Medical image data retrieval method, device, medium, program product and terminal

Abstract

The application provides a medical image data retrieval method, device, medium, program product and terminal, and semantic understanding of a user retrieval request is realized through a medical language recognition model. A user input retrieval request is subjected to semantic analysis through a medical language recognition model to generate a user intention feature set. The user intention feature set is converted into a database query instruction according to a preset standard, a query is performed in a medical image database, and matched original retrieval data is obtained. According to a user type requirement, a preprocessing operation is performed on the original retrieval data to generate a standardized medical image retrieval result. The application breaks through the limitation of traditional medical image retrieval, and solves the problem that the retrieval accuracy of an existing medical image retrieval system is not high and the user experience is poor.

Description

Technical Field

[0001] This application relates to the field of medical image data retrieval, and in particular to medical image data retrieval methods, devices, media, program products and terminals. Background Technology

[0002] Currently, common solutions to the problem of medical image data retrieval include keyword matching and metadata-annotated retrieval. Traditional technologies typically employ static labels or simple classification systems, requiring doctors and researchers to navigate complex search interfaces and use precise search terms to locate the desired images. Existing technologies heavily rely on manually pre-defined labels and classification systems, often necessitating specialized knowledge and precise search strategies in the retrieval process.

[0003] However, existing technologies for medical image data retrieval suffer from several problems. From the user's perspective, retrieval systems lack flexibility, making it difficult to simultaneously meet the diverse needs of clinicians and researchers. From a data organization perspective, the metadata tag structure is simplistic, failing to accurately represent the multidimensional and complex characteristics of medical images. In terms of system intelligence, the retrieval process lacks real-time and personalization, unable to dynamically adjust and optimize results based on user search habits. More importantly, existing systems present a high retrieval threshold and a poor user experience for users with diverse professional backgrounds. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a medical image data retrieval method, device, medium, program product and terminal to solve the problems of low retrieval accuracy and poor user experience of existing medical image retrieval systems.

[0005] To achieve the above and other related objectives, a first aspect of this application provides a medical image data retrieval method, comprising: acquiring an image retrieval request input by a user; inputting the image retrieval request into a medical language recognition model to perform semantic parsing operations to generate a set of user intent features; converting the set of user intent features into a database query instruction based on a preset standard; performing a query operation in a medical image database based on the database query information to obtain raw retrieval data; and performing corresponding preprocessing operations on the raw retrieval data based on the image retrieval request to generate standard medical image retrieval results.

[0006] In some embodiments of the first aspect of this application, the process of inputting the image retrieval request into a medical language recognition model to perform semantic parsing operations to generate a set of user intent features includes: identifying medical entity information in the image retrieval request, performing standardization processing on the identified medical entity information, and inserting semantic tagging information into the standardized medical entities to generate medical structured entities; semantically encoding each medical entity in the medical structured entities according to a pre-trained model to generate a semantic vector; extracting the relationships between each medical entity from a medical knowledge graph based on the medical structured entities, and constructing corresponding relationship vectors; and generating the set of user intent features based on the semantic vectors and the relationship vectors.

[0007] In some embodiments of the first aspect of this application, the process of converting the set of user intent features into a database query instruction based on a preset standard includes: constructing a query structure based on the medical image database; and sequentially inserting parent / child retrieval conditions into the query structure in a preset multi-level nested structure according to the semantic vector and the relation vector to generate a database query instruction.

[0008] In some embodiments of the first aspect of this application, the image retrieval request includes a query user type; the process of performing corresponding preprocessing operations on the original retrieval data based on the image retrieval request to generate standard medical image retrieval results includes: in response to the query user type being a developer, performing normalization and standardization operations on the labeled data in the original retrieval data, and performing pixel value adjustment, resampling, and rescaling operations on the image data in the original retrieval data to generate standard medical image retrieval results.

[0009] In some embodiments of the first aspect of this application, the medical image database includes medical image data, as well as detection annotation information, segmentation annotation information, and key point annotation information generated by performing data annotation operations based on the CT image data.

[0010] In some embodiments of the first aspect of this application, the method further performs the following steps: based on multiple medical image retrieval requests entered by the user, the frequency of occurrence of each search keyword in the medical image retrieval request is counted; for search terms with a frequency higher than a preset threshold, corresponding index tags are created in the medical image database.

