Medical case intelligent matching method and system based on artificial intelligence

Abstract

The embodiment of the invention relates to the technical field of information, in particular to a medical case intelligent matching method and system based on artificial intelligence. The method comprises the following steps: reading historical medical cases; identifying and intercepting a key local image in the diagnosis and treatment picture, and generating and associating local image description; identifying and intercepting key segments of the diagnosis and treatment video, and generating and associating segment descriptions; according to the disease description, the diagnosis and treatment process description, the diagnosis and treatment data, the key local image, the local image description, the key segment and the segment description, generating medical case knowledge items; generating extended medical case knowledge items by increasing diseases; constructing a medical case library according to the medical case knowledge items and the extended medical case knowledge items; receiving disease query description to be queried, and respectively calculating the matching degree between the disease description of each medical case knowledge item and the disease query description; and taking the plurality of medical case knowledge items with the highest matching degree as a matching result.

Description

Technical Field

[0001] Several embodiments of this specification relate to the field of information technology, specifically to an artificial intelligence-based intelligent matching method and system for medical cases. Background Technology

[0002] Traditional medical case retrieval systems, largely based on keyword matching or structured database queries, suffer from weak semantic understanding and insufficient processing capabilities for unstructured data (such as images, videos, and free text descriptions). This results in low relevance of search results, failing to meet clinicians' needs for accurate and comprehensive diagnostic information. Existing systems typically only statically store and index cases, lacking in-depth analysis of diagnostic images and videos. For example, medical images may contain multiple key lesion areas, and surgical videos involve multiple key operational steps. However, traditional methods often treat the entire image or video as a whole, failing to achieve fine-grained information extraction and correlation, thus limiting the accuracy and practicality of case matching. With the increasingly widespread application of artificial intelligence (AI) technology in the healthcare field, particularly in areas such as assisted diagnosis, clinical decision support, and medical education, it is necessary to research AI-based medical case matching solutions. Summary of the Invention

[0003] This specification describes a method and system for intelligent matching of medical cases based on artificial intelligence through several embodiments.

[0004] Firstly, embodiments of this specification provide an intelligent medical case matching method based on artificial intelligence, including the following steps:

[0005] Read historical medical cases, which include disease descriptions, treatment process descriptions, desensitized treatment data, treatment images, and treatment videos;

[0006] Identify and extract key local images from the diagnostic images, and generate and associate local image descriptions;

[0007] Identify and extract key segments from the medical video, and generate and associate segment descriptions;

[0008] Based on the description of symptoms, the description of the diagnosis and treatment process, the diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions, generate medical case knowledge items;

[0009] By adding symptoms and modifying the symptom descriptions, diagnosis and treatment process descriptions, and diagnosis and treatment data accordingly, as well as adding key local images, local image descriptions, key segments, and segment descriptions, extended medical case knowledge items are generated.

[0010] A medical case library is constructed based on medical case knowledge items and extended medical case knowledge items;

[0011] Receive the description of the symptom to be queried, and calculate the matching degree between the symptom description of each medical case knowledge item and the symptom query description;

[0012] The system uses the medical case knowledge items with the highest matching degree as the matching results, and responds to the user's operation to display the description of the diagnosis and treatment process, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions.

[0013] Secondly, embodiments of this specification provide an intelligent medical case matching system based on artificial intelligence, comprising:

[0014] The reading module reads historical medical cases, which include disease descriptions, treatment process descriptions, desensitized treatment data, treatment images, and treatment videos.

[0015] The first recognition module identifies and extracts key local images from the diagnostic images, and generates and associates local image descriptions.

[0016] The second identification module identifies and extracts key segments from the medical video, and generates and associates segment descriptions.

[0017] The generation module generates medical case knowledge items based on the symptom description, diagnosis and treatment process description, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions;

[0018] The extension module generates extended medical case knowledge items by adding symptoms, modifying the symptom description, diagnosis and treatment process description and diagnosis and treatment data accordingly, and adding key local images, local image descriptions, key segments and segment descriptions accordingly.

[0019] The module constructs a medical case library based on medical case knowledge items and extended medical case knowledge items.

[0020] The query module receives the description of the disease to be queried and calculates the matching degree between the description of the disease and the description of the disease query for each medical case knowledge item.

[0021] The matching module uses the medical case knowledge items with the highest matching degree as the matching results. In response to the user's operation, it displays the description of the diagnosis and treatment process, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions to the user.

[0022] Thirdly, embodiments of this specification provide an electronic device, including a processor and a memory;

[0023] The processor is connected to the memory;

[0024] The memory is used to store executable program code;

[0025] The processor runs a program corresponding to the executable program code stored in the memory to perform the method described in any of the above aspects.

[0026] Fourthly, embodiments of this specification provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the methods described in any of the above aspects.

