Data processing method, system and electronic device for medical images

Abstract

The application provides a medical image-oriented data processing method, system and electronic equipment, and relates to the technical field of medical images. The method comprises the following steps: training a data analysis decision model by using a training database; generating a target data analysis execution flow by using the trained data analysis decision model according to data analysis requirements; the data analysis requirements comprise a processing target for medical image data; the target data analysis execution flow comprises a plurality of medical image data analysis links, related parameters of each link and a corresponding execution sequence; and the actual medical image data obtained is processed based on the target data analysis execution flow to generate a final processing result, so as to realize automatic analysis and processing of medical image data. The above method alleviates the problems of great difficulty and low efficiency in medical image data processing, guarantees the quality of data processing and reduces the workload of data processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical imaging technology, in particular to a data processing method and system for medical imaging and an electronic device. BACKGROUND

[0002] With the rapid development of artificial intelligence, big data, and neuroimaging technology, people are increasingly interested in the mechanisms of brain development, aging, and neurological diseases. Furthermore, there is a need for a more systematic understanding of the components of cognition, emotion, and disease and their relationships, including brain computing, which involves multi-modal neuroimaging data analysis.

[0003] Depending on different research objectives, data processing for neuroimaging involves numerous steps, such as preprocessing, feature extraction, statistical analysis, and modeling. In these data processing steps, numerous software packages are generally involved. According to incomplete statistics, there are at least several hundred neuroimaging-related software packages. To correctly and effectively apply these software for image processing, data processing personnel need to have a deep understanding of the algorithms, parameters, result formats and meanings, design methods, etc. contained in these software. For neuroscientists, the threshold is high and the difficulty is great. Moreover, each neuroimaging data processing step often involves many sub-steps, and different software needs to be combined to form a data processing flow. Therefore, it is necessary to have a deep understanding of the matching between software, which is also a difficult and time-consuming problem. Similarly, the above problems also exist in the data processing of other medical images besides neuroimaging. That is, the existing data processing scheme for medical images has the problems of high processing difficulty and low processing efficiency. SUMMARY

[0004] The present application aims to provide a data processing method and system for medical imaging and an electronic device to alleviate the technical problems of high processing difficulty and low processing efficiency in the prior art.

[0005] In a first aspect, the present application provides a data processing method for medical imaging, comprising: training a data analysis decision model using a training database; the training database is generated based on a medical imaging data processing knowledge base; the medical imaging data processing knowledge base includes a data processing target and a data analysis step;

[0006] According to the data analysis requirements, a target data analysis execution flow is generated using the trained data analysis decision model; the data analysis requirements include the processing target of the medical imaging data; the target data analysis execution flow includes several medical imaging data analysis steps, related parameters of each step, and the corresponding order of execution;

[0007] Based on the above target data analysis execution flow, the acquired actual medical image data is processed to generate a final processing result, so as to realize automatic analysis and processing of medical image data.

[0008] In some possible implementation manners, the method further includes: constructing a medical image data processing knowledge base; a framework of the medical image data processing knowledge base includes a plurality of top-level categories and a plurality of sub-level categories; each current sub-level category is obtained by subdividing a previous level category of the current sub-level category; and the top-level categories include: data processing target, data analysis link, object data, data storage, server, file and task.

[0009] In some possible implementation manners, the method further includes: constructing a training database based on the medical image data processing knowledge base; the training database includes: data analysis requirement and data analysis execution flow; the data analysis requirement is a coded character sequence expression of a processing target of medical image data; and the data analysis execution flow includes: a character sequence expression of a component corresponding to at least one medical image data analysis link; each component includes at least one module, and the module is used to represent a specific algorithm of a corresponding link.

[0010] In some possible implementation manners, the method further includes: constructing a computing facility decision model; the computing facility decision model includes: corresponding rule and machine learning model; generating a corresponding rule between user data processing requirement and data storage and server according to the medical image data processing knowledge base; training a machine learning model based on relevant data in a server running process; the machine learning model includes: data storage decision model and server decision model; an input of the data storage decision model is the user data processing requirement, and an output is a data storage target; an input of the server decision model is the user data processing requirement, and an output is a server target; determining a target computing facility based on the computing facility decision model; and the target computing facility is used to execute a data analysis link of the target data analysis execution flow.

[0011] In some possible implementation manners, the step of determining a target computing facility based on the computing facility decision model includes: determining a data storage target based on the data storage decision model; the data storage target includes: storage type, capacity and target storage device; the target storage device is used to store initial data to be processed and data generated by each data analysis link; determining a server target based on the server decision model; the server target includes: target server type and target server; and the target server is used to perform calculation of each data analysis link.

[0012] In some possible implementation manners, the method further includes: generating a processing strategy related to the target data analysis execution flow according to a data analysis requirement by using the trained data analysis decision model; the processing strategy includes a data calculation strategy and a data analysis strategy; the data analysis strategy is used to represent a manner of executing each analysis link in the target data analysis execution flow; the data analysis strategy includes: single analysis, voting analysis, and multi-dimensional analysis; the data calculation strategy is used to represent a data calculation manner in the analysis link; and the data calculation manner includes: CPU calculation, GPU calculation, and hybrid calculation.

[0013] In some possible implementation manners, the method further includes: filtering the data calculation strategy and the data analysis strategy respectively based on a predefined rule between the data storage target, the server target, and the processing strategy, to generate filtered data calculation strategy and data analysis strategy.