[0011] To achieve the above and other related objectives, a second aspect of this application provides a medical image data retrieval device, comprising: a request processing module for acquiring an image retrieval request input by a user; inputting the image retrieval request into a medical language recognition model to perform semantic parsing operations to generate a set of user intent features; a query construction module for converting the set of user intent features into a database query instruction based on a preset standard, and performing a query operation in a medical image database based on the database query information to obtain raw retrieval data; and a result acquisition module for performing corresponding preprocessing operations on the raw retrieval data based on the image retrieval request to generate standard medical image retrieval results.

[0012] To achieve the above and other related objectives, a third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the medical image data retrieval method.

[0013] To achieve the above and other related objectives, a fourth aspect of this application provides a computer program product comprising computer program code, which, when executed on a computer, enables the computer to implement the medical image data retrieval method.

[0014] To achieve the above and other related objectives, a fifth aspect of this application provides an electronic terminal, including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the medical image data retrieval method.

[0015] As described above, the medical image data retrieval method, device, medium, program product, and terminal of this application have the following beneficial effects: They achieve medical entity standardization, semantic encoding, and relationship construction through a medical language recognition model, significantly improving the depth of semantic understanding in retrieval; they establish a multi-layered nested query structure, enabling precise location of massive amounts of medical image data; they provide personalized data preprocessing for different user types (such as developers), achieving normalization of labeled data and resampling of image data; and they innovatively generate index tags dynamically based on retrieval frequency, continuously optimizing retrieval performance. This application breaks through the limitations of traditional retrieval technologies, significantly lowers the professional threshold for retrieval, improves retrieval accuracy and intelligence, and provides a brand-new solution for medical big data retrieval, possessing significant technological innovation value and practical significance. Attached Figure Description

[0016] Figure 1 The diagram shows a flowchart of an embodiment of the medical image data retrieval method of this application.

[0017] Figure 2The diagram shows a schematic representation of the semantic parsing operation in one embodiment of the medical image data retrieval method of this application.

[0018] Figure 3 The diagram shows a schematic representation of the medical language recognition model in one embodiment of the medical image data retrieval method of this application.

[0019] Figure 4 This paper shows a schematic diagram of the structure of a medical knowledge graph in one embodiment of the medical image data retrieval method of this application.

[0020] Figure 5 A schematic diagram of an embodiment of the medical image data retrieval device of this application is shown.

[0021] Figure 6 The diagram shows a structural schematic of an embodiment of the medical image data retrieval terminal of this application. Detailed Implementation

[0022] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0023] Before providing a further detailed description of the present invention, the nouns and terms used in the embodiments of the present invention are explained, and the nouns and terms used in the embodiments of the present invention are subject to the following interpretations:

[0024] <1> The International Classification of Diseases (ICD-10 / ICD-11) is a globally unified classification and coding system for diseases and related health problems developed by the World Health Organization (WHO). It is used to standardize the classification of diseases, symptoms, signs, external causes, and health conditions, and is widely used in various scenarios such as clinical diagnosis, health statistics, disease surveillance, health insurance, and research.

[0025] <2> Systematic Medical Nomenclature - Clinical Terminology (SNOMED CT): An internationally recognized comprehensive clinical medical terminology system that covers medical concepts such as diseases, clinical manifestations, procedures, body structures, and drugs. It is used to achieve structured data expression and semantic interoperability in medical information systems such as electronic health records (EHRs).

[0026] <3> RadLex (Radiology Terminology Standard): A unified terminology system for radiology developed by the American College of Radiology (ACR). It covers imaging examinations, tissue structures, diagnoses, and the instruments and parameters involved in the process. It is used to standardize the annotation of medical imaging data and radiological reports, and improve the consistency and interoperability of imaging information.

[0027] <4> BERT: A deep pre-trained language understanding model based on the Transformer architecture proposed by Google. It learns the contextual representation of words and sentences through masked language modeling of large-scale unsupervised text corpora and next sentence prediction tasks. It is widely used in natural language processing applications such as text classification, question answering systems, and named entity recognition.