[0027] Fifthly, embodiments of this specification provide a computer program product, including a computer program that, when executed by a processor, implements the methods described in any of the above aspects.

[0028] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following:

[0029] In several embodiments of this specification, the provided intelligent medical case matching method and system improve the semantic understanding and matching accuracy of medical case retrieval by combining multi-dimensional information such as keyword matching degree, ICD encoding level matching degree, and symptom weight to correct the matching degree. It employs an image recognition model to automatically detect key local areas in diagnostic images and combines this with an image-text generation model to generate local image descriptions, achieving fine-grained analysis of medical image and video data, enhancing the expressive power of case knowledge, and providing more intuitive and specific reference information. It provides intelligent symptom expansion for original cases, automatically generating extended case knowledge items containing reasonable concurrent symptoms, effectively expanding the scale and diversity of the medical case database, making it closer to real complex clinical scenarios, and improving its value in assisting decision-making.

[0030] Other features and advantages of various embodiments of this specification will be further revealed in the following detailed description and accompanying drawings. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of intelligent matching of medical cases provided in the embodiments of this specification.

[0033] Figure 2 This is a schematic diagram of the intelligent medical case matching method provided in the embodiments of this specification.

[0034] Figure 3 An additional symptom diagram is provided for the embodiments of this specification.

[0035] Figure 4 This is a schematic diagram of the intelligent medical case matching system provided in the embodiments of this specification.

[0036] Figure 5 A schematic diagram of an electronic device provided in an embodiment of this specification. Detailed Implementation

[0037] The technical solutions of the embodiments of this specification will be explained and described below with reference to the accompanying drawings. However, the following embodiments are only preferred embodiments of this specification and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments in the implementation methods without creative effort are all within the protection scope of this specification.

[0038] The terms "first," "second," "third," etc., in the description, claims, and accompanying drawings are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0039] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to facilitate the description of the embodiments and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this specification.

[0040] All data involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0041] Before introducing the technical solutions described in this manual, the application scenarios and related technologies of the technical solutions will be introduced.

[0042] In modern clinical medical practice, medical cases serve as an irreplaceable link between theoretical knowledge and practical clinical experience. Each verified medical case encapsulates real-life symptoms, treatment processes, examination data, and imaging information, providing valuable reference for doctors in developing treatment plans and playing a crucial role in medical education, clinical training, and consultations for complex cases. Especially when dealing with rare diseases, complex comorbidities, or atypical symptoms, searching and referencing similar historical cases can effectively assist doctors in broadening their diagnostic approaches and optimizing treatment strategies. Currently, most medical institutions have established electronic medical record systems and medical case databases, supporting case searches by disease name, ICD code, or keywords. However, existing search methods generally have limitations.

[0043] For example, when a doctor sees a patient presenting with "chest pain accompanied by intermittent syncope," they may only be able to match cases explicitly labeled as "coronary artery disease" or "arrhythmia," failing to identify potentially related cases with highly similar clinical presentations, such as certain metabolic diseases or autonomic dysfunction. Furthermore, traditional systems struggle to effectively utilize unstructured diagnostic images14, surgical videos, or free text descriptions, resulting in a large amount of multimodal data containing crucial information being overwhelmed and unable to participate in matching.

[0044] The limitations of retrieval technology restrict the efficiency of reusing medical case knowledge and the realization of its clinical auxiliary value. Therefore, please refer to the appendix. Figure 1 This specification provides an AI-based intelligent medical case matching method and system, capable of extracting multi-source information such as symptom descriptions 11, treatment processes, anonymized data, images, and videos from historical medical cases 10. It automatically identifies key local areas in images and generates descriptions using AI technology, extracts key segments from videos and generates segmented descriptions, and then integrates this fine-grained information to form medical case knowledge items 20. Based on symptom coexistence rules and clinical knowledge graphs, it automatically adds reasonable symptoms and expands their descriptions and multimedia content accordingly, generating expanded case knowledge items and constructing a complete medical case library. It receives user symptom query descriptions 31, calculates their matching degree with the symptom descriptions 11 of each case, and returns the cases with the highest matching degrees as results, supporting the display of their complete treatment processes, data, and key images and video clips.

[0045] This manual first presents an artificial intelligence-based intelligent matching method for medical cases; please refer to the appendix. Figure 2 The steps include:

[0046] Step S1) Read historical medical cases 10, which include a disease description 11, a treatment process description 12, anonymized treatment data 13, treatment images 14, and treatment videos 15. Through interface or file import methods, authorized and compliantly anonymized historical medical cases 10 are read from hospital electronic medical record systems, image archiving and communication systems, laboratory information systems, pathology systems, surgical anesthesia systems, etc. Each historical medical case 10 contains a disease description 11, a treatment process description 12, anonymized treatment data 13, treatment images 14, and treatment videos 15.