[0014] In some possible implementation manners, the medical image data analysis link includes: any of preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, and processing strategy; processing the acquired actual medical image data based on the target data analysis execution flow includes: sequentially executing the medical image data analysis link on the acquired actual medical image data in the corresponding order, and generating a processing result of each link; calculating a data quality control index for the processing result of the data analysis link; the data quality control index is used to evaluate the usability of the data; and generating a result evaluation parameter for the final processing result.

[0015] In a second aspect, an embodiment of the present application provides a data processing system for medical images, and the system includes:

[0016] a model training module configured to train a data analysis decision model by using a training database; the training database is generated based on a medical image data processing knowledge base; the medical image data processing knowledge base includes: a data processing target and a data analysis link;

[0017] an execution flow generation module configured to generate a target data analysis execution flow by using the trained data analysis decision model according to a data analysis requirement; the data analysis requirement includes: a processing target for medical image data; and the target data analysis execution flow includes a plurality of medical image data analysis links, related parameters of each link, and a corresponding execution order;

[0018] a processing result generation module configured to process acquired actual medical image data based on the target data analysis execution flow, and generate a final processing result, so as to realize automatic analysis and processing of the medical image data.

[0019] In a third aspect, an electronic device is provided, which comprises a memory and a processor, the memory storing a computer program executable on the processor, and the processor executes the computer program to implement the steps of the method of any one of the first aspect.

[0020] The present application provides a medical image-oriented data processing method and system, and an electronic device. The method comprises: training a data analysis decision model using a training database; generating a target data analysis execution flow using the trained data analysis decision model according to a data analysis requirement; the data analysis requirement comprises a processing target for medical image data; the target data analysis execution flow comprises a plurality of medical image data analysis links, related parameters of each link, and a corresponding execution order; and processing actual medical image data obtained based on the target data analysis execution flow to generate a final processing result, so as to realize automatic analysis and processing of medical image data. The above method alleviates the problems of high difficulty and low efficiency in medical image data processing, ensures the quality of data processing, and reduces the workload of data processing. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the following will briefly introduce the drawings needed in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0022] Figure 1 A flowchart of a medical image-oriented data processing method provided by an embodiment of the present application is shown in the figure.

[0023] Figure 2 A flowchart of a medical image-oriented data processing method provided by an embodiment of the present application is shown in the figure.

[0024] Figure 3 A structure diagram of a medical image-oriented data processing system provided by an embodiment of the present application is shown in the figure.

[0025] Figure 4 A structure diagram of an electronic device provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0026] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0028] It should be noted that similar reference numerals and letters refer to similar items throughout the drawings, and therefore, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings. Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0029] With the rapid development of artificial intelligence, big data, neural imaging technology, and the like, people are increasingly in-depth in the mechanism research of brain development, aging, and nervous system diseases. Further, it is necessary to more systematically understand the constituent elements of cognition, emotion, and disease and the relationship therebetween, and brain computing including multi-modal neural image data analysis will be a core driving force. In combination with different research objectives, data processing of neural images includes numerous links with different objectives, such as preprocessing, feature extraction, statistical analysis, modeling, and the like. In these data processing links, a large number of software packages are generally involved, such as FreeSurfer, FSL (FMRIB software library), SPM (Statistical parametric mapping), AFNI (Analysis functional neuroImage), tensorflow, pytorch, and the like. According to incomplete statistics, there are at least several hundred kinds of neural image related software.

[0030] To address this, data processing personnel need a deep understanding of the algorithms, parameters, result formats and meanings, and design methods of these software programs in order to correctly and effectively apply them for image processing. This presents a significant barrier to entry and considerable difficulty for neuroscience researchers. Furthermore, each neuroimaging data processing stage often involves numerous sub-stages, requiring the combination of different software programs to form a data processing flow. This necessitates a thorough understanding of the compatibility between software programs, which is also a challenging and time-consuming issue. Similarly, the same problems exist in the processing of other medical images besides neuroimaging. In other words, existing data processing solutions for medical images suffer from high processing difficulty and low efficiency. Although some target-specific neuroimaging data processing flow software exists, users still need a deep understanding of the software's internal logic to configure parameters and select methods via the command-line interface before processing. Therefore, the aforementioned problems remain unresolved.

[0031] Based on this, embodiments of the present invention provide a data processing method, system, and electronic device for medical images to alleviate the technical problems of high processing difficulty and low processing efficiency in the prior art.

[0032] To facilitate understanding of this embodiment, we will first describe one aspect of the present invention.

[0033] For a detailed introduction to data processing methods for medical images, please refer to [link / reference]. Figure 1 The diagram shows a flowchart of a data processing method for medical images. This method can be executed by electronic devices and mainly includes...

[0034] Next steps S110 to S130:

[0035] S110: Train the data analysis decision model using a training database; the training database is generated based on a medical image data processing knowledge base; the medical image data processing knowledge base includes: data processing objectives.

[0036] The process includes target setting and data analysis.

[0037] In this embodiment, a medical image data processing knowledge base is first constructed. (Medical image data processing)

[0038] The knowledge base framework includes multiple top-level categories and multiple sub-level categories; each current sub-level category is derived from the previous level category; the top-level categories include: data processing objectives, data analysis stages, object data, data storage, servers, files, and tasks.

[0039] The framework of the knowledge base is a multi-level and multi-link category system. As a specific example, the top-level categories include processing target, object data, data storage, server, computing process (i.e., data analysis link), file, task, and the like. According to the knowledge base framework, the evidence-based literature approach is adopted to record the processing target, object data, data storage, server, computing process, and the like that have clear and high-quality evidence support, thereby establishing the knowledge base.

[0040] Then, in the present embodiment, a training database is constructed based on the medical image data processing knowledge base; the training database includes: data analysis requirements and data analysis execution flow; the data analysis requirements are coded literal sequence expression of the processing target of the medical image data; the data analysis execution flow includes: literal sequence expression of components corresponding to at least one medical image data analysis link; each component includes at least one module, and the module is used to represent the specific algorithm of the corresponding link.

[0041] As a specific example, the data analysis requirements mainly include disease, focus point-constitute a coded literal sequence after standardization, and calculation efficiency and performance requirements-according to the pre-determined rules, the efficiency and performance requirements are corresponded to the options of extremely high, high, high, medium, and the like-and constitute a coded literal sequence. The data analysis execution flow is composed of one or more components, and the component mainly refers to preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, processing strategy, and the like; the component is composed of one or more modules, and the module refers to a specific algorithm applied to a specific calculation, such as head motion correction algorithm (which can be multiple) in preprocessing; from this perspective, the data analysis execution flow is a directed graph, the node is the module, and the edge is the sequential relationship. The training data is divided into training, verification, and test sets in the ratio of 7:1.5:1.5.

[0042] S120: generating target data analysis execution flow according to data analysis requirements and using the trained data analysis decision model; the data analysis requirements include: processing target for medical image data; the target data analysis execution flow includes several medical image data analysis links, related parameters of each link, and corresponding execution order;

[0043] In the present embodiment, the medical image data analysis link can include any of the following: preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, and processing strategy. The target data analysis execution flow can also include corresponding: software tools, algorithms, parameter values, and the like.

[0044] In the present embodiment, the medical image data analysis link can include any of the following: preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, and processing strategy. The target data analysis execution flow can also include corresponding: software tools, algorithms, parameter values, and the like.

[0045] S130: Based on the target data analysis execution flow, the acquired actual medical image data is processed to generate a final processing result, so as to realize automatic analysis and processing of the medical image data.

[0046] In an embodiment, the method further comprises:

[0047] S21: A computing facility decision model is constructed; the computing facility decision model comprises: corresponding rules and a machine learning model;

[0048] S22: According to the medical image data processing knowledge base, corresponding rules between user data processing requirements and data storage and servers are generated;

[0049] S23: Based on relevant data in the server running process, a machine learning model is trained and generated;

[0050] S24: Based on the computing facility decision model, a target computing facility is determined; the target computing facility is used to execute the data analysis link of the target data analysis execution flow.

[0051] In this embodiment, the step of training and generating a machine learning model based on relevant data in the server running process in S23 specifically comprises: based on relevant data in the server running process, a data storage decision model and a server decision model are respectively constructed by using a machine learning method. Wherein, the input of the data storage decision model is user data processing requirements (such as: data processing purpose, data itself, data processing expectation, etc.), and the output is a data storage target; the input of the server decision model is user data processing requirements (such as: data processing purpose, data itself, data processing expectation, etc.), and the output is a server target.

[0052] In this embodiment, the step of determining a target computing facility based on a computing facility decision model in S24 specifically comprises: (S231) determining a data storage target based on a data storage decision model; (S232) determining a server target based on a server decision model.

[0053] Wherein, the data storage target comprises: storage type, capacity and target storage device; the target storage device is used to store initial data to be processed and data generated by each data analysis link; the server target comprises: target server type and target server; the target server is used to execute the calculation of each data analysis link.

[0054] In this embodiment, the data stored in the target storage device is subjected to format checking, and the data not conforming to the agreed format is subjected to format conversion.

[0055] That is, according to the determined storage type, capacity and specific storage selection, the data specified in the user information configuration is transmitted to the storage. For the data object transmitted to the destination location, it is checked according to the pre-defined data organization method (such as bids), if the data does not conform to the agreed format, the data object is converted to the agreed format, and the data is saved.

[0056] In an embodiment, the method further comprises: (S233) generating a processing strategy related to the target data analysis execution flow according to the data analysis requirements and using the trained data analysis decision model; the processing strategy includes a data calculation strategy and a data analysis strategy; wherein the data analysis strategy is used to represent the way of executing each analysis link in the target data analysis execution flow; the data analysis strategy includes: single analysis, voting analysis, multi-dimensional analysis, etc.; the data calculation strategy includes the data calculation method in the analysis link; the data calculation method includes: CPU calculation, GPU calculation, hybrid calculation.

[0057] As a specific example, according to the different analysis strategies, the data analysis process includes the execution analysis process in different ways such as single analysis, voting analysis, multi-dimensional analysis (weighted), etc., and according to the different calculation strategies, the data analysis process can be divided into different calculation methods such as cpu calculation, gpu calculation, hybrid calculation, etc. The above strategies are automatically recommended (including referring to literature or according to default settings - such as single analysis, cpu calculation, etc.) in the process of generating data analysis execution flow and filtered by rules between storage servers and strategies (determined in advance, used to determine the analysis and calculation strategies in a specific storage and server environment).

[0058] In an embodiment, the method further comprises: filtering the data calculation strategy and the data analysis strategy respectively based on the pre-defined rules between the data storage target, the server target and the processing strategy, to generate the filtered data calculation strategy and the data analysis strategy.

[0059] In the above S130, based on the target data analysis execution flow, the acquired actual medical image data is processed, including: S31 sequentially executing the medical image data analysis links on the acquired actual medical image data according to the corresponding order, and generating the processing result of each link; S32 calculating a data quality control index for the processing result of the data analysis link; the data quality control index is used to evaluate the usability of the data; S33 generating a result evaluation parameter for the final processing result. The final processing result is evaluated from multiple dimensions through different result evaluation parameters; the usability of the data is evaluated by calculating the data quality control index.

[0060] The result evaluation parameter is mainly to objectively evaluate the execution result of data analysis, and the data quality control is mainly to evaluate the whole process of data analysis (along with the data analysis process) so as to provide the process quality basis for reasonably regarding the result. As a specific example, the evaluation parameters can include: the significance level in statistical analysis, the correlation coefficient size, the regression coefficient size, the accuracy of the machine learning model, the recall rate, F1, ROC, AUC, etc.

[0061] The result of data quality control is a dimensionless number (for each data object, etc.), called data quality control index (0-100), which is divided into several ranges according to the size of the index: less than 60, the data is unusable (the data in this state will not participate in the data analysis process), 60-80, the data is usable, and greater than 80, the data is better.

[0062] The calculation process of the data quality control index is, for example, for structural images, first, the feature data of the image data is extracted, including parameter difference (such as repetition time), signal-to-noise ratio, difference between structural images and templates, brain tissue symmetry, gray matter outside the template, and difference between different tool calculation results, etc. After standardization and normalization of these data (standardization and normalization according to the pre-determined gold standard structural image and its numerical value), the data is weighted (the weight is determined by the Delphi method in advance) and normalized to 0-100.

[0063] The calculation result of data analysis, result evaluation, data quality control index, etc. are saved to the location specified by the user in the format (such as bids) mode determined in advance.

[0064] The embodiment of the application provides an intelligent data processing method for medical images. The method combines the data characteristics of medical images (such as neural images), realizes the automatic analysis and processing process of data through the steps of knowledge base construction, information configuration, calculation facility determination, data acquisition, data analysis, result evaluation and result acquisition, reduces the difficulty of data processing, improves the efficiency of data processing, guarantees the quality of data processing, and reduces the workload of data processing. As a specific example, refer to Figure 2 As shown in the figure, the method can specifically include the following steps 1 to 7.

[0065] Step 1, knowledge base construction

[0066] To support intelligent data processing computing for neuroimaging, a neuroimaging intelligent data processing knowledge base framework is first constructed, which is constructed by using a multi-level multi-link method. The category relationship of the knowledge base framework is described as follows: the top-level categories include: processing target, object data, data storage, server, computing process (i.e. data analysis link), file, task, etc. The secondary categories are obtained by subdividing the top-level categories, for example: the categories under the processing target include: disease, focus point, efficiency, performance, etc.; the categories under the object data include: original, conversion, intermediate result, final result, etc.; the categories under the data storage include: physical storage, logical storage, etc.; the categories under the server include: virtual machine, physical server; the categories under the computing process include: preprocessing, feature extraction, statistical analysis, machine learning modeling (shallow learning), deep learning modeling, visualization, data quality control, result annotation, auxiliary tool, processing strategy, etc.; the categories under the file include: configuration file, data file, log file, etc.; the categories under the task include: file transmission task, data processing task, etc.

[0067] The third-level categories are subdivided from the second-level categories, such as: the categories under disease can include: dementia, autism, etc.; the categories under focus can include: cognition, social cognition, emotion, quality of life, body function, etc.; the categories under efficiency can include: operation time, space occupation, etc.; the categories under performance can include: accuracy, recall rate, significance level, etc.; the categories under raw data include: demographic characteristics, group, mode, field strength, sequence, additional features, related data, etc.; the categories under conversion / intermediate results include: steps, input, algorithm, results, etc.; the categories under final results include: statistical analysis results (including subjects / group, algorithm, results), model modeling results (including subjects / group, algorithm, model), etc.; the categories under physical storage include: file storage, small file storage, object storage, relational storage; the categories under logical storage include: local, remote, primary, secondary, restricted, unrestricted, etc.; the categories under physical servers include: cpu servers, gpu servers, cluster servers, etc.; the categories under preprocessing include: format conversion, removal of time points, time layer correction, head motion correction, standardization, regression covariates, linear drift removal, smoothing, filtering, deletion of time points, generation, etc.; the categories under feature extraction include: statistical indicators, graph theory indicators, structural networks, functional networks, feature selection and dimensionality reduction, etc.; the categories under statistical analysis can include: difference, correlation, regression, etc.; the categories under machine learning modeling can include: classification, clustering, integration, etc.; the categories under deep learning can include: based on cnn model, based on rnn model, based on attention model, integrated model, etc.; the categories under data quality control can include: quality control indicators, quality control methods, etc.; the categories under result annotation can include: annotation methods, annotation algorithms, etc.; the categories under auxiliary tools can include: format conversion, anonymization, etc.; the categories under processing strategy can include: analysis strategy, calculation strategy, etc.; the categories under data files can include: original files, converted files, intermediate process files, calculation result files, etc.; the categories under data processing tasks can include: data conversion tasks, preprocessing tasks, feature extraction tasks, statistical analysis tasks, machine learning tasks, deep learning tasks, etc.

[0068] The fourth level, fifth level, etc. can be obtained by sequentially decomposing, for example, the cognitive category can be further divided into: calculation, reasoning, problem solving, decision making, perception, memory, attention, visual space, execution, learning, language; the social cognition category can be further divided into: interpersonal relationship, social adaptation, etc.; the emotion category can be further divided into: anxiety, depression, hostility, dullness, confusion, etc.; the graph theory index category can be further divided into: node degree, shortest path, clustering coefficient, global efficiency, rich hub, etc.; the structural network category can be further divided into: white matter fiber, structural covariance, structural causality, etc.; the functional network category can be further divided into: dynamic, resting state, and undirected, directed, etc.; the quality control index category can be further divided into: signal-to-noise ratio, difference between structural images and templates, brain tissue symmetry, gray matter outside the template, difference between tool results, etc.; the quality control method category can be further divided into: supervised method, unsupervised clustering, etc.; the analysis strategy category can be further divided into: separate analysis, voting analysis, multi-dimensional analysis, etc.; the computing strategy category can be further divided into: cpu computing, gpu computing, hybrid computing, etc.; and so on, until it cannot be divided.

[0069] In addition, each subcategory under the computing process category includes input, output, parameters, algorithms, tools, steps, templates (if any), etc. detailed parameters; each subcategory under the storage category includes type, capacity, average access rate, etc. subcategories; each subcategory under the server category includes cpu / gpu core number, cpu / gpu frequency, internal storage, gpu video memory, current load, etc. subcategories; each subcategory under the object data category includes total data capacity, format, small file ratio, etc. subcategories.

[0070] The relationship categories in the knowledge base framework mainly include is-a and attribute relationships. According to the above knowledge base framework, using the evidence-based literature approach, for the processing target, object data, data storage, server, computing process, etc. content records that have clear high-quality evidence support, establish a knowledge base (entity and relationship knowledge content can be obtained by information extraction method or artificial sorting method, and the constructed knowledge base is essentially a knowledge graph); In addition, for the running case related data in the running process of the service system, it can be extracted from the log file and added to the knowledge base after artificial audit confirmation.

[0071] Step 2, information configuration

[0072] To maximize user convenience in processing neuroimaging data, a graphical interface is provided for users to express their requirements. The content for expressing these requirements in the graphical interface can be divided into three categories: data processing purpose, the data itself, and data processing expectations. Referring to the knowledge base framework mentioned above, the data processing purpose mainly includes: the targeted disease (can be empty), focus, etc. The data itself includes: data size, format, modality, current location (and directory hierarchy—e.g., subjects, groups, etc.). The data processing expectations include: expected data processing time, planned storage space, expected performance metrics, etc. (training, validation, and test datasets required for modeling are also specified here).

[0073] In addition, the physical hardware information of the service system can also be based on a graph in this embodiment.

[0074] The interface is obtained through the initial settings of the highest administrator. The physical hardware information of the service system can be divided into: storage information and execution server information; the storage information here includes storage type, location, IP, capacity, access rate, access key, etc.; the execution server includes type, location, IP, number of CPU / GPU cores, CPU / GPU frequency, internal storage, GPU memory, current load (which can be dynamically obtained automatically during the process), access key, etc.

[0075] Step 3: Determine the computing facilities

[0076] To achieve the user's data processing goals, this step requires determining the type, capacity, and other specifications of the storage.

[0077] This includes physical storage, server type, and specific server (all of which are computing facilities within the user's permission scope).

[0078] Considering that the initial knowledge base may not be sufficient to build a good computing facility decision model (machine learning model), the construction of the decision model here includes two parts: the rule part and the machine learning model part.

[0079] First, based on relevant expert knowledge, establish user data processing requirements and storage / server requirements.

[0080] The rules governing the data processing and decision-making are established based on these rules. Then, based on relevant data from server operation, machine learning methods are used to construct models related to user data processing requirements, storage, and the server (training data is derived from a knowledge base and confirmed by expert annotation). If the model's effectiveness is...

[0081] If the requirements are met, the above rules are replaced; here, the algorithm 0 used in model building includes integration based on decision tree, neural network; the input of the storage model is related to user data processing

[0082] Requirements: data processing purposes, data itself, data processing expectations, etc. (encode related parameters), model output for storage selection, storage model loss function considering storage balance, minimum transmission time, etc. The input of the server model is related to user data processing requirements: data processing purposes, data itself, data processing expectations, etc. (encode related parameters), and storage selection, etc. The output is server selection, and the loss function of the server model considers server balanced load, minimum execution time, etc.

[0083] The determined computing facility result is automatically displayed in a graphical manner, and user self-customized modification options for the information are provided.

[0084] In addition, the embodiment can also provide settings such as adding, changing, etc. of new storage, servers, etc.

[0085] Step 4, data acquisition

[0086] According to the storage type, capacity and specific storage selection determined in the above steps, the data specified in the user information configuration is transmitted to the determined storage. During transmission, it can be transmitted through http protocol, or it can be quickly transmitted through ftp or sftp protocol for large capacity neural image data. According to the specific storage determined in the above steps, if the physical storage location exceeds 1, that is, when the data is distributedly stored, the data needs to be segmented to different storage locations according to the determined capacity.

[0087] For the data object transmitted to the destination storage, according to the pre-defined data organization method (such as bids), if it meets the agreed format, it is checked and the data acquisition is completed; if the data does not meet the agreed format, the data object is converted according to the agreed format, and the data acquisition is completed. The agreed data format provides a data basis for subsequent flexible selection, automatic execution of data processing flow, and data sharing within a certain range.

[0088] Step 5, data analysis; this step includes two sub-steps: data analysis execution flow and parameter determination, and data analysis execution process.