[0028] <5> Tag embedding: Mapping words, subwords or other tokens in text into dense low-dimensional vectors to capture their semantic and syntactic properties as input features for natural language processing models (such as neural networks), and is the foundation of natural language understanding and generation tasks.

[0029] <6> Segment embedding: Assigning unique vector representations to different paragraphs or semantic segments in a text to indicate their syntactic or semantic affiliation. It is often used to distinguish sentence pairs or multi-segment text structures for information modeling.

[0030] <7> Position embedding: Assigning specific vectors to each element in a sequence according to its sequential position, enabling the model to capture and utilize temporal information from data such as text and sequences. In structures such as Transformer, it is used to compensate for the deficiencies of position-independent mechanisms in the structure.

[0031] <8> UMLS / SNOMED CT / RadLex: UMLS (Unified Medical Language System) is a medical terminology integration platform developed by the U.S. National Library of Medicine. It integrates multiple medical vocabularies and standards (such as ICD, SNOMED CT, MeSH, etc.) to achieve semantic mapping and interoperability between heterogeneous medical terminology systems. SNOMED CT is a globally leading clinical health terminology standard that supports structured expression of medical information and efficient data retrieval. RadLex is an internationally unified radiology terminology system that standardizes the representation of imaging information and promotes radiology data sharing and automated processing.

[0032] <9> DisGeNET is a disease-gene relationship knowledge base that integrates multiple authoritative databases and literature, describing information such as pathogenic genes, related diseases, and association evidence, and widely supporting research on the genetic basis of human diseases and applications in genomic medicine.

[0033] <10> PharmGKB: A comprehensive database focused on pharmacogenomics research, containing detailed information on drugs, genes, variants, diseases, and drug-gene interactions, providing knowledge support for personalized drug therapy, gene-guided medication, and clinical drug safety.

[0034] <11> Multi-level nested structure: refers to a complex hierarchical system formed by multiple nested hierarchical structures. It is often used to describe the grammatical and semantic structure of natural language, or the hierarchical classification relationship of medical knowledge, to improve the organization and semantic expression of data.

[0035] <12> DICOM format: Digital Imaging and Communications in Medicine, defines the standards for storing, exchanging, transmitting and managing medical images and related information, and is widely used by medical institutions worldwide for the integration and interoperability of medical image data.

[0036] <13> Nifti format: A commonly used data storage and exchange standard in neuroimaging (Neuroimaging Informatics Technology Initiative), supporting multidimensional (such as three-dimensional and four-dimensional) neuroimaging data and spatial distribution information, facilitating quantitative analysis and data sharing in neuroscience and medical imaging.

[0037] To facilitate understanding of the embodiments of this application, firstly, in conjunction with Figure 1 Detailed explanation. Figure 1 A flowchart illustrating a medical image data retrieval method according to an embodiment of the present invention is shown. The medical image data retrieval method in this embodiment mainly includes the following steps:

[0038] Step S11: Obtain the image retrieval request input by the user.

[0039] In one embodiment of this application, the image retrieval request includes, but is not limited to, image type, time, patient information, disease description, etc.

[0040] In this embodiment, the image type parameter is used to limit the modal category of the medical image to be retrieved, including but not limited to standard types in medical image classification systems such as X-ray films, CT scans, and MRI. The time parameter adopts a start-end timestamp format or a relative time expression, supporting both precise time point retrieval and time range retrieval modes. The patient information field includes at least patient identification elements such as medical record number, initials of name in pinyin, and date of birth. The disease description section receives natural language text input, extracts and standardizes keywords through a preset medical terminology dictionary, and converts it into standardized query conditions conforming to the International Classification of Diseases (ICD) coding system.

[0041] Step S12: Input the image retrieval request into the medical language recognition model to perform semantic parsing operations, so as to generate a set of user intent features.

[0042] In one embodiment of this application, the process of inputting the image retrieval request into a medical language recognition model to perform semantic parsing to generate a set of user intent features includes: identifying medical entity information in the image retrieval request, performing standardization processing on the identified medical entity information, and inserting semantic tagging information into the standardized medical entities to generate medical structured entities; semantically encoding each medical entity in the medical structured entities according to a pre-trained model to generate a semantic vector; extracting the relationships between each medical entity from a medical knowledge graph based on the medical structured entities and constructing corresponding relationship vectors; and generating the set of user intent features based on the semantic vectors and the relationship vectors.