[0047] Among them, the description of symptoms (11) refers to the patient's chief complaint, present illness, past medical history, physical examination results, and final diagnosis at the time of consultation. It is usually recorded in natural language and covers symptoms, signs, disease name, and its severity. For example, "The patient has experienced recurrent upper abdominal pain for 3 months, accompanied by acid reflux and belching. Gastroscopy revealed chronic atrophic gastritis with intestinal metaplasia."

[0048] The diagnosis and treatment process description (section 12) records the complete clinical pathway from initial diagnosis to the end of treatment, including diagnostic reasoning, examination arrangements, treatment plans (including medication use and surgical procedures), efficacy evaluation, and follow-up. It reflects the physician's clinical decision-making logic and actual intervention measures and is an important component of case knowledge.

[0049] Desensitized medical data (13) refers to quantitative medical indicators extracted from medical records, such as laboratory test results (e.g., complete blood count, liver and kidney function), vital signs (e.g., blood pressure, heart rate), pathological scores, and genetic testing data. All data undergoes privacy protection processing after being read to ensure that patient identity information (e.g., name, ID number, address, etc.) is completely removed or encrypted, complying with relevant regulations.

[0050] Diagnostic images 14 include various medical image screenshots or static images, such as X-ray films, CT / MRI tomographic images, ultrasound images, endoscopic photographs, skin lesion photographs, pathological slide images, and other images generated during the treatment process. Diagnostic images 14 visually reflect the morphology, location, and changes of lesions and are important evidence for diagnosis.

[0051] Clinical video 15 encompasses videos of treatment processes, including surgical recordings, endoscopic examination videos, dynamic ultrasound images, and rehabilitation training records. Clinical video 15 contains detailed operational procedures and the evolution of the disease over time, exhibiting high information density. Current retrieval systems struggle to effectively index and utilize clinical video 15.

[0052] Step S2) Identify and extract key local images from the diagnostic image 14, and generate and associate local image descriptions.

[0053] The method for identifying and extracting key local images from the diagnostic image 14, and generating and associating local image descriptions includes:

[0054] A pre-trained image recognition model is used to detect key regions in the diagnostic image 14, and key local regions with key pathological features are located in the image. The key pathological features are obtained from a pre-configured pathological reference library.

[0055] The detected key local regions are cropped to form key local images;

[0056] The key local image is input into the connected image and text generation model to generate a local image description.

[0057] Image recognition models, employing either Convolutional Neural Networks (CNNs) or Visual Transformers, can perform lesion localization analysis on the input diagnostic images 14. During training, the image recognition model learns from medical image data annotated with pathological features, enabling it to identify disease-related regions, such as ground-glass nodules in lung CT scans, ulcers in gastroscopy images, the boundaries of pigmented nevi under dermoscopy, and abnormal cell aggregation areas in pathological sections. The definitions of pathological features are derived from a built-in pathological reference library, provided by techniques publicly available in the field. This reference library contains typical imaging manifestations and spatial distribution patterns of various diseases.

[0058] Precise image cropping or masking segmentation is performed on the detected areas to generate key local images. This operation separates the core diagnostic information from background interference in the original image, facilitating subsequent storage, comparison, and display. For the same diagnostic image 14, multiple key local images (such as multiple lesions) can be detected and cropped to achieve fine-grained information extraction. The cropped key local images are input into an integrated image-text generation model (exemplary models such as BLIP-2 or Flamingo). The image-text generation model combines image visual features with medical knowledge to generate a natural language description. As a recommendation, the description includes the location, size, shape, boundary clarity, color features, and surrounding tissue relationships of the lesion. An example is: "A circular ulcer with a diameter of approximately 1.5 cm is visible on the posterior wall of the gastric body, with raised edges, a dirty coating at the base, and surrounding mucosa radiating outwards." The generated key local images are bound to their corresponding local image descriptions using metadata, and an association index is established with the original diagnostic image 14 and the relevant case record.

[0059] Step S3) Identify and extract key segments from the diagnostic video 15, and generate and associate segment descriptions. Specifically, this includes:

[0060] The diagnostic video 15 is subjected to frame sequence extraction. The action detection model is used to identify the time periods containing key diagnostic operations or key symptom changes as key segments. The key diagnostic operations and key lesion changes are obtained from a pre-configured operation reference library and symptom change reference library.

[0061] Key frames are extracted from each key segment, and the key frames are input into the connected image and text generation model to generate image descriptions of the key frames;

[0062] Identify key segments of video content, and generate segment descriptions based on the video content and the image descriptions.