[0089] (1) Data analysis execution flow and parameter determination

[0090] According to the relevant requirements of the user on the data processing in the information configuration, including diseases, concerns and data conditions, etc., based on the established data analysis decision model, the specific execution procedure of the data analysis flow is obtained, including preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, processing strategy, etc.

[0091] Wherein, in order to obtain the data analysis decision model, first, a training database is constructed, the training data comes from the content in the knowledge base and is confirmed by expert annotation, and the specific data mainly includes analysis requirements (mainly including diseases, concerns-constituting an encoded text sequence after standardization, and calculation efficiency and performance requirements-according to the rules determined in advance, the efficiency and performance requirements are corresponded to the options of extremely high, high, medium, etc.-and constitute an encoded text sequence) and data analysis execution flow; the data analysis execution flow is composed of one or more components, the components mainly refer to preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, processing strategy, etc.; the components are composed of one or more modules, the modules refer to specific algorithms applied to specific calculations, such as head movement correction algorithm (there can be multiple) in preprocessing; from this point of view, the data analysis execution flow is a directed graph, the nodes are modules, and the edges are sequential relationships; the text sequence of the data analysis requirements and the text sequence of the data analysis execution flow (both are standardized) are represented in the following way; and the training data is divided into training, validation and test sets in the ratio of 7:1.5:1.5.

[0092] The encoded text sequence expression of the analysis requirements is composed of: the diseases, concerns, etc. after standardization constitute an encoded text sequence, the calculation efficiency, performance requirements, etc. are corresponded to the options of extremely high, high, medium, etc. according to the rules determined in advance and constitute an encoded text sequence, and then the text sequences are combined together to constitute a unified text sequence (the connection symbol is comma,, ), such as: (Alzheimer's disease, memory, high efficiency, high performance). The text sequence expression of the data analysis execution flow is composed of: the main part is the component, the component includes the module, the specific algorithm and the related parameter setting recommendation (connected by comma), adopts a hierarchical method for representation, the sequential relationship between the components is represented by "->", the sequential relationship between the modules is represented by "=>", and the parallel relationship is represented by comma ", ", such as:

[0093] (preprocessing (

[0095] (remove time point, algorithm =, first = 10, …) => (…) ) )

[0098] -> (feature extraction (…) => …)

[0099] ->(…) )

[0101] Correspondingly, the input of the data analysis decision model is the relevant requirements of data analysis, i.e. the textual sequence expression composed of disease, focus, calculation efficiency and performance requirements, etc. (after standardization). The output of the data analysis decision model is the textual sequence expression composed of components (i.e. preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, processing strategy, etc. and specific modules constituting the components, etc.). The model adopts a sequence-to-sequence generation model, such as rnn combined with attention or T5, etc.

[0102] According to the data analysis requirements of the user, the data analysis execution flow can be obtained by the data analysis decision model. The software-related preset parameters in the data analysis execution flow are preset according to the evidence-based literature and the source is indicated; the data object-related preset parameters are obtained by pre-detection of the software (including the user preset part).

[0103] The results of the data analysis execution flow obtained by the data analysis decision model are expressed in a graphical manner on the user interaction interface, and user self-customization selection is provided.

[0104] In addition, the embodiment can also provide settings such as addition, change, etc. of new modules, components, execution flows, etc.

[0105] (2) Data analysis execution process

[0106] After the user customizes the data analysis execution flow, the data analysis execution flow can be executed according to the user setting (based on the above determined server, the corresponding category of task is started), and the data analysis process (such as general data preprocessing, feature extraction, statistical analysis, machine learning modeling, etc.) is automatically completed.

[0107] The intermediate result files and the result files in the data analysis process are stored according to the previously agreed format (such as reference to bids) (the results are marked according to date, software name, etc. for distinction).

[0108] It is worth noting that the analysis strategy and the calculation strategy are an important factor in the execution of the data analysis execution flow (the storage and the server determined in the above steps affect the analysis and calculation strategy as implicit parameters, such as recommending weighted multi-dimensional analysis when multiple servers are allocated). According to the different analysis strategies, the data analysis process includes separate analysis, voting analysis, multi-dimensional analysis (weighted), and other different ways of executing the analysis process, while according to the different calculation strategies, the data analysis process can be divided into cpu calculation, gpu calculation, hybrid calculation, and other different calculation methods. The above strategies are automatically recommended (including referring to literature or following the default settings - such as separate analysis, cpu calculation, etc.) in the generation of the data analysis execution flow and are filtered by the rules between the storage server and the strategy (determined in advance, used to determine the analysis and calculation strategy in a specific storage and server environment).

[0109] Step 6, analysis evaluation

[0110] This step mainly includes data quality control, result evaluation, etc. Among them, the result evaluation is mainly to objectively evaluate the data analysis execution result; while the data quality control is mainly evaluated from the whole process of data analysis (along with the data analysis process), so as to provide process quality basis for reasonably viewing the result.

[0111] The result evaluation here mainly provides different evaluation parameters to make an objective understanding of the result itself. The evaluation parameters include the significance level in statistical analysis, the correlation coefficient size, the regression coefficient size, the accuracy rate of machine learning model, the recall rate, F1, ROC, AUC, etc.

[0112] The result of data quality control is a dimensionless number (for each data object, etc.), called data quality control index (0-100), which is divided into several ranges according to the size of the index: less than 60, the data is not available (the data in this state will not participate in the data analysis process), 60-80, the data is available, and greater than 80, the data is better.