[0043] In this embodiment, the process of standardizing the identified medical entity information includes: transforming the input information according to a preset medical terminology dictionary and predefined specifications to generate medical entity information conforming to a structured form. Subsequently, semantic tagging information is inserted into the standardized medical entities to complete the construction of the structured medical entities. The preset medical terminology dictionary and predefined specifications include, but are not limited to, standardized medical classification systems and clinical data specifications such as the International Classification of Diseases (ICD-10 / ICD-11), Systematic Medical Nomenclature-Clinical Terminology (SNOMED CT), and RadLex (Radiology General Terminology Standard). Inserting semantic tagging information refers to adding identifiers or metadata that can represent the semantic attributes, types, and relationships of the standardized medical entities to enhance their semantic expressiveness and computer understandability; for example, adding semantic dimension tags such as "disease type" and "anatomical location" to "DDH".

[0044] For example, such as Figure 2As shown, the process of identifying medical entity information in the image retrieval request includes: when a user inputs "I need some full-length lower limb CT scans with DDH" as an image retrieval request, the system will first identify basic medical entity information through a semantic parsing process, including "need", "DDH" (medical term for developmental dysplasia of the hip), "CT" (medical imaging type), and "full-length lower limb" (shooting range). The entity information will be standardized to ensure that it conforms to the specifications of the Unified Medical Language System (UMLS) or other medical terminology dictionaries, and the following information will be input: {'disease':('DDH',),'modality':('CT',),'shooting_range':('full-length lower limb'),'operation':('download')}.

[0045] Figure 3 This illustration shows a schematic diagram of the medical language recognition model in one embodiment of this application. The input image retrieval request is "I need some full-length CT scans of the lower limbs with DDH". This text information is first processed by word segmentation to divide it into individual tokens. A [CLS] and a separator [SEP] are added to each token to perform a classification task, generating the following word segmentation information: [CLS], I, need, some, have, DDH, lower limbs, full length, of, CT, [SEP]. The word segmentation information is integrated into different types of embedding representations, and a complete input representation is generated by adding the different embedding representations. These include: token embeddings, which convert each token into a vector representation, such as E[CLS], E I, E need, etc.; segment embeddings, which use EA and EB to distinguish different sentences; and position embeddings, which encode the position information of each token in the sequence. The combination of embeddings is input into a multi-layer Transformer encoder (labeled "Trm"), where there are full connections between Transformer layers.

[0046] Further, after processing by the embedding layer, the resulting embedded representations are combined additively to form an input vector, which is then fed into a multi-layer Transformer encoder layer (Trm). It's important to clarify that BERT itself is built on the Transformer architecture; more precisely, it utilizes the encoder portion of the Transformer. Therefore, Trm represents a Transformer encoder layer. Within the encoder layer, through a self-attention mechanism, each token can focus on other tokens in the entire sequence, thereby capturing intra-sequence dependencies and constructing context-sensitive representations. Ultimately, the model outputs a contextualized vector representation, which can be used for various downstream tasks, such as classification, named entity recognition, or question answering systems. This structure enables BERT to understand the semantic content of specialized medical queries like "DDH lower limb full-length CT scan."

[0047] In one embodiment of this application, the process of semantically encoding each medical entity in the medical structured entity according to a pre-trained model includes: extracting the relationships between each medical entity from an existing medical knowledge graph, and generating corresponding relationship vectors using knowledge graph-based embedding techniques. The relationship vectors include descriptions of the semantic connections between entities, such as the association between human anatomical structures and related diseases, or the diagnostic range supported by a certain medical imaging device. Existing medical knowledge graphs mainly include standardized knowledge bases such as UMLS, SNOMED CT, RadLex, DisGeNET, and PharmGKB, as well as professional domain knowledge systems such as clinical practice guideline graphs and traditional Chinese medicine knowledge graphs.

[0048] Furthermore, the medical language recognition model can employ the BERT model. Specifically, the BERT model in this embodiment is trained using a structured corpus and employs entity embedding and relation embedding techniques to support the construction and reasoning capabilities of the medical knowledge graph. During the pre-training phase, the model constructs a graph of medical entities and their relationships using a specialized corpus. For example, it transforms medical knowledge points such as "skeleton - contains - pelvis", "pelvis - prone to - fractures", and "CT - supports diagnosis - hip joint disease (HOA)" into trainable graph vector representations.

[0049] Specifically, such as Figure 4As shown in the figure, semantic relationships between medical entities are modeled and analyzed based on the medical knowledge graph. The knowledge graph contains multiple key medical concept nodes (blue circular nodes such as "anatomical structure," "disease," "imaging modality," and "surgical procedure") and their related sub-concepts (light-colored elliptical nodes). Semantic vectors are generated for each medical entity, and relation vectors are generated based on three different types of relationships in the figure (blue line represents "subordination," red line represents "diagnostic support," and green line represents "predisposing diseases"). The semantic and relation vectors are standardized to eliminate numerical scale differences between different concept nodes. Subsequently, an attention mechanism is used to fuse the vectors, focusing particularly on the semantic relevance between medical entities shown in the figure, such as the subordinate relationship between "DDH" (developmental dysplasia of the hip) and "disease," and the diagnostic support relationship between CT scans and various diseases. By calculating the connection strength between entity nodes, a semantic relevance matrix is ​​constructed, encompassing dimensions such as "anatomical structure" (e.g., bone, hip bone, pelvis), "disease" (e.g., DDH, OA, KOA), "imaging modality" (e.g., CT, MR, X-ray), and "surgical procedure" (e.g., THA, TKA, HTO). Finally, principal component analysis (PCA) is used to extract the most representative feature dimensions from the high-dimensional semantic space to generate a feature set that accurately expresses the user's query intent (e.g., "full-length CT scan of the lower extremities for DDH"), thereby supporting the system's understanding of the specific needs and context of users in medical image retrieval.

[0050] Step S13: Convert the set of user intent features into a database query instruction based on a preset standard, and perform a query operation in the medical image database based on the database query information to obtain the original retrieval data.

[0051] In one embodiment of this application, the process of converting the set of user intent features into a database query instruction based on a preset standard includes: constructing a query structure based on the medical image database; and sequentially inserting parent / child retrieval conditions into the query structure in a preset multi-level nested structure according to the semantic vector and the relation vector to generate a database query instruction.

[0052] In this embodiment, the hierarchical structure of parent / subclass search conditions includes a multi-level nested structure: first-level categories such as "operative procedure / module", "disease", "location", "includes prosthesis", "image modality / image type", and "data status"; subclass search conditions correspond to second-level or deeper-level category nodes under the main branch. For example, "location" as a parent category search condition includes subclass search conditions such as "lower limb", "pelvis", "spine", and "foot and ankle"; "image modality / image type" as a parent category search condition includes subclass conditions such as "CT", "MR", "DX", "CR", "RF", "SR", and "PNG".

[0053] When constructing database query commands, the main search categories in the user intent feature set are first identified to determine which parent category search conditions should be used. Then, the search conditions are further refined to the corresponding sub-category search conditions. For example, based on the appearance of "lower limbs" and "CT" in the user query, the "lower limbs" sub-category under the "location" parent category and the "CT" sub-category under the "image modality" parent category are matched respectively. Through this multi-layered nested structure, database query commands that conform to the knowledge system of the medical professional field can be accurately constructed, achieving precise retrieval of medical imaging data and reducing the difficulty of balancing recall and precision.

[0054] In this embodiment, the database query structure adopts a multi-layered nested metadata organization form to facilitate dynamic adjustment of the tag hierarchy order according to user needs. For example, for doctor users, their search needs typically focus on data attributes such as symptoms, image capture date, and image capture range. Therefore, image data related to the user's target intent is prioritized for filtering, with DICOM or Nifti format as the primary output format. For instance, when filtering CT image data with DDH symptoms, non-CT image data is first excluded, and then further filtered from the CT image data to include images of the lower limbs, pelvis, or mid-femur. In the generated query structure, the parent tag is CT, and the child tags are lower limbs, pelvis, or femur.

[0055] Furthermore, for developers, their retrieval needs typically focus on filtering target labeled data and generating model training data. For example, if a developer needs to obtain data for training a pelvic segmentation model, the database query must filter image data containing the pelvis from the segmentation labeled data. After filtering, the preprocessing module standardizes the format and content of the original image data. Specifically, the original image data is stored in .nii.gz format, the image grayscale values ​​are extracted using the itk module, and the image is isotropically converted so that all voxels are 1×1×1. Subsequently, the pixel data is stored as a NumPy array, and the array is normalized, mapping the pixel values ​​to the range (-1, 1). The normalized data is then cropped or downsampled according to the model input requirements; for example, the original 512×512×512 image data is downsampled to a target size of 256×256×256, and finally stored in .npy format for developers to use in model training.

[0056] Furthermore, to meet the presentation needs of doctors or academic exchanges, the labeled data can be visualized. For example, for keypoint labeled data, the number and coordinates of keypoints can be obtained based on keywords (such as "ACJoint") in the labeling file, and a three-dimensional sphere can be drawn on the image to visualize the keypoint information. Each keypoint corresponds to a sphere, which facilitates professionals to verify the consistency and reliability of the labeling results, thereby determining whether the labeling can be used as the gold standard in medicine.

[0057] It is worth noting that in the database query command generation process of this embodiment, metadata such as tags, time, and type of image data is automatically extracted from the parsed user intent feature set and transformed into query commands expressed in database language. The database engine processes the query commands, enabling rapid filtering of data that meets the criteria from a massive medical image dataset. Furthermore, this embodiment employs a multi-layered nested metadata structure to organize data, allowing the order of parent and child tags to be adjusted according to the needs of different user roles, thereby improving the accuracy and efficiency of retrieval. This dynamic adjustment mechanism ensures that both doctors and developers can obtain the most suitable search results in their respective use cases.

[0058] In one embodiment of this application, the medical image database includes medical image data, as well as detection annotation information, segmentation annotation information, and key point annotation information generated by performing data annotation operations based on the CT image data. The medical image data includes one or more of the following image data: CT image data, X-ray image data, and MRI image data.

[0059] In this embodiment, the data in the medical image database includes two types: medical image data and AI-annotated data. The medical image data is collected by patients at medical institutions, while the AI-annotated data is generated by importing the original medical image data and annotating it using auxiliary software such as Mimics and Labelme. The annotation content includes the spatial location and pixel range information of specific anatomical sites.

[0060] In this embodiment, the data storage format in the medical image database is shown in Table 1. This includes data type, image modality, storage format, and corresponding AI annotation storage format. All data types are medical images, and the image modalities include three types: CT, X-ray, and MRI. The original storage format for all modalities is dcm\nil.gz. The corresponding AI annotation storage format varies depending on the annotation type. Segmentation annotations include nil.gz, png, and json formats; keypoint annotations include nil.gz, json, and xml formats; and detection annotations include txt, xml, and json formats.

[0061]

[0062] Table 1: Data storage formats in medical image databases

[0063] Step S14: Based on the image retrieval request, perform corresponding preprocessing operations on the original retrieval data to generate standard medical image retrieval results.

[0064] In one embodiment of this application, the image retrieval request includes a query user type; the process of performing corresponding preprocessing operations on the original retrieval data based on the image retrieval request to generate standard medical image retrieval results includes: in response to the query user type being a developer, performing normalization and standardization operations on the labeled data in the original retrieval data, and performing pixel value adjustment, resampling, and rescaling operations on the image data in the original retrieval data to generate standard medical image retrieval results.

[0065] In this embodiment, corresponding preprocessing operations are performed based on the type of querying user. Specifically, for scenarios where the querying user is a developer, the raw medical image data and labeled data are converted into a standard format suitable for direct display and AI training, thus meeting the developer's data needs. During preprocessing, normalization and standardization operations are performed on the labeled data in the raw retrieval data to ensure consistency and comparability. Simultaneously, for image data, pixel value adjustment, resampling, and rescaling operations are performed to generate standard-compliant medical image retrieval results. These preprocessing operations effectively improve the efficiency and accuracy of subsequent image data usage and analysis.

[0066] Furthermore, this embodiment can automatically filter image data based on a multi-layered nested metadata structure to meet complex query needs. This multi-layered nested structure enables precise matching of user search requirements, ensuring data accuracy and relevance. In addition, it can statistically analyze the frequency of search keywords based on multiple user-entered medical image search requests and create corresponding index tags in the database for search terms with frequencies exceeding a preset threshold, thereby optimizing search efficiency and user experience.

[0067] In one embodiment of this application, the method further performs the following steps: based on multiple medical image retrieval requests input by the user, the frequency of occurrence of each search keyword in the medical image retrieval request is counted; for search terms with a frequency higher than a preset threshold, corresponding index tags are created in the medical image database.

[0068] This embodiment provides a method for optimizing a database based on multiple user-inputted medical image retrieval requests. Specifically, based on multiple user-inputted medical image retrieval requests, the frequency of each search keyword is statistically analyzed, and for search terms with a frequency exceeding a preset threshold, corresponding index tags are created in the medical image database. Specifically, firstly, multiple medical image retrieval requests submitted by the user are recorded, and each search keyword is extracted from them. Subsequently, the frequency of the search keywords is statistically analyzed, for example, counting the number of times each keyword is used within a certain time range (e.g., 20 times). When the number of times a keyword appears exceeds this threshold, the index tag creation process is triggered.

[0069] Furthermore, for search keywords that meet the criteria, corresponding index tags are created in the medical imaging database so that subsequent search operations can quickly locate relevant data entries. Index tags can be the keywords themselves or semantic extensions related to the keywords. For example, irrelevant search terms (such as personal pronouns and other words that do not affect search results) are identified and filtered out, and the semantic relationships of the keywords are optimized to ensure the accuracy and effectiveness of the index tags.

[0070] It should be noted that, in the embodiments of this application, the words "exemplary" or "for example" indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0071] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0072] Figure 5 This is a schematic block diagram of the medical image data retrieval device 500 provided in an embodiment of this application. Figure 5 As shown, the device includes a request processing module 501, a query construction module 502, and a result acquisition module 503.

[0073] Request processing module 501: used to obtain the image retrieval request input by the user; input the image retrieval request into the medical language recognition model to perform semantic parsing operation to generate a set of user intent features.

[0074] Query construction module 502: is used to convert the set of user intent features into database query instructions based on preset standards, and to perform query operations in the medical image database based on the database query information to obtain raw retrieval data.

[0075] Result acquisition module 503: Based on the image retrieval request, performs corresponding preprocessing operations on the original retrieval data to generate standard medical image retrieval results.

[0076] It should be understood that the specific process of each module performing the above-mentioned steps has been described in detail in the above method embodiments, and will not be repeated here for the sake of brevity.

[0077] It should also be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division; in actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0078] Figure 6 This is a schematic block diagram of the electronic terminal provided in an embodiment of this application. Figure 6 As shown, the electronic terminal includes: at least one processor 601, a memory 602, at least one network interface 603, and a user interface 605. The various components in the device are coupled together via a bus system 604. It is understood that the bus system 604 is used to implement communication between the components. In addition to a data bus, the bus system 604 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 6 The general will label all buses as bus systems.

[0079] The user interface 605 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0080] It is understood that memory 602 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memory described in the embodiments of this invention is intended to include, but is not limited to, any other suitable class of memory.

[0081] In this embodiment of the invention, the memory 602 is used to store various types of data to support the operation of the electronic terminal 600. Examples of data include: any executable program for operation on the electronic terminal 600, such as the operating system 6021 and application program 6022; the operating system 6021 includes various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 6022 may include various applications, such as media players, browsers, etc., for implementing various application services. The medical image data retrieval method provided in this embodiment of the invention can be included in the application program 6022.

[0082] The methods disclosed in the above embodiments of the present invention can be applied to processor 601, or implemented by processor 601. Processor 601 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 601 or by instructions in the form of software. The processor 601 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 601 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. General-purpose processor 601 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0083] In an exemplary embodiment, the electronic terminal 600 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0084] According to the method provided in the embodiments of this application, this application also provides a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute the medical image data retrieval method of any of the embodiments shown above.

[0085] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when run on a computer, causes the computer to perform the medical image data retrieval method of any of the embodiments shown above.

