Text processing-based AI auxiliary diagnosis method, apparatus and device, and medium

Abstract

The invention relates to an AI auxiliary diagnosis method and device based on text processing, equipment and a medium. The method comprises the following steps: acquiring medical text data, constructing a secure channel based on data characteristics to generate a data load, processing the data load through an encryption algorithm, sending the data load to a target interface, and analyzing response data to obtain an original medical text set; cleaning the original set by adopting a multilayer regular filtering system of a text processing engine, inputting a semantic analysis model to calculate a feature association value, generating a semantic classification tag when the feature association value accords with a preset threshold value, constructing a classification index database, and extracting potential semantic association features to obtain an extended semantic feature set; and finally, classifying and marking the extended feature set, arranging and determining a content structure through a layered content organizer, and dynamically updating and generating a final output document. By adopting the method, the security of data transmission and storage can be ensured, the specialty of medical text cleaning and semantic analysis is improved, and automatic conversion from semantics to document structures is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computer software application, and particularly relates to an AI auxiliary diagnosis method, device, equipment and medium based on text processing. BACKGROUND

[0002] In the medical field, accurate and efficient diagnosis is of great significance to disease treatment and patient prognosis. Currently, doctors face many challenges in the diagnosis process, with a misdiagnosis rate of about 20% at the grassroots level, and traditional auxiliary diagnosis methods have problems such as dependence on structured input and complex training, resulting in a significant gap between medical prediction algorithms and clinical actual application. With the development of artificial intelligence technology, pre-training language models have promoted the progress of natural language processing (NLP), and their application in the biomedical field has gradually expanded. However, existing medical large language models (LLMs) lack in clinical scene design, and their diagnosis reasoning ability needs to be further explored. There are also problems such as hallucinations, factual errors, and incomplete evaluation. For example, some models can process medical text, but they are difficult to accurately simulate the diagnosis reasoning process of doctors. At the same time, medical texts in medical data, such as electronic health records (EHRs) and clinical notes, have accumulated a large amount of information, providing rich resources for disease diagnosis. However, these unstructured text data contain a lot of noise and redundant information, making it difficult to extract effective features and build reliable disease diagnosis models. Some existing technologies only focus on single text analysis or lack of office scene fusion. SUMMARY

[0003] Therefore, it is necessary to provide an AI auxiliary diagnosis method, device, equipment and medium based on text processing, which can realize the whole process processing of medical text and improve the reliability of diagnosis assistance.

[0004] In a first aspect, the application provides an AI auxiliary diagnosis method based on text processing, comprising:

[0005] Obtaining medical text data, constructing a secure channel based on the data and generating data payload, sending to a target interface after encryption algorithm processing to obtain an original medical text set.

[0006] The original medical text set is filtered and cleaned by multi-layer regularization, and the feature correlation value is obtained by inputting the semantic analysis model. If the feature correlation value meets the preset threshold, an extended semantic feature set is generated.

[0007] The extended semantic feature set is classified and labeled by color coding based on semantics, and the final content structure is determined by hierarchical arrangement to dynamically update and generate a final output document.

[0008] In one embodiment, constructing a secure channel based on the data and generating data payload, sending to a target interface after encryption algorithm processing to obtain an original medical text set, comprises:

[0009] configure security protocol parameters based on medical text data characteristics, and construct security channel information.

[0010] Parse the protocol specification in the security channel information, and encapsulate the original data payload based on the medical text data content.

[0011] The original data payload is processed using a preset encryption algorithm to obtain an encrypted data payload.

[0012] Based on the encrypted data payload, a request is initiated to the target interface to obtain response data.

[0013] The response data is parsed, and the complete original medical text set is extracted and integrated.

[0014] In one embodiment, the original medical text set is filtered and cleaned using a multi-layer regular filter, and a feature correlation value is obtained by inputting a semantic analysis model; if the feature correlation value meets a preset threshold, an extended semantic feature set is generated, including:

[0015] The original medical text set is cleaned using a multi-layer regular filter system of a text processing engine to obtain a cleaned text data set.

[0016] Based on the cleaned text data set, a semantic analysis model is inputted to calculate a feature correlation value.

[0017] If the feature correlation value meets a preset threshold, a corresponding semantic classification label is generated.

[0018] According to the semantic classification label, a classification index database is constructed to determine the semantic classification direction of the text.