[0063] The diagnostic video 15 is decoded and decomposed into a continuous sequence of image frames at a set frame rate (e.g., 5 frames per second or keyframe extraction). A pre-trained action detection model or temporal behavior detection network is used to perform temporal analysis on the frame sequence to identify time periods containing key diagnostic operations or key symptom changes. The definition of key events is derived from a system-preset operation reference library and symptom change reference library. For example, key events include "vascular clipping," "lesion resection," and "lymph node dissection" in surgical videos; "biopsy sampling," "hemostasis," and "passage through stenosis" in endoscopic videos; and "abnormal opening and closing of heart valves," "sudden changes in blood flow signals," and "mass movement characteristics" in ultrasound videos. The model identifies key time periods by recognizing specific action patterns, instrument movement trajectories, or dynamic changes in tissue morphology, and segments these periods into key segments from the diagnostic video 15. The action detection model or temporal behavior detection network can be any model provided by publicly available technology in this field.

[0064] For the identified key segments, the most representative keyframes (such as action start, peak, or end frames) are automatically extracted. The keyframes are then input into the connected image generation model (such as BLIP-2, Med-Flamingo, and other medical multimodal large models) to generate a preliminary image description of the content, such as: "The cystic duct is seen to be clamped by titanium clips under laparoscopy, with no bile leakage."

[0065] Then, based on the keyframe image descriptions, and combined with the temporal dynamic information of the video (such as the action process, duration, tissue response, etc.), a semantic segment description of the entire key segment is further generated. The segment description includes static visual information and also reflects the dynamic process. For example: "During the operation, the cystic duct was found to be dilated. The proximal cystic duct was successfully clamped with titanium clips. It was confirmed that there was no bile leakage. The operation was smooth and lasted about 45 seconds."

[0066] Each key segment is bound to its corresponding segment description, and an index is established to associate it with the diagnostic video 15 and the relevant case. Keyframes and their image descriptions are also stored as auxiliary information, forming a multi-level structured knowledge unit.

[0067] Step S4) Generate medical case knowledge item 20 based on the symptom description 11, diagnosis and treatment process description 12, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions.

[0068] The heterogeneous information from historical medical cases 10 is structurally integrated and semantically fused to construct a unified, complete, and computable medical case knowledge item 20. Each medical case knowledge item 20 is constructed as a multimodal knowledge tuple. For example, its structure can be represented as follows:

[0069] {

[0070] Case ID: XXX,

[0071] Symptom description 11: "...",

[0072] Description of the diagnosis and treatment process 12: "...",

[0073] Clinical data: [Indicator 1: Value, Indicator 2: Value, ...],

[0074] Key local image set: [

[0075] { Image ID: IMG_001, Local image description: "..."},

[0076] { Image ID: IMG_002, Local image description: "..."}

[0077] ],

[0078] Key segment set: [

[0079] { Video Clip ID: VID_01, Segment Description: "..."},

[0080] { Video Clip ID: VID_02, Segment Description: "..."} ]

[0082] }

[0083] Through this step, historical medical cases 10 are transformed into standardized knowledge units that are structurally clear, semantically rich, and multimodal. Each medical case knowledge item 20 retains relevant clinical context information, has the ability to be understood, stored, and retrieved by machines, and can be used to build a medical case library.

[0084] Step S5) By adding symptoms and modifying the symptoms description 11, the diagnosis and treatment process description 12 and the diagnosis and treatment data accordingly, as well as adding key local images, local image descriptions, key segments and segment descriptions, an extended medical case knowledge item 20 is generated.

[0085] Building upon the basic medical case knowledge item 20, by intelligently adding reasonable new symptoms and simultaneously expanding the corresponding text descriptions, quantitative data, and multimodal content, extended medical case knowledge items 20 are generated, improving the coverage and clinical representativeness of the case library and making it closer to complex scenarios in the real world.

[0086] Please see the appendix Figure 3 Methods to increase symptoms include:

[0087] Read the pre-configured disease coexistence entry table 21 and disease rejection entry table 23;

[0088] Based on the coexistence entry table 21, add several other diseases 22 that can coexist with disease description 11;

[0089] Obtain a clinical knowledge graph, and infer and add several concurrent diseases based on the clinical knowledge graph;

[0090] Filter according to the disease rejection list 23 to exclude diseases that are rejected.

[0091] Table 21, the list of coexisting diseases, records combinations of diseases that are common or likely to coexist in clinical practice. For example, "type 2 diabetes" often coexists with "hypertension," "hyperlipidemia," and "diabetic nephropathy." Table 23, the list of excluded diseases, records combinations of diseases that are mutually exclusive in terms of pathological mechanism, clinical manifestation, or diagnostic logic. For example, although "acute appendicitis" and "intestinal obstruction" can coexist, "functional abdominal pain" usually excludes organic acute abdomen.