[0086] As used in this specification, the terms "component," "module," "system," etc., are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. As illustrated, applications running on computing devices and computing devices can both be components. One or more components may reside in a process and / or an execution thread, and components may be located on a single computer and / or distributed among two or more computers. Furthermore, components may execute from various computer-readable media on which various data structures are stored. Components may communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0087] Those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether a function is implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0088] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0089] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0090] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0092] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. A computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0093] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0094] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0095] In summary, this application provides a medical image data retrieval method, device, medium, program product, and terminal. Through a medical language recognition model, it achieves semantic understanding of user retrieval requests. The user-input retrieval request is semantically parsed using the medical language recognition model to generate a set of user intent features. This set of user intent features is converted into a database query instruction according to preset standards, and the query is executed in the medical image database to obtain matching raw retrieval data. Based on user type requirements, preprocessing operations are performed on the raw retrieval data to generate standardized medical image retrieval results. This application overcomes the limitations of traditional medical image retrieval and solves the problems of low retrieval accuracy and poor user experience in existing medical image retrieval systems. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0096] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for retrieving medical image data, characterized in that, include: Obtain the image retrieval request input by the user; The image retrieval request is input into a medical language recognition model to perform semantic parsing operations, thereby generating a set of user intent features; The user intent feature set is converted into a database query instruction based on a preset standard, and a query operation is performed in the medical imaging database based on the database query information to obtain the raw retrieval data. Based on the image retrieval request, corresponding preprocessing operations are performed on the original retrieval data to generate standard medical image retrieval results.

2. The medical image data retrieval method according to claim 1, characterized in that, The process of inputting the image retrieval request into a medical language recognition model to perform semantic parsing to generate a set of user intent features includes: Identify medical entity information in the image retrieval request, perform standardization processing on the identified medical entity information, and insert semantic tag information into the standardized medical entities to generate medical structured entities; Semantic encoding is performed on each medical entity in the medical structured entity according to the pre-trained model to generate a semantic vector; the relationships between each medical entity are extracted from the medical knowledge graph based on the medical structured entity, and the corresponding relationship vectors are constructed. Based on the semantic vector and the relation vector, the set of user intent features is generated.

3. The medical image data retrieval method according to claim 2, characterized in that, The process of converting the set of user intent features into a database query command based on a preset standard includes: Construct a query structure based on the aforementioned medical image database; Based on the semantic vector and the relation vector, parent / child search conditions are sequentially inserted into the query structure in a preset multi-level nested structure to generate a database query instruction.

4. The medical image data retrieval method according to claim 1, characterized in that, The image retrieval request includes the query user type; The process of performing corresponding preprocessing operations on the original search data based on the image retrieval request to generate standard medical image retrieval results includes: In response to the query user type being a developer, normalization and standardization operations are performed on the labeled data in the original retrieval data, and pixel value adjustment, resampling, and rescaling operations are performed on the image data in the original retrieval data to generate standard medical image retrieval results.

5. The medical image data retrieval method according to claim 1, characterized in that, The medical image database includes medical image data, as well as detection annotation information, segmentation annotation information, and key point annotation information generated by performing data annotation operations based on the CT image data.

6. The medical image data retrieval method according to claim 1, characterized in that, The method also performs the following steps: Based on multiple medical image retrieval requests entered by the user, the frequency of occurrence of each search keyword in the medical image retrieval requests is statistically analyzed; For search terms that appear more frequently than a preset threshold, corresponding index tags are created in the medical imaging database.

7. A medical image data retrieval device, characterized in that, include: Request processing module: used to obtain image retrieval requests input by the user; input the image retrieval requests into the medical language recognition model to perform semantic parsing operations to generate a set of user intent features; Query construction module: used to convert the user intent feature set into a database query instruction based on a preset standard, and to perform a query operation in the medical image database based on the database query information to obtain the original retrieval data; Result acquisition module: Based on the image retrieval request, it performs corresponding preprocessing operations on the original retrieval data to generate standard medical image retrieval results.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the medical image data retrieval method according to any one of claims 1 to 6.

9. A computer program product, characterized in that, The computer program product includes computer program code, which, when run on a computer, causes the computer to implement the medical image data retrieval method as described in any one of claims 1 to 6.

10. An electronic terminal, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the medical image data retrieval method according to any one of claims 1 to 6.

Images