CN-121561129-B - Electronic medical record retrieval method and device

CN121561129BCN 121561129 BCN121561129 BCN 121561129BCN-121561129-B

Abstract

The application discloses a method and equipment for retrieving an electronic medical record. The method comprises the steps of analyzing an image description text, determining the focus type described by the image description text, extracting image symptom description related to the focus type from the image description text, predicting the lesion type corresponding to a target object based on the image symptom description, generating a retrieval prompt text based on the lesion type, retrieving a plurality of candidate medical record fragments related to the lesion type based on the retrieval prompt text, and realizing accurate retrieval of medical record fragments based on the image characteristics so as to avoid omission or false detection caused by natural language expression blurring or incomplete retrieval conditions. And a plurality of target medical record fragments with the highest matching degree with the retrieval prompt text are screened out from the plurality of candidate medical record fragments based on the importance values corresponding to the plurality of candidate medical record fragments, so that the efficiency and the accuracy of medical record retrieval are improved.

Inventors

LI JIA
LV HAN
WANG ZHENCHANG
WANG LIHUA
ZHOU ZICHUN
WANG XINGHAO
WANG JIXIANG
LI JUNWEI

Assignees

首都医科大学附属北京友谊医院

Dates

Publication Date: 20260512
Application Date: 20250829

Claims (12)