[0092] The system analyzes the primary disease (e.g., "coronary heart disease") in the original case and searches for other diseases (e.g., "heart failure," "arrhythmia," "diabetes") that are comorbid with it in the co-occurrence list 21. These diseases are then added as candidate new diseases to the expansion set. A structured clinical knowledge graph (e.g., UMLS, SNOMED CT) is accessed, and semantic relationships such as "complications," "secondary to," and "often accompanied" are used for inference. For example, "cirrhosis" and "hepatocellular carcinoma" can be inferred from "chronic hepatitis B." Or, "lupus nephritis" and "hematologic system involvement" can be inferred from "systemic lupus erythematosus." This process can discover potential comorbidities not covered by the rule table, enhancing the intelligence and comprehensiveness of the expansion. For all candidate new diseases, the system verifies them in the disease exclusion list 23, excluding items that have medical logical conflicts with the original disease or the selected new diseases.

[0093] The method for generating extended medical case knowledge items 20 by modifying the symptom description 11, the diagnosis and treatment process description 12, and the diagnosis and treatment data accordingly, and by adding key local images, local image descriptions, key segments, and segment descriptions, includes:

[0094] Based on the expanded set of symptoms, update symptom description 11 and treatment process description 12 to cover the clinical symptoms and matching treatment methods of the newly added symptoms;

[0095] Read several typical diagnosis and treatment data corresponding to the newly added disease, and synthesize matching and desensitized diagnosis and treatment data;

[0096] If the newly added symptom involves diagnostic image 14, then the corresponding key local image and local image description will be matched from the medical case database.

[0097] If a new symptom involves a key segment of the diagnosis and treatment video 15, then the corresponding key segment and segment description will be matched from the medical case database.

[0098] The newly added symptoms are integrated into the existing text descriptions to generate a comprehensive description covering multiple symptoms. The diagnosis and treatment process description 12 is updated, supplementing information such as examination items, treatment plans (e.g., the addition of neurological medications), and follow-up plans for the newly added symptoms. Reference ranges are retrieved from a standard database based on typical clinical indicators of the newly added symptoms, serving as desensitized data that conforms to medical principles.

[0099] If a new symptom involves imaging features (such as "diabetic retinopathy"), the system retrieves key local images with corresponding characteristics from the medical case database (such as microaneurysms in fundus photography), borrows and associates existing local image descriptions, and adds them to the current knowledge item. If a new symptom involves specific procedures or dynamic manifestations (such as "renal biopsy"), the system retrieves relevant key segmented videos and their segmented descriptions from the case database, reuses or adapts them, and then incorporates them.

[0100] Step S6) Construct a medical case library based on medical case knowledge item 20 and extended medical case knowledge item 20.

[0101] This paper unifies and integrates 20 medical case knowledge items and 20 extended medical case knowledge items, optimizes their structured storage and indexing, and constructs an intelligent medical case library that is multimodal, semantically rich, and supports efficient retrieval. The original knowledge items from real historical cases and the extended knowledge items generated through intelligent reasoning are merged to form a unified data set. Each knowledge item is described using a standardized data structure to ensure that multimodal information such as text, numerical values, images, and videos is logically consistent and semantically parsable. Furthermore, the medical case library supports continuous incremental updates.

[0102] Step S7) Receive the symptom query description 31 to be queried, and calculate the matching degree between the symptom description 11 of each medical case knowledge item 20 and the symptom query description 31. The specific method includes:

[0103] Input the symptom description 31 of the symptom query description and the symptom description 11 of the medical case knowledge item 20 into the accessed large language model to obtain semantic vectors;

[0104] Calculate the similarity between the semantic vector of the disease query description 31 and the semantic vector of each disease description 11, as the preliminary matching degree;

[0105] Based on keyword matching degree, ICD encoding level matching degree, and symptom weight, a correction coefficient is generated. The initial matching degree is then corrected using the correction coefficient to obtain the corrected matching degree.

[0106] The corrected matching degree is the matching degree between the symptom description 11 of the medical case knowledge item 20 and the symptom query description 31;

[0107] The initial values ​​for keyword matching degree, ICD coding level matching degree, and symptom weight factor are all 1.

[0108] When both symptom description 31 and symptom description 11 match the preset key medical term entries, the keyword matching degree is set to be greater than 1.

[0109] Map symptom query description 31 and symptom description 11 to ICD codes, calculate the semantic distance between the ICD codes of symptom query description 31 and symptom description 11, and when the semantic distance is less than a preset reference value, set the ICD code level matching degree to be greater than 1.

[0110] Construct a symptom weight table, obtain the weight values ​​of the symptoms contained in both symptom query description 31 and symptom description 11 based on the symptom weight table, and obtain the symptom weight based on the maximum value of the weight values ​​of all symptoms.