[0019] Based on the classification index database, potential semantic correlation features are extracted to obtain an extended semantic feature set.

[0020] In one embodiment, the feature correlation value is calculated by the following formula:

[0021]

[0022] where S 关联 represents the feature correlation value, represents the word vector of the i-th medical term, represents the context semantic vector of the current text segment, ω i represents the weight coefficient of the i-th medical term.

[0023] In one embodiment, the extended semantic feature set is classified and labeled using semantic-based color coding, the final content structure is determined by hierarchical arrangement, and the final output document is dynamically updated and generated, including:

[0024] The semantic feature set is classified and labeled by a semantic color coding module based on an intelligent formatting module adopting a smart format.

[0025] The semantic feature data is arranged by a hierarchical content organizer to determine a content structure and generate preliminary content framework information.

[0026] The preliminary content framework information is dynamically updated to generate structure data and output structured document content.

[0027] The structured document content is extracted by semantic association analysis to obtain a semantic association matrix.

[0028] The logical level of the structured document content is optimized by the semantic association matrix to generate an optimized final output document.

[0029] In one embodiment, the semantic association matrix is calculated by the following formula:

[0030]

[0031] wherein M represents the semantic association matrix, represents a vector set of key semantic units in the structured document, represents the context semantic feature of the i-th semantic unit, TF-IDE(·) represents a term frequency-inverse document frequency matrix, Attention(·,·,·) represents a self-attention mechanism, and σ represents an activation function.

[0032] In a second aspect, the present application also provides an AI-assisted diagnosis device based on text processing, which comprises:

[0033] A data security module is configured to obtain medical text data, construct a secure channel based on the data and generate data payloads, and send the data payloads to a target interface after encryption algorithm processing to obtain an original medical text set.

[0034] A semantic processing module is configured to adopt multi-layer regular filtering and cleaning on the original medical text set and input a semantic analysis model to obtain a feature correlation value; if the feature correlation value meets a preset threshold, an extended semantic feature set is generated.

[0035] An intelligent formatting module is configured to utilize semantic-based color coding classification labeling on the extended semantic feature set, determine a final content structure through hierarchical arrangement, and dynamically update to generate a final output document.

[0036] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the foregoing method when executing the computer program.

[0037] In a fourth aspect, the present application also provides a computer readable storage medium, having stored thereon a computer program, which when executed by a processor implements the method as above.

[0038] The AI-assisted diagnosis method and device based on text processing, computer device and storage medium described above, after obtaining medical text data, a secure channel is constructed based on data characteristics to generate data payloads, which are sent to a target interface after being processed by an encryption algorithm, and the original medical text set is obtained by analyzing the response data; the original set is cleaned by a multi-layer regular filtering system of a text processing engine, the feature correlation value is calculated by inputting the semantic analysis model, the semantic classification label is generated when the preset threshold is met, the classification index database is constructed and the potential semantic correlation features are extracted, and the expanded semantic feature set is obtained; finally, the semantic color coding module of the intelligent formatting module is used to classify and label the expanded feature set, the content structure is determined by arranging the hierarchical content organizer, and the final output document is dynamically updated and generated. The method realizes the whole process of medical text processing through modular design, the security architecture ensures the security of data transmission and storage, the text processing engine improves the professionalism of medical text cleaning and semantic analysis, and the intelligent formatting realizes the automatic conversion of semantics to document structure. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiment or related art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0040] Figure 1 The flowchart of the AI-assisted diagnosis method based on text processing provided by the embodiment of the present application;

[0041] Figure 2 The structural block diagram of the AI-assisted diagnosis device based on text processing provided by the embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0043] In one embodiment, as shown in Figure 1 The AI-assisted diagnosis method based on text processing provided by the present application can include the following steps:

[0044] Step S101, obtain medical text data, construct a secure channel based on the data and generate data payload, send to the target interface after encryption algorithm processing to get the original medical text set.

[0045] Specifically, after obtaining the medical text data, configure the security protocol parameters based on the data characteristics, construct the security channel information through the security HTTPS protocol processor of the API communication module, parse the protocol specification, use the JSON payload constructor to encapsulate the medical text data into the original data payload, and then use the preset encryption algorithm to process to get the encrypted data payload. Based on this, request is initiated to the target interface and the response data is parsed, and the complete original medical text set is extracted and integrated. The process ensures the security and integrity of data transmission through the API key encryption subsystem of the security architecture and the local text purification layer, avoiding sensitive information leakage or data tampering.

[0046] Step S102, the original medical text set is filtered and cleaned by multiple regular expressions and input into a semantic analysis model to obtain a feature correlation value; if the feature correlation value meets the preset threshold, an extended semantic feature set is generated.

[0047] Specifically, for the original medical text set, the multi-layer regular expression filtering system of the text processing engine is used for cleaning to remove noise, garbled code and non-medical professional content in the text to obtain a cleaned text data set. The data set is input into a pre-established semantic analysis model to obtain a feature correlation value by calculating the weighted cosine similarity of the medical term word vector and the context semantic vector. If the feature correlation value reaches the preset threshold, the corresponding semantic classification label is generated, and a classification index database is constructed to extract potential semantic correlation features and form an extended semantic feature set. The professional nature of medical text processing and the accuracy of semantic analysis are improved.

[0048] Step S103, the extended semantic feature set is classified and labeled using a semantic-based color coding, and the final content structure is determined by hierarchical arrangement to dynamically update and generate the final output document.

[0049] For the extended semantic feature set, the semantic-based color coding module of the intelligent formatting module is used for classification and labeling, and different semantic categories of features are mapped to visual color identifiers to obtain semantic feature data with color coding. The hierarchical content organizer arranges the labeled data by level and structure to determine the content structure and generate preliminary content framework information. Then, the framework is dynamically updated based on real-time interactive input or semantic correlation analysis results to generate final document structure data and output structured document content. Finally, the semantic correlation matrix is used to optimize the document logical level to ensure that the medical text information is presented in a way that meets the clinical diagnosis logic, realizing the automatic generation of semantic features to diagnostic documents.

[0050] The AI-assisted diagnosis method based on text processing described above, after obtaining medical text data, constructs a secure channel based on data characteristics to generate data payloads, sends them to the target interface after encryption algorithm processing, and obtains the original medical text set by analyzing the response data; the original set is cleaned by using the multi-layer regular filtering system of the text processing engine, the feature correlation value is calculated by inputting the semantic analysis model, the semantic classification label is generated when it meets the preset threshold, the classification index database is constructed and the potential semantic correlation feature is extracted, and the expanded semantic feature set is obtained; finally, the semantic color coding module of the intelligent formatting module is used to classify and label the expanded feature set, the content structure is determined by arranging the hierarchical content organizer, and the final output document is dynamically updated and generated. This method realizes the whole process of medical text processing through modular design, the security architecture ensures the security of data transmission and storage, the text processing engine improves the professionalism of medical text cleaning and semantic analysis, and the intelligent formatting module realizes the automatic conversion of semantics to document structure.

[0051] In one embodiment, constructing a secure channel based on data and generating data payloads, sending them to the target interface after encryption algorithm processing to obtain the original medical text set, can include the following steps:

[0052] Step S201, configure security protocol parameters based on medical text data characteristics to construct secure channel information.

[0053] Further, the medical text data characteristics refer to the attributes of the medical text data itself, including data type (such as electronic medical record, diagnosis report, medical literature, etc.), sensitivity (containing patient privacy information), format specification (unstructured text is the main one, mixed with professional terms), and data size, etc.

[0054] The security protocol parameters are a set of parameters for constructing a secure communication channel, including encryption algorithm type (such as AES-256), key length, authentication method (such as API key verification), data transmission protocol (such as HTTPS), etc.

[0055] Step S202, parse the protocol specification in the secure channel information, and encapsulate the original data payload in combination with the medical text data content.

[0056] Preferably, the protocol specification is the communication rule defined in the secure channel information, including data format, transmission process, error handling mechanism, etc. Parsing the protocol specification is to ensure that the encapsulation of the data payload meets the receiving requirements of the target interface, avoiding transmission failure due to format incompatibility.

[0057] Step S203, process the original data payload using a preset encryption algorithm to obtain an encrypted data payload.

[0058] Preferably, the encryption algorithm matching the current medical text data security level is called from the key management library by the API key encryption subsystem in the security architecture, such as the AES-256 encryption algorithm used in the equivalent 256-bit encryption security framework. The original data payload containing the original text content, metadata (such as data type identification) and necessary protocol header information is input into the key encryption subsystem for byte-by-byte encryption processing, and the local text purification layer is triggered synchronously during encryption to desensitize sensitive information in the data, such as replacing private data such as patient name and ID number with specific placeholders. After encryption, the encrypted data payload with encryption check code is generated, which is used for integrity verification in subsequent data transmission process to ensure the security and confidentiality of the data before transmission to the target interface, effectively preventing medical text data from being stolen or tampered with during network transmission, and meeting the requirements of medical data security standards for data encryption transmission.

[0059] Step S204, based on the encrypted data payload, a request is initiated to the target interface to obtain response data.

[0060] Step S205, the response data is parsed to extract and integrate the complete original medical text set.

[0061] Specifically, based on the characteristics of medical text data, the security protocol parameters are configured, the security channel information is constructed, the protocol specification is parsed, and the medical text data content is combined to encapsulate and generate the original data payload through the JSON payload constructor. The original data payload is processed by a preset encryption algorithm to obtain the encrypted data payload, and then a request is initiated to the target interface based on the encrypted data payload to obtain response data and parse it to extract and integrate the complete original medical text set.

[0062] The API key encryption subsystem in the security architecture and the local text purification layer are used to realize the whole process security protection of medical text data from constructing a security channel to generating an original set, ensuring that the data is not tampered with or leaked during transmission. Compared with the prior art, the security HTTPS protocol processor and the response verification system are used to improve the reliability and integrity of medical text data in network transmission, providing a secure and complete data source for subsequent text processing, while meeting the strict requirements of medical data for security and compliance.

[0063] In one embodiment, the original medical text set is filtered and cleaned by multiple layers of regular expressions and input into a semantic analysis model to obtain a feature correlation value; if the feature correlation value meets a preset threshold, an extended semantic feature set is generated, which can include the following steps:

[0064] Step S301, a multi-layer regular expression filtering system of a text processing engine is used to clean the original medical text set to obtain a cleaned text data set.

[0065] Preferably, the text processing engine includes a multi-layer regular expression filtering system, a medical term dedicated processor, and a context-aware segmentation unit, which are responsible for cleaning, semantic analysis, and segmentation processing of medical text. The multi-layer regular expression filtering system cleans the original medical text through multi-layer regular expression rules to remove garbled codes, non-medical terms, and noise data, thereby improving the quality of the text data.

[0066] Step S302, based on the cleaned text data set, a pre-established semantic analysis model is input to calculate the feature correlation value.

[0067] Further, with the help of the medical term dedicated processor in the text processing engine, the cleaned text data is converted into a numerical representation containing medical term word vectors, each word vector representing the semantic features (such as disease type, symptoms, examination indicators, etc.) of the corresponding term. At the same time, the context semantic vector of the current text segment is extracted through the context-aware segmentation unit to capture the semantic dependency relationship between the terms in the text. The word vector and the context semantic vector are input into the pre-set semantic analysis model, and the weighted cosine similarity algorithm is used to calculate the correlation degree of the two, wherein the weight coefficient is determined by the importance of the term in the medical field (such as frequency of occurrence, clinical value, etc.). Finally, the feature correlation value representing the correlation strength of the medical term and the context semantics is obtained. The feature correlation value is a quantitative indicator of the correlation strength of the medical term and the context semantics, which is calculated by the weighted cosine similarity algorithm and is used to determine whether the correlation between the term and the text context meets the pre-set standard.

[0068] Step S303, if the feature correlation value meets the pre-set threshold, the corresponding semantic classification label is generated.

[0069] Preferably, the pre-set threshold is a critical value of the feature correlation value set by humans. If the calculated feature correlation value reaches or exceeds the threshold, it is considered that the corresponding medical term and the context semantics have effective correlation, triggering the generation of the semantic classification label.

[0070] Step S304, according to the semantic classification label, a classification index database determining the semantic classification direction of the text is constructed.

[0071] Further, the medical terms or text segments that meet the threshold requirement are assigned a semantic category identifier (such as "disease diagnosis", "treatment plan", "examination indicator", etc.) to standardize the semantic classification direction of the text.

[0072] Step S305, based on the classification index database, potential semantic correlation features are extracted to obtain an expanded semantic feature set.

[0073] The classification index database is a database constructed based on semantic classification labels, stores the mapping relationship between text segments and semantic categories, supports fast retrieval and semantic association feature extraction, and provides structured data support for the generation of an extended semantic feature set.

[0074] Specifically, for an original medical text collection, a multi-layer regular filtering system of a text processing engine is used to clean up the noise, garbled codes and non-professional content therein, to obtain a cleaned text dataset. The dataset is input into a pre-established semantic analysis model to obtain a feature association value by calculating the weighted cosine similarity of the medical term word vector and the context semantic vector; if the feature association value meets a preset threshold, a corresponding semantic classification label is generated, and a classification index database is constructed therefrom to extract potential semantic association features and obtain an extended semantic feature set.

[0075] The multi-layer regular filtering system of the text processing engine improves the quality of the medical text data, providing a clean and professional data source for semantic analysis. The semantic analysis model combines a medical term dedicated processor and a context-aware segmentation unit to accurately calculate the feature association value and generate semantic classification labels, enabling the medical text data to have a structured semantic classification direction. The construction of the classification index database and the extraction of potential semantic association features effectively solve the problem of the lack of professionalism in the processing of medical text by existing similar products, improving the accuracy and efficiency of medical text analysis.

[0076] In one embodiment, the feature association value can be calculated by the following formula:

[0077]

[0078] wherein S 关联 represents the feature association value, represents the word vector of the i-th medical term, represents the context semantic vector of the current text segment, ω i represents the weight coefficient of the i-th medical term.

[0079] The medical term dedicated processor in the text processing engine generates the word vector, combined with the context semantic vector extracted by the context-aware segmentation unit, to ensure that the feature association value accurately reflects the semantic relevance of the medical term in the specific context. The weight coefficient can be set based on factors such as the clinical importance and frequency of occurrence of the medical term, making the calculation result more consistent with medical professional logic. The feature association value obtained by the formula can effectively filter out medical terms closely related to the context semantics, providing a quantitative basis for subsequent generation of semantic classification labels and construction of a classification index database, solving the problem of the lack of professionalism in the semantic analysis of medical text by existing similar products, and improving the accuracy and efficiency of medical text processing.

[0080] In one embodiment, the extended semantic feature set is classified and labeled using semantic-based color coding, the final content structure is determined through hierarchical arrangement, and the final output document is dynamically updated and generated, which can include the following steps:

[0081] Step S401, based on the extended semantic feature set, the semantic color coding module of the intelligent formatting module is used for classification and labeling to obtain visual semantic feature data.

[0082] Specifically, based on the extended semantic feature set, the semantic color coding module of the intelligent formatting module is used for classification and labeling to obtain visual semantic feature data. Specifically, according to the category of semantic features (such as disease diagnosis, treatment plan, examination index, etc.), the module assigns corresponding color labels to different semantic units, for example, marks "disease name" as red and "drug recommendation" as blue. Through this visual method, abstract semantic features are converted into intuitive color-coded data, which facilitates the organization and analysis of subsequent content structure, and also provides a more intuitive semantic understanding method for medical professionals, improving the efficiency and accuracy of document processing.

[0083] Step S402, the semantic feature data is arranged using the arrangement algorithm of the hierarchical content organizer to determine the content structure and generate preliminary content framework information.

[0084] Step S403, the preliminary content framework information is dynamically updated to generate structure data and output structured document content.

[0085] Step S404, the structured document content is extracted through semantic association analysis to obtain a semantic association matrix.

[0086] Preferably, the structured document content data generated after dynamic update has a clear hierarchical structure and semantic classification, and is output as a document conforming to the WPS native format through the WPS Office extension framework, supporting subsequent editing and format saving.

[0087] Specifically, the structured document content is extracted through semantic association analysis to obtain a semantic association matrix. Using natural language processing technology, the association between semantic units in the document is analyzed, for example, the treatment association between "hypertension" and "hypotensive drugs". By calculating the association strength between semantic units, a matrix is constructed to represent their relationship, and each element value in the matrix represents the association degree of the corresponding semantic unit. This semantic association matrix can clearly present the logical context of the semantic units in the document, ensuring that the finally generated document content conforms to the logical order and professional requirements of medical diagnosis.

[0088] Step S405, the logical level of the structured document content is optimized by using the semantic association matrix to generate the optimized final output document.

[0089] Specifically, based on the extended semantic feature set, the semantic color coding module of the intelligent formatting module is used for classification labeling to obtain visual semantic feature data; the arrangement algorithm of the hierarchical content organizer is used on the data to determine the content structure and generate preliminary content framework information, which is then dynamically updated to generate structure data and output structured document content. Finally, key semantic units are extracted from the structured document through semantic association analysis to obtain a semantic association matrix, which is used to optimize the logical level of the document to generate the optimized final output document.

[0090] The semantic color coding module and the hierarchical content organizer of the intelligent formatting module realize the automatic conversion from semantic features to document structure, solving the problem of text generation with garbled code in existing products. The dynamic updating mechanism ensures that the content framework is synchronized with the semantic analysis results in real time, improving the accuracy of document generation. The application of the semantic association matrix optimizes the logical level of the document, making the medical text content more consistent with the clinical diagnosis logic. In addition, combined with the WPS Office extension framework, the final output document can maintain the original WPS format and meet the medical data security requirements through the security architecture, reducing the manual formatting time and improving the office efficiency compared with traditional methods.

[0091] In one of the embodiments, the semantic association matrix can be calculated by the following formula:

[0092]

[0093] wherein M represents the semantic association matrix, represents the vector set of key semantic units in the structured document, represents the context semantic features of the i-th semantic unit, TF-IDE(·) represents the term frequency-inverse document frequency matrix, Attention(·,·,·) represents the self-attention mechanism, and σ represents the activation function.

[0094] Preferably, the self-attention mechanism calculates the association degree between semantic units through query-key-value triple mapping:

[0095]

[0096] wherein, d k represents the vector dimension, and the output matrix represents the dependency relationship between semantic units (e.g., high association degree between “insulin injection” and “hypoglycemia prevention”).

[0097] The embodiment can effectively capture long-distance semantic dependencies in medical text (such as the association between disease symptoms and treatment plans) by dynamically calculating the association weight between semantic units through the self-attention mechanism, and can make the semantic association matrix more consistent with the clinical diagnosis logic by combining the quantification of the importance of medical terms by the TF-IDF matrix. For example, when processing semantic units such as "high blood pressure" and "hypotensive drugs", the treatment association strength can be accurately calculated. The generated semantic association matrix can optimize the logical hierarchy of the document, so that the final output diagnosis document is arranged in a professional order of "symptom description -> etiology analysis -> treatment recommendation", which reduces the manual formatting time compared with the prior art, and improves the medical professionalism of the document through highly accurate context-aware analysis, thereby meeting the efficient document generation and security requirements in the WPS Office office scenario.

[0098] In one of the embodiments, as shown in Figure 2 The application also provides an AI-assisted diagnosis device based on text processing, which can include:

[0099] The data security module 501 is configured to obtain medical text data, construct a secure channel based on the data and generate data payloads, and send the data payloads to a target interface after encryption algorithm processing to obtain an original medical text set.

[0100] The semantic processing module 502 is configured to perform multi-layer regular filtering and cleaning on the original medical text set and input a semantic analysis model to obtain a feature association value; if the feature association value meets a preset threshold, an extended semantic feature set is generated.

[0101] The intelligent formatting module 503 is configured to use semantic-based color coding classification labeling on the extended semantic feature set, determine a final content structure through hierarchical arrangement, and dynamically update and generate a final output document.

[0102] The above-mentioned AI auxiliary diagnosis device based on text processing, the data security module is responsible for obtaining medical text data, constructing a secure channel based on data characteristics and generating data payload, sending to the target interface after encryption algorithm processing, so as to obtain the original medical text set. The semantic processing module carries out multi-layer regular filtering and cleaning to the original set, inputs the cleaned data into the semantic analysis model to calculate the feature correlation value, and generates an extended semantic feature set when the feature correlation value meets the preset threshold. The intelligent formatting module carries out semantic color coding classification labeling on the extended semantic feature set, determines the content structure through hierarchical arrangement and dynamically updates, and finally generates an output document. The data security module ensures the security of medical text data transmission and storage through a secure HTTPS protocol processor, an API key encryption subsystem and the like, and solves the data leakage risk problem of existing products. The semantic processing module improves the professionalism of medical text processing and the accuracy of feature correlation value calculation by means of a multi-layer regular filtering system and context-aware parsing technology, effectively avoiding problems such as garbled code caused by non-professional processing. The intelligent formatting module combines WPS native format protection technology to realize automatic conversion of semantics to document structure, reduces manual formatting time, and optimizes the logical level of the document through a semantic correlation matrix, so that the final output document meets the clinical diagnosis logic and meets the dual needs of medical data security and office efficiency.

[0103] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0104] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the AI auxiliary diagnosis method, device, equipment and medium based on text processing as described above when executing the computer program.