1. The electronic medical record retrieval method is characterized by comprising the following steps of: Responding to an imaging description text input by a user aiming at a target object in a medical record retrieval interface, and extracting an image symptom description from the imaging description text, wherein the image symptom description is used for describing an image representation corresponding to focus tissues and comprises a plurality of image symptom characteristics; If the table item corresponding to the combination of the image symptom characteristics is retrieved from an image symptom database, determining the lesion type in the table item as the lesion type corresponding to the target object, wherein the image symptom database stores the corresponding relation between various image symptom combinations and the corresponding lesion types; if the table item corresponding to the combination of the plurality of image symptom characteristics is not retrieved from the image symptom database, analyzing the imaging description text by combining a medical knowledge graph and the focus types described in the imaging description text, and predicting the lesion types corresponding to the target object; Identifying a plurality of medical entity words included in the imaging description text, wherein the medical entity words are named entity words with medical information; searching target entity nodes corresponding to the medical entity words from the medical knowledge graph; Determining the centrality of each target entity node in the medical knowledge graph based on the number of directly connected edges of each target entity node in the medical knowledge graph; determining a focus attention field corresponding to the imaging description text based on the degree centrality corresponding to each target entity node; Based on the key focus field, searching relevant factor information corresponding to the key focus field from the medical knowledge graph; Determining a risk factor that causes the lesion type based on the lesion type and the related factor information; filling the lesion type and the risk factors into a preset retrieval prompt template, generating a retrieval prompt text, and retrieving a plurality of candidate medical record fragments matched with the retrieval prompt text from the electronic medical record text corresponding to the target object; Determining an importance value corresponding to each candidate medical record segment, wherein the importance value is used for representing the clinical influence degree of the candidate medical record segment on the image problems described in the imaging description text; screening a plurality of target medical record fragments with importance values meeting preset values from the plurality of candidate medical record fragments based on the importance values corresponding to the plurality of candidate medical record fragments; And carrying out structuring treatment on the plurality of target medical record fragments to obtain a plurality of pieces of structured medical record information, and displaying the plurality of pieces of structured medical record information in the medical record retrieval interface.
2. The method of claim 1, wherein the analyzing the imaging descriptive text in combination with the medical knowledge-graph and the lesion type predicts a lesion type corresponding to the target object, comprising: retrieving knowledge segments related to the lesion type from the medical knowledge graph; Generating a first prompt word according to the knowledge segment, the imaging description text and the focus type; And inputting the first prompt word into a large language model to obtain the lesion type output by the large language model.
3. The method according to claim 1, wherein the searching for relevant factor information corresponding to the focused attention field from the medical knowledge graph based on the focused attention field includes: Generating a query statement based on the focused attention field; and searching relevant factor information corresponding to the key attention field from the medical knowledge graph by using the query statement.
4. The method of claim 1, wherein determining the importance value corresponding to each candidate medical record segment comprises: coding each candidate medical record segment to obtain a coding vector corresponding to each candidate medical record segment; Identifying each medical entity word contained in a target candidate medical record segment aiming at the target candidate medical record segment, wherein the target candidate medical record segment is any one of the plurality of candidate medical record segments; determining weight values corresponding to the medical entity words, wherein the weight values are used for representing the importance degrees corresponding to the medical entity words; determining associated influence values corresponding to the medical entity words, wherein the associated influence values are used for representing the importance degree and associated influence force of the medical entity words in the medical knowledge graph; Determining a degree centrality score corresponding to the target candidate medical record segment based on the weight value corresponding to each medical entity word and the association influence value corresponding to each medical entity word; determining a risk degree score corresponding to the target candidate medical record segment based on the number of preset high-risk medical entity words contained in the target candidate medical record segment; And determining an importance value corresponding to the target candidate medical record segment based on the centrality score corresponding to the target candidate medical record segment and the risk degree score corresponding to the target candidate medical record segment.
5. The method of claim 4, wherein the determining the importance value for the target candidate medical record segment based on the centrality score for the target candidate medical record segment and the risk level score for the target candidate medical record segment comprises: Respectively determining the association degree between each entity word and the image symptom description, wherein the association degree is used for representing the association degree between the entity information corresponding to the entity word and the image symptom description; Determining a sign association score corresponding to the target candidate medical record segment based on the weight value corresponding to each medical entity word and the association between each entity word and the image sign description; And determining an importance value corresponding to the target candidate medical record segment based on the centrality score of the degree corresponding to the target candidate medical record segment, the risk degree score corresponding to the target candidate medical record segment and the sign association score corresponding to the target candidate medical record segment.
6. The method of claim 1, wherein retrieving a plurality of candidate medical record segments from the electronic medical record text corresponding to the target object that match the retrieval hint text comprises: retrieving an electronic medical record text corresponding to the target object from a medical record database, and segmenting the electronic medical record text to obtain a plurality of medical record segments; Inputting the plurality of medical record fragments and the retrieval prompt text into a pre-trained text retrieval model to obtain a plurality of candidate medical record fragments matched with the retrieval prompt text; In the text retrieval model, semantic coding is carried out on the retrieval prompt text to obtain first semantic feature vectors corresponding to the retrieval prompt text, semantic coding is carried out on the medical record fragments to obtain second semantic feature vectors corresponding to the medical record fragments, similarity between the second semantic feature vectors corresponding to each medical record fragment and the first semantic feature vectors corresponding to the retrieval prompt text is calculated, and a plurality of candidate medical record fragments are screened out from the medical record fragments according to a preset similarity threshold.
7. The method of claim 1, wherein prior to extracting an image symptom description from the imaging descriptive text, the method further comprises: rewriting the imaging description text to obtain a structured imaging description text; Wherein the extracting the image symptom description from the image description text comprises the following steps: generating a second prompt word based on the structured imaging description text; Inputting the second prompt word into a large language model, so that the large language model determines the focus type corresponding to focus tissues described in the structured imaging description text by analyzing the structured imaging description text; And extracting an image symptom description related to the focus type from the structured imaging description text.
8. The method of claim 7, wherein the rewriting the imaging descriptive text to obtain a structured imaging descriptive text comprises: inputting the imaging description text into a pre-trained text rewrite model to obtain standardized entity words corresponding to each entity word in the imaging description text; Generating a structured imaging description text based on the rewritten text corresponding to each entity word; In the text rewrite model, each entity word in the imaging description text is encoded to obtain word vectors corresponding to each entity word in the imaging description text, the word vectors corresponding to each entity word, the medical entity embedding vectors corresponding to each entity word and the relation embedding vectors among medical entities corresponding to each entity word are fused to obtain fused vector representations corresponding to each entity word, and decoding is carried out on the fused vector representations corresponding to each entity word to obtain rewrite text corresponding to each entity word; The medical entity embedding vector is used for representing semantic information of each entity word in the medical knowledge graph, and the relation embedding vector is used for representing association relations among medical entities corresponding to each entity word.
9. The method of claim 4, wherein the identifying individual medical entity terms contained in the target candidate medical record segment comprises: And inputting the target candidate medical record fragments into a pre-trained medical entity recognition model to obtain a plurality of medical entity words, wherein the medical entity recognition model is used for recognizing the medical entity words included in the description text.
10. The method of claim 1, wