnormal objects, the first decomposition error is calculated, and the first decomposition error is used to represent the decomposition error of the preset difference matrix.

[0168] For each other preset abnormal object subset outside the preset abnormal object subset, the preset difference matrix of the other preset abnormal object subset is decomposed according to the pre-optimal number, the non-negative matrix factorization algorithm and the first difference factor, to obtain the second weight matrix of the other preset abnormal object subset.

[0169] The second decomposition error is calculated based on the preset difference matrix, the second weight matrix, and the first difference factor of the other preset abnormal object subset. The second decomposition error is used to represent the decomposition error of the preset difference matrix of the other preset abnormal object subset when the first difference factor is applied.

[0170] Based on all first decomposition errors and all second decomposition errors, calculate the target value of the pre-optimal number. The target value is used to represent the degree of deviation of the decomposition matrix corresponding to the pre-optimal number.

[0171] Based on the target calculated value, determine the preset optimal number.

[0172] It is evident that implementing this optional embodiment can improve the accuracy of determining the preset optimal number, which is beneficial to improving the accuracy of determining the target difference factor, improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, and further improving the scalability of individual magnetic resonance anomaly analysis.

[0173] In this optional embodiment, as an optional implementation method, the method for determining the preset optimal number before determining the preset optimal number based on the target calculated value may further include the following operations:

[0174] Determine whether the number of partitions of the preset abnormal object set is greater than or equal to the preset number threshold. If it is determined that the number of partitions of the preset abnormal object set is greater than or equal to the preset number threshold, then trigger the operation of determining the preset optimal number based on the target calculated value.

[0175] When it is determined that the number of times the preset abnormal object set is divided is less than the preset number of times threshold, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered.

[0176] It is evident that implementing this optional embodiment can provide a method for multiple verifications, thereby further improving the accuracy of determining the target calculated value, which is conducive to further improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, and is conducive to further improving the scalability of individual magnetic resonance anomaly analysis.

[0177] In this optional embodiment, as another optional implementation, determining the preset optimal number based on the target calculated value may include the following operations:

[0178] The average target calculation value for each pre-optimal number is obtained by averaging all the target calculation values ​​for the same pre-optimal number.

[0179] Among all the average target calculation values, the minimum number of the preliminary optimal values ​​is determined as the preset optimal number.

[0180] It is evident that implementing this optional embodiment can provide a specific method for determining the preset optimal number, thereby further improving the accuracy of determining the preset optimal number. This is beneficial for further improving the accuracy of grasping the common characteristics of abnormal objects and the abnormal weight of each abnormal object, further improving the scalability of individual magnetic resonance anomaly analysis, improving the efficiency of individual magnetic resonance image anomaly analysis, and protecting the individual's health.

[0181] Example 3

[0182] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of another individual magnetic resonance imaging anomaly analysis device disclosed in an embodiment of the present invention. Figure 4 As shown, the individual magnetic resonance imaging anomaly analysis device may include:

[0183] Memory 301 that stores executable program code.

[0184] Processor 302 coupled to memory 301.

[0185] The processor 302 calls the executable program code stored in the memory 301 to execute the steps in the individual magnetic resonance imaging anomaly analysis method described in Embodiment 1 of the present invention.

[0186] Example 4

[0187] This invention discloses a computer storage medium storing computer instructions, which, when invoked, execute the steps in the individual magnetic resonance imaging anomaly analysis method described in Embodiment 1 of this invention.

[0188] Example 5

[0189] This invention discloses a computer program product, which includes a non-transitory computer storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the individual magnetic resonance imaging anomaly analysis method described in Embodiment 1.

[0190] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0191] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0192] Finally, it should be noted that the individual magnetic resonance imaging anomaly analysis method and apparatus disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention, and are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing anomalies in individual magnetic resonance imaging, characterized in that, The method includes: For each analysis object, the target voxel information of the analysis object is obtained; based on the target voxel information and a preset voxel difference calculation model, the target difference score of the analysis object is calculated, whereby the target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object; it is determined whether the target difference score is greater than or equal to a preset difference score threshold; if it is determined that the target difference score is greater than or equal to the preset difference score threshold, then the analysis object is identified as an abnormal object. Based on the target voxel information of all the aforementioned abnormal objects, determine the target difference matrix; Based on the nonnegative matrix factorization algorithm and the target difference matrix, a target difference factor and a target weight matrix are determined. The target difference factor is used to represent the common features of all the abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each abnormal object. Furthermore, before calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model, the method further includes: Obtain the physiological information of the analysis subject, wherein the physiological information includes at least one of age information, gender information, height information, skin color information, and body fluid information; Based on the physiological information and the pre-trained Gaussian process regression model, the predicted voxel information of the analysis object is calculated, and the operation of calculating the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model is triggered. The predicted voxel information is the normal voxel distribution range information of the analysis object. Furthermore, the preset voxel difference calculation model is as follows: in, Used to represent the target difference score Used to represent the predicted voxel information. Used to represent the target voxel information. Used to represent the preset expected variance. This is used to represent the prediction variance, which is calculated based on the prediction voxel information of all the analyzed objects.

2. The method for analyzing individual magnetic resonance imaging anomalies according to claim 1, characterized in that, The nonnegative matrix factorization algorithm is as follows: ， in, Used to represent the target difference matrix Used to represent the target weight matrix Used to represent the target difference factor. Used to indicate a preset bias.

3. The method for analyzing individual magnetic resonance imaging anomalies according to claim 2, characterized in that, The target difference factor has a corresponding preset optimal number, and the preset optimal number is determined by the following methods: Divide the preset set of exception objects into at least two preset subsets of exception objects; For each of the preset abnormal object subsets, a preset difference matrix is ​​obtained for the preset abnormal object subset. The preset difference matrix is ​​determined by the preset voxel information of each preset abnormal object in the preset abnormal object subset. For each of the pre-optimal numbers in the pre-optimal number set, the pre-optimal difference matrix of the pre-optimal abnormal object subset is decomposed according to the pre-optimal number and the non-negative matrix factorization algorithm to obtain the first weight matrix and the first difference factor of the pre-optimal abnormal object subset, wherein the first weight matrix is ​​the product of the pre-optimal number and the number of the pre-optimal abnormal objects in the pre-optimal abnormal object subset. Based on the preset difference matrix, the first weight matrix, and the first difference factor of the preset subset of abnormal objects, a first decomposition error is calculated, and the first decomposition error is used to represent the decomposition error of the preset difference matrix. For each other preset abnormal object subset outside of the preset abnormal object subset, the preset difference matrix of the other preset abnormal object subset is decomposed according to the pre-optimal number, the non-negative matrix factorization algorithm and the first difference factor to obtain the second weight matrix of the other preset abnormal object subset; A second decomposition error is calculated based on the preset difference matrix, the second weight matrix, and the first difference factor of the other preset abnormal object subset. The second decomposition error is used to represent the decomposition error of the preset difference matrix of the other preset abnormal object subset when the first difference factor is applied. Based on all the first decomposition errors and all the second decomposition errors, calculate the target calculated value of the pre-optimal number, which is used to represent the degree of deviation of the decomposition matrix corresponding to the pre-optimal number; The preset optimal number is determined based on the target calculated value.

4. The method for analyzing individual magnetic resonance imaging anomalies according to claim 3, characterized in that, Before determining the preset optimal number based on the target calculated value, the method for determining the preset optimal number further includes: Determine whether the number of times the preset abnormal object set is divided is greater than or equal to a preset number threshold. If it is determined that the number of times the preset abnormal object set is divided is greater than or equal to the preset number threshold, then trigger the operation of determining the preset optimal number based on the target calculated value. When it is determined that the number of times the preset abnormal object set is divided is less than the preset number threshold, the operation of dividing the preset abnormal object set into at least two preset abnormal object subsets is triggered.

5. The method for analyzing individual magnetic resonance imaging anomalies according to claim 3 or 4, characterized in that, Determining the preset optimal number based on the target calculated value includes: The average target calculated value for each of the same pre-optimal numbers is obtained by averaging all the target calculated values ​​of the same pre-optimal number. Among all the average target calculation values, the preliminary optimal number of the smallest average target calculation value is determined as the preset optimal number.

6. A device for analyzing individual magnetic resonance imaging anomalies, characterized in that, The apparatus is used to perform the individual magnetic resonance imaging anomaly analysis method as described in any one of claims 1-5, and the apparatus comprises: The acquisition module is used to acquire the target voxel information of each analysis object; The first calculation module is used to calculate the target difference score of the analysis object based on the target voxel information and the preset voxel difference calculation model. The target difference score is used to represent the degree of deviation between the target voxel information and the normal voxel distribution range information of the analysis object. The judgment module is used to determine whether the target difference score is greater than or equal to a preset difference score threshold. The determination module is used to determine the analysis object as an abnormal object when the judgment module determines that the target difference score is greater than or equal to the preset difference score threshold. The determining module is further configured to determine a target difference matrix based on the target voxel information of all the abnormal objects; The determining module is further configured to determine a target difference factor and a target weight matrix based on a non-negative matrix factorization algorithm and the target difference matrix. The target difference factor is used to represent the common features of all the abnormal objects, and the target weight matrix includes the abnormal weight corresponding to each of the abnormal objects.

7. A device for analyzing individual magnetic resonance imaging anomalies, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the individual magnetic resonance imaging anomaly analysis method as described in any one of claims 1-5.

8. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the individual magnetic resonance imaging anomaly analysis method as described in any one of claims 1-5.

Images