[0111] The symptom query description 31, such as "recurrent chest pain with shortness of breath for 3 days," and the symptom description 11 from medical case knowledge item 20, such as "acute myocardial infarction, anterior ST-segment elevation, accompanied by left ventricular dysfunction," are input into a large-scale language model. Examples include ClinicalBERT, BioGPT, and Med-PaLM. The large-scale language model maps natural language text into high-dimensional semantic vectors, achieving semantic representation. The cosine similarity between the semantic vector of symptom query description 31 and the semantic vector of each case symptom description 11 is calculated as the initial matching degree. This value reflects the closeness of the two in the semantic space and can identify cases with different expressions but similar meanings, such as "angina pectoris" and "chest pressure." To overcome potential biases in pure semantic matching (such as ignoring key terms, coding levels, or symptom importance), this embodiment uses three correction factors to construct correction coefficients and perform weighted optimization of the initial matching degree.

[0112] Keyword matching score, default value is 1. Maintain a preset table of key medical terms (such as high-risk diagnostic terms such as "myocardial infarction", "cerebral hemorrhage", "shock"). When both the symptom description 31 and the case symptom description 11 contain the same key terms from this table, the keyword matching score is increased (e.g., set to 1.2 or higher) to highlight the matching value of the key diagnostic terms.

[0113] The default value for ICD encoding level matching is 1. The symptom query description 31 and case symptom description 11 are converted to their corresponding ICD-10 codes using a mapping model or terminology standardization tool (such as MetaMap, SNOMED CT to ICD mapping).

[0114] Calculate the semantic distance between two ICD codes (such as path length based on the ICD coding tree, common ancestor level depth, etc.).

[0115] If the semantic distance is less than a preset threshold (e.g., belonging to the category "I20-I25 Angina and Ischemic Heart Disease"), the codes are considered highly correlated, and the ICD coding level matching degree is set to be greater than 1 (e.g., 1.15). This can enhance the accuracy of structured matching by utilizing international standard coding systems.

[0116] A built-in symptom weight table assigns weight values ​​to symptoms based on factors such as clinical specificity, severity, and diagnostic relevance. For example, "loss of consciousness" has a weight of 0.9; "cough" has a weight of 0.3; and "chest pain radiating to the left arm" has a weight of 0.8. The system extracts the set of symptoms common to both the query description 31 and the case description 11, finds their corresponding values ​​in the symptom weight table, and selects the maximum value as the symptom weight factor for this match (range 0-1), highlighting the matching influence of high-value symptoms.

[0117] Multiply the three correction factors mentioned above to obtain the comprehensive correction coefficient. Correction coefficient = Keyword matching degree × ICD coding level matching degree × (1 + Symptom weight). Then calculate the corrected matching degree = Initial matching degree × Correction coefficient.

[0118] Step S8) The medical case knowledge items 20 with the highest matching degree are used as matching results 32. In response to the user's operation, the user is shown the diagnosis and treatment process description 12, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions.

[0119] Based on the corrected matching degree of each medical case knowledge item 20, they are sorted from high to low, and the top 5, top 10, or user-defined number of cases with the highest matching degree are selected as the final matching result set 32.

[0120] On the other hand, additional filtering conditions can also be supported, such as time range (e.g., cases within the last 5 years), case source (e.g., displaying only real cases or including extended cases), and treatment type (e.g., only surgical cases or cases from specific departments), to meet personalized needs in different scenarios.

[0121] In response to user actions (such as clicking "View Details" or "Expand Case"), dynamically display complete multimodal information of the selected matching case to the user.

[0122] The diagnosis and treatment process description 12 can be used to display the complete clinical pathway from initial diagnosis, examination, diagnosis to treatment, helping users understand the decision-making logic and intervention measures of the case. Diagnosis and treatment data are presented in structured tables or charts, showing desensitized key medical indicators such as complete blood count, biochemical indicators, imaging scores, and vital sign trends, facilitating horizontal comparative analysis. Key local images and their descriptions display images of key lesion areas automatically identified and extracted from the original diagnosis and treatment images 14; simultaneously, descriptions of the local images generated by the image-text generation model are displayed, explaining the pathological features in the images (e.g., "A 2.3cm diameter spiculated nodule is seen in the upper lobe of the right lung"). Key segmented videos and their descriptions support the playback of key operational or lesion dynamic segments extracted from the diagnosis and treatment video 15 (e.g., "Clamping of the cystic artery during laparoscopic cholecystectomy"), accompanied by textual descriptions explaining the clinical significance and key operational points of the key segments.

[0123] Users can quickly access and query the most relevant and concise content without having to browse the entire medical record. The fusion and display of multimodal information makes abstract descriptions concrete, which is especially beneficial for understanding imaging and surgical scenarios. The treatment pathways of highly matched cases can provide guidance for current patients, playing an important role, especially in rare diseases or complex comorbidities.

[0124] On the other hand, this manual provides an artificial intelligence-based intelligent medical case matching system; please refer to the appendix. Figure 4 ,include:

[0125] The reading module 100 reads historical medical cases 10, which include disease descriptions 11, treatment process descriptions 12, desensitized treatment data 13, treatment images 14, and treatment videos 15.