[0113] The calculation process of data quality control index, such as for structural images, first extract the feature data of image data - including parameter difference (such as repetition time), signal-to-noise ratio, difference between structural image and template, brain tissue symmetry, gray matter outside the template, difference between different tool calculation results, etc. After standardization and normalization (standardization and normalization according to the pre-determined gold standard structural image and its numerical value), the data is weighted (the weight is determined by the Delphi method in advance) and normalized to 0-100.

[0114] The calculation result of data analysis, result evaluation, data quality control index, etc. are saved to the user-specified location in the format (such as bids) mode determined in advance.

[0115] Step 7, obtaining results

[0116] After the data calculation and analysis process such as data analysis, analysis evaluation and data quality control is completed, the results can be viewed, downloaded and used for further analysis by using the graphical interface provided by the method.

[0117] Compared with the prior art, the above-mentioned embodiments of the application have the following characteristics: (1) the target, data related situation and the like of the data processing are configured through the interface, and the whole process from preprocessing, feature extraction, statistical analysis, machine learning modeling and deep learning modeling can be selected and completed in one station; (2) based on the reasoning method combining knowledge and learning, combined with the standardized literal sequence expression method, the automatic recommendation of the data analysis process can be supported; (3) based on the specially designed evidence-based medicine knowledge base, the automatic recommendation of the parameter setting in the analysis process can be supported; (4) on the basis of automatic recommendation, the execution flow, component and module of the data analysis are supported to be selected by hot plug mode; (5) by the automatic selection and recommendation of the data processing hardware facilities, the automatic recommendation of the data analysis process, the automatic setting and recommendation of the related parameters and the method of labeling the related references, the difficulty of medical image data processing is reduced, and the work efficiency and work quality are improved; (6) by setting a unified data format (including data directory and the like), on the one hand, the foundation for automatic execution of data processing and analysis is laid, and on the other hand, the data and results within a certain range are also helpful for sharing.

[0118] In addition, the embodiment of the application also provides a data processing system for medical images, which is described with reference to Figure 3 The system comprises:

[0119] The model training module 310 is configured to train the data analysis decision model by using the training database; the training database is generated based on the medical image data processing knowledge base; the medical image data processing knowledge base comprises a data processing target and a data analysis link;

[0120] The execution flow generation module 320 is configured to generate a target data analysis execution flow by using the trained data analysis decision model according to the data analysis requirement; the data analysis requirement comprises a processing target for medical image data; the target data analysis execution flow comprises a plurality of medical image data analysis links, related parameters of each link and a corresponding execution order;

[0121] The processing result generation module 330 is configured to process the obtained actual medical image data based on the target data analysis execution flow to generate a final processing result, so as to realize automatic analysis and processing of the medical image data.

[0122] The medical image-oriented data processing system provided by the embodiments of the present application can be specific hardware on a device or software or firmware installed on the device, etc. The system provided by the embodiments of the present application has the same implementation principles and technical effects as the method embodiments described above, and for brevity of description, the part not mentioned in the device embodiments can refer to the corresponding content in the method embodiments described above. Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the method embodiments described above, which will not be described herein. The medical image-oriented data processing system provided by the embodiments of the present application has the same technical features as the medical image-oriented data processing method provided by the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0123] The embodiments of the present application also provide an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program performs the method of any one of the above embodiments when executed by the processor.

[0124] Figure 4 The structure schematic diagram of an electronic device provided by the embodiments of the present application is shown in the figure, and the electronic device 400 includes a processor 40, a memory 41, a bus 42 and a communication interface 43, the processor 40, the communication interface 43 and the memory 41 are connected through the bus 42; the processor 40 is used to execute the executable modules stored in the memory 41, such as computer programs.

[0125] The memory 41 can contain a high-speed random access memory (RAM, Random Access Memory) and can also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0126] The bus 42 can be an ISA bus, a PCI bus or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 4 Only one bidirectional arrow is used to represent the bus, but it does not mean that there is only one bus or one type of bus.

[0127] The memory 41 is used to store programs, and the processor 40 executes the programs after receiving execution instructions. The method executed by the device defined by the flow process disclosed in any one of the embodiments of the present application can be applied to the processor 40 or realized by the processor 40.

[0128] Processor 40 may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method can be completed through integrated logic circuits in the processor 40 or through software instructions. The processor 40 described above can be a general-purpose processor, including a central processing unit (CPU).

[0129] This can be a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute this invention.

[0130] The methods, steps, and logic block diagrams disclosed in the embodiments are shown below. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc.

[0131] The storage medium is located in memory 41. The processor 40 reads the information from memory 415 and, in conjunction with its hardware, completes the steps of the above method.

[0132] Corresponding to the above method, this application embodiment also provides a computer-readable storage medium storing machine-executable instructions. When the machine-executable instructions are called and run by a processor, the machine-executable instructions cause the processor to perform the steps of the above method.

[0133] In the embodiments provided in this application, it should be understood that the disclosed apparatus and method can be used through...

[0134] Other methods can be used to achieve this. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and other division methods may be used in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be...

[0135] may be ignored or not performed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed may be indirect coupling or communication connection through some communication interface, device or unit,

[0136] may be electrical, mechanical or other forms.

[0137] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e. they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.

[0138] In addition, the functional units in the embodiments provided in the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0139] The functions, if realized in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part of the present application that essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, an electronic device or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk and various program code storage media.