[0105] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps in each method embodiment described above.

[0106] For the device embodiment, since it basically corresponds to the method embodiment, the relevant part can be seen from the part of the method embodiment. The device embodiment described above is only schematic, wherein the components shown as separate components can or can not be physically separate, and the components shown as a unit can or can not be a physical unit, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement without creative labor.

[0107] The above-described embodiments only express several implementation manners of the present application, which are described in detail, but should not be understood as a limitation on the patent scope of the application. It should be pointed out that, for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. AI-assisted diagnosis method based on text processing, characterized in that: The method comprises: Obtain medical text data, build a secure channel based on the data and generate a data payload, which is then processed by an encryption algorithm and sent to the target interface to obtain the original medical text collection; The original medical text set is cleaned by multi-layer regularization filtering and input into a semantic analysis model to obtain a feature association value; if the feature association value meets a preset threshold, an extended semantic feature set is generated; The extended semantic feature set is classified and annotated using semantic-based color coding, and the final content structure is determined through hierarchical arrangement, and the final output document is dynamically updated.

2. The method according to claim 1, characterized in that The data-based secure channel is constructed and a data payload is generated, which is then processed by an encryption algorithm and sent to a target interface to obtain a set of original medical texts, including: Configuring security protocol parameters based on the characteristics of the medical text data and constructing security channel information; Parsing the protocol specification in the secure channel information, combining it with the medical text data content and encapsulating it to generate an original data payload; Processing the original data payload using a preset encryption algorithm to obtain an encrypted data payload; Initiate a request to the target interface based on the encrypted data payload and obtain response data; The response data is parsed, extracted and integrated to obtain a complete set of original medical texts.

3. The method according to claim 1, characterized in that The original medical text collection is cleaned by multi-layer regular filtering and input into a semantic analysis model to obtain feature association values; If the feature association value meets the preset threshold, an extended semantic feature set is generated, including: Cleaning the original medical text collection using a multi-layer regular filtering system of a text processing engine to obtain a cleaned text dataset; Inputting a pre-established semantic analysis model based on the cleaned text data set to calculate feature association values; If the feature association value meets the preset threshold, a corresponding semantic classification label is generated; Constructing a classification index database for determining the semantic classification direction of the text according to the semantic classification labels; Potential semantic association features are extracted based on the classification index database to obtain an extended semantic feature set.

4. The method according to claim 1, wherein The feature association value is calculated using the following formula: Among them, S 关联 represents the feature association value, Represents the word vector of the i-th medical term, Represents the contextual semantic vector of the current text segment, ω i represents the weight coefficient of the i-th medical term.

5. The method according to claim 1, wherein The method of labeling the extended semantic feature set using semantically based color coding, determining the final content structure through hierarchical arrangement, and dynamically updating to generate the final output document includes: Based on the extended semantic feature set as data, the semantic color coding module of the intelligent formatting module is used for classification and annotation to obtain visual semantic feature data; Using an arrangement algorithm of a hierarchical content organizer on the semantic feature data, determining a content structure and generating preliminary content framework information; Dynamically updating the preliminary content framework information, generating structure data and outputting structured document content; Extracting key semantic units from the structured document content through semantic association analysis to obtain a semantic association matrix; The semantic association matrix is ​​used to optimize the logical hierarchy of the structured document content to generate an optimized final output document.

6. The method according to claim 5, characterized in that The semantic association matrix is ​​calculated by the following formula: Where M represents the semantic association matrix, A vector set representing key semantic units in a structured document, represents the contextual semantic features of the i-th semantic unit, TF-IDE(·) represents the term frequency-inverse document frequency matrix, Attention(·,·,·) represents the self-attention mechanism, and σ represents the activation function.

7. AI-assisted diagnosis device based on text processing, characterized in that: The device comprises: The data security module is used to obtain medical text data, build a secure channel based on the data, generate a data payload, and send it to the target interface after encryption algorithm processing to obtain the original medical text collection; A semantic processing module is used to clean the original medical text set using multi-layer regularization filtering and input it into a semantic analysis model to obtain a feature association value; if the feature association value meets a preset threshold, an extended semantic feature set is generated; The intelligent formatting module is used to classify and annotate the extended semantic feature set using semantic-based color coding, determine the final content structure through hierarchical arrangement, and dynamically update to generate the final output document.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images