[0126] The first recognition module 200 identifies and extracts key local images from the diagnostic image 14, and generates and associates local image descriptions.

[0127] The second recognition module 300 identifies and extracts key segments of the diagnostic video 15, and generates and associates segment descriptions.

[0128] The generation module 400 generates medical case knowledge items 20 based on the symptom description 11, the diagnosis and treatment process description 12, the diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions.

[0129] The extension module 500 generates extended medical case knowledge items 20 by adding symptoms and modifying the symptom description 11, diagnosis and treatment process description 12 and diagnosis and treatment data accordingly, as well as adding key local images, local image descriptions, key segments and segment descriptions accordingly.

[0130] Module 600 is used to build a medical case library based on medical case knowledge item 20 and its extended components.

[0131] The query module 700 receives the symptom query description 31 to be queried, and calculates the matching degree between the symptom description 11 of each medical case knowledge item 20 and the symptom query description 31.

[0132] The matching module 800 takes the medical case knowledge items 20 with the highest matching degree as the matching result 32, and in response to the user's operation, displays the diagnosis and treatment process description 12, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions to the user.

[0133] See also Figure 5 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this specification.

[0134] like Figure 5As shown, the electronic device 1100 may include: at least one processor 1101, at least one network interface 1104, a user interface 1103, a memory 1105, and at least one communication bus 1102. The communication bus 1102 can be used to connect and communicate with the various components mentioned above. The user interface 1103 may include buttons, and optionally may include standard wired or wireless interfaces. The network interface 1104 may include, but is not limited to, a Bluetooth module, an NFC module, or a Wi-Fi module. The processor 1101 may include one or more processing cores. The processor 1101 connects to various parts within the electronic device 1100 using various interfaces and lines, and performs various functions of the routing device and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1105, and by calling data stored in the memory 1105. Optionally, the processor 1101 may be implemented using at least one hardware form of DSP, FPGA, or PLA. The processor 1101 may integrate one or more combinations of CPU, GPU, and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content that the display screen needs to show; and the modem is used for wireless communication.

[0135] It is understandable that the aforementioned modem may not be integrated into the processor 1101, but may be implemented using a separate chip.

[0136] The memory 1105 may include RAM or ROM. Optionally, the memory 1105 may include a non-transitory computer-readable medium. The memory 1105 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 1105 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 1105 may also be at least one storage device located remotely from the aforementioned processor 1101. As a computer storage medium, the memory 1105 may include an operating system, a network communication module, a user interface module, and application programs. The processor 1101 may be used to call the application programs stored in the memory 1105 and execute the methods in the above-described embodiments.

[0137] This specification also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform multiple steps as described in the above embodiments. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0138] This specification also provides a computer program product, including a computer program that, when executed by a processor, implements the multiple steps described in the above embodiments.

[0139] Where there is no conflict, the technical features in this embodiment and implementation scheme can be combined arbitrarily.

[0140] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes multiple computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are 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 through the computer-readable storage medium. The 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 accessible to a computer or a data storage device such as a server or data center integrating multiple available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0141] When implemented through hardware or firmware, the aforementioned method flow is programmed into the hardware circuit to obtain the corresponding hardware circuit structure and achieve the corresponding function. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit, whose logic function is determined by the user programming the device. Designers can program a digital system onto a PLD themselves, eliminating the need for chip manufacturers to design and fabricate dedicated integrated circuit chips. Furthermore, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, similar to the software compiler used in program development. The original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There is not just one HDL, but many. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of the aforementioned hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logic method flow can be easily obtained.

[0142] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Any modifications and improvements made by those skilled in the art to the technical solutions of this specification without departing from the spirit of this specification should fall within the protection scope defined by the claims of this specification.

Claims

1. A medical case intelligent matching method based on artificial intelligence, characterized in that, Including the following steps: Read historical medical cases, which include disease descriptions, treatment process descriptions, desensitized treatment data, treatment images, and treatment videos; Identify and extract key local images from the diagnostic images, and generate and associate local image descriptions; Identify and extract key segments from the medical video, and generate and associate segment descriptions; Based on the description of symptoms, the description of the diagnosis and treatment process, the diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions, generate medical case knowledge items; By adding symptoms and modifying the symptom descriptions, diagnosis and treatment process descriptions, and diagnosis and treatment data accordingly, as well as adding key local images, local image descriptions, key segments, and segment descriptions, extended medical case knowledge items are generated. A medical case library is constructed based on medical case knowledge items and extended medical case knowledge items; Receive the description of the symptom to be queried, and calculate the matching degree between the symptom description of each medical case knowledge item and the symptom query description; The system uses the medical case knowledge items with the highest matching degree as the matching results, and responds to the user's operation to display the description of the diagnosis and treatment process, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions.

2. The intelligent medical case matching method based on artificial intelligence according to claim 1, characterized in that, The method for identifying and extracting key local images from the diagnostic images, and generating and associating local image descriptions includes: A pre-trained image recognition model is used to detect key regions in the diagnostic images and locate key local regions with key pathological features in the images. The key pathological features are obtained from a pre-configured pathological reference library. The detected key local regions are cropped to form key local images; The key local image is input into the connected image and text generation model to generate a local image description.

3. The intelligent medical case matching method based on artificial intelligence according to claim 1 or 2, characterized in that, The method for identifying and extracting key segments from the medical video, and generating and associating segment descriptions includes: The diagnosis and treatment video is subjected to frame sequence extraction. The action detection model is used to identify time periods containing key diagnosis and treatment operations or key symptom changes as key segments. The key diagnosis and treatment operations and key lesion changes are obtained from a pre-configured operation reference library and symptom change reference library. Key frames are extracted from each key segment, and the key frames are input into the connected image and text generation model to generate image descriptions of the key frames; Identify key segments of video content, and generate segment descriptions based on the video content and the image descriptions.

4. The intelligent matching method for medical cases based on artificial intelligence according to claim 1 or 2, characterized in that, Methods to increase symptoms include: Read the pre-configured disease coexistence entry table and disease rejection entry table; Based on the aforementioned list of coexisting diseases, several other diseases that can coexist with the disease description will be added; Obtain a clinical knowledge graph, and infer and add several concurrent diseases based on the clinical knowledge graph; Filter the entries based on the disease rejection list to exclude diseases that are subject to rejection.

5. The intelligent medical case matching method based on artificial intelligence according to claim 4, characterized in that, The methods for generating expanded medical case knowledge items, based on the added symptoms, include modifying the symptom description, treatment process description, and treatment data accordingly, as well as adding key local images, local image descriptions, key segments, and segment descriptions. Based on the expanded set of symptoms, update the symptom descriptions and treatment process descriptions to include the clinical symptoms and matching treatments for the newly added symptoms; Read several typical diagnosis and treatment data corresponding to the newly added disease, and synthesize matching and desensitized diagnosis and treatment data; If a new symptom involves images related to diagnosis and treatment, then the corresponding key local images and descriptions of the local images will be matched from the medical case database. If a new symptom involves a key segment of a medical video, the corresponding key segment and segment description will be matched from the medical case database.

6. A medical case intelligent matching method based on artificial intelligence according to claim 1 or 2, characterized in that, The method for calculating the matching degree between the symptom description of each medical case knowledge item and the symptom query description includes: The symptom descriptions of the symptom query and the symptom descriptions of the medical case knowledge items are respectively input into the large language model to obtain semantic vectors; Calculate the similarity between the semantic vector of the disease query description and the semantic vector of each disease description, as the preliminary matching degree; Based on keyword matching degree, ICD encoding level matching degree, and symptom weight, a correction coefficient is generated. The initial matching degree is then corrected using the correction coefficient to obtain the corrected matching degree. The corrected matching degree is the matching degree between the symptom description of the medical case knowledge item and the symptom query description; The initial values ​​for keyword matching degree, ICD coding level matching degree, and symptom weight factor are all 1. When both the symptom query description and the symptom description match the preset key medical term entries, the keyword matching degree is set to be greater than 1. The symptom query description and the symptom description are mapped to ICD codes. The semantic distance between the symptom query description and the ICD code of the symptom description is calculated. When the semantic distance is less than a preset reference value, the ICD code level matching degree is set to be greater than 1. Construct a symptom weight table, obtain the weight values ​​of symptoms that are present in both the symptom query description and the symptom description based on the symptom weight table, and obtain the symptom weight based on the maximum value of all symptom weight values.

7. A medical case intelligent matching system based on artificial intelligence, characterized in that, include: The reading module reads historical medical cases, which include disease descriptions, treatment process descriptions, desensitized treatment data, treatment images, and treatment videos. The first recognition module identifies and extracts key local images from the diagnostic images, and generates and associates local image descriptions. The second identification module identifies and extracts key segments from the medical video, and generates and associates segment descriptions. The generation module generates medical case knowledge items based on the symptom description, diagnosis and treatment process description, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions; The extension module generates extended medical case knowledge items by adding symptoms, modifying the symptom description, diagnosis and treatment process description and diagnosis and treatment data accordingly, and adding key local images, local image descriptions, key segments and segment descriptions. The module constructs a medical case library based on medical case knowledge items and extended medical case knowledge items. The query module receives the description of the disease to be queried and calculates the matching degree between the description of the disease and the description of the disease query for each medical case knowledge item. The matching module uses the medical case knowledge items with the highest matching degree as the matching results. In response to the user's operation, it displays the description of the diagnosis and treatment process, diagnosis and treatment data, key local images, local image descriptions, key segments and segment descriptions to the user.

8. An electronic device, characterized in that, Including the processor and memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code stored in the memory to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images