[0140] It should be noted that: similar reference numerals and letters represent similar items in the drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings, in addition, the terms "first", "second", "third" and the like are only used to distinguish description, and cannot be understood as indicating or implying relative importance.

[0141] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A data processing method for medical images, characterized by, The method comprises the following steps: training a data analysis decision model by using a training database; the training database is generated based on a medical image data processing knowledge base; the medical image data processing knowledge base comprises data processing targets and data analysis links; the training database comprises data analysis requirements and data analysis execution flows; the data analysis requirements are coded text sequence expressions of the processing targets of medical image data; the processing targets of medical image data comprise diseases, concerns, efficiency and performance; the data analysis execution flows comprise text sequence expressions of components corresponding to at least one medical image data analysis link; the medical image data analysis links comprise any of preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation and processing strategies; each component comprises at least one module for representing a specific algorithm of the corresponding link; generating a target data analysis execution flow by using the trained data analysis decision model according to the data analysis requirements; the target data analysis execution flow comprises a plurality of medical image data analysis links, related parameters of each link and a corresponding sequence of execution; sequentially executing the medical image data analysis links on the acquired actual medical image data according to the corresponding sequence, and generating processing results of each link; calculating a data quality control index for the processing results of the data analysis links; the data quality control index is used to evaluate the usability of data; generating a final processing result based on the processing results of each link, and generating a result evaluation parameter for the final processing result, so as to realize automatic analysis and processing of medical image data.

2. The medical image-oriented data processing method according to claim 1, characterized in that, The method further comprises the following steps: constructing a medical image data processing knowledge base; the framework of the medical image data processing knowledge base comprises a plurality of top-level categories and a plurality of sub-level categories; each current sub-level category is obtained by subdividing a previous level category of the current sub-level category; the top-level categories comprise data processing targets, data analysis links, object data, data storage, servers, files and tasks; constructing a training database based on the medical image data processing knowledge base.

3. The medical image-oriented data processing method according to claim 1, characterized in that, The method further comprises the following steps: constructing a computing facility decision model; the computing facility decision model comprises corresponding rules and machine learning models; generating corresponding rules between user data processing requirements and data storage and servers according to the medical image data processing knowledge base; training a machine learning model based on related data in the running process of the servers; the machine learning model comprises a data storage decision model and a server decision model; the input of the data storage decision model is the user data processing requirements, and the output is a data storage target; the input of the server decision model is the user data processing requirements, and the output is a server target; determining a target computing facility based on the computing facility decision model; the target computing facility is used to execute the data analysis links of the target data analysis execution flow.

4. The medical image-oriented data processing method according to claim 3, characterized in that, The step of determining a target computing facility based on the computing facility decision model comprises the following steps: determining a data storage target based on the data storage decision model; the data storage target comprises a storage type, a capacity, and a target storage device; the target storage device is used to store initial data to be processed and data generated by each data analysis link; determining a server target based on the server decision model; the server target comprises a target server type and a target server; the target server is used to perform calculation of each data analysis link.

5. The medical image-oriented data processing method according to claim 4, characterized in that, The method further comprises: generating a processing strategy related to the target data analysis execution flow according to the data analysis requirement and using the trained data analysis decision model; the processing strategy comprises a data calculation strategy and a data analysis strategy; the data analysis strategy is used to represent a manner of executing each analysis link in the target data analysis execution flow; the data analysis strategy comprises individual analysis, voting analysis, and multi-dimensional analysis; the data calculation strategy comprises a data calculation manner in the analysis link; the data calculation manner comprises CPU calculation, GPU calculation, and hybrid calculation.

6. The medical image-oriented data processing method according to claim 5, characterized in that, The method further comprises: filtering the data calculation strategy and the data analysis strategy based on a predefined rule between the data storage target, the server target, and the processing strategy, to generate filtered data calculation strategy and data analysis strategy.

7. A medical image-oriented data processing system, characterized by The system comprises: a model training module configured to train a data analysis decision model using a training database; the training database is generated based on a medical image data processing knowledge base; the medical image data processing knowledge base comprises a data processing target and a data analysis link; the training database comprises a data analysis requirement and a data analysis execution flow; wherein the data analysis requirement is a coded character sequence expression of a processing target of medical image data; the processing target of medical image data comprises a disease, a focus, an efficiency, and a performance; the data analysis execution flow comprises a character sequence expression of a component corresponding to at least one medical image data analysis link; the medical image data analysis link comprises any of preprocessing, feature extraction, statistical analysis, machine learning modeling, deep learning modeling, visualization, result annotation, and processing strategy; each component comprises at least one module, and the module is used to represent a specific algorithm of a corresponding link; an execution flow generation module configured to generate a target data analysis execution flow according to a data analysis requirement and using a trained data analysis decision model; the target data analysis execution flow comprises a plurality of medical image data analysis links, related parameters of each link, and a corresponding execution order. The processing result generation module is configured to sequentially execute the medical image data analysis links on the acquired actual medical image data in the corresponding order, generate processing results of each link, calculate a data quality control index for the processing results of the data analysis links, wherein the data quality control index is used to evaluate the usability of the data, generate a final processing result based on the processing results of each link, and generate a result evaluation parameter for the final processing result, so as to realize automatic analysis and processing of the medical image data.

8. An electronic device comprising a memory, a processor, the memory having stored therein a computer program executable on the processor, characterized in that, The processor implements the steps of the method of any one of claims 1 to 6 when executing the computer program.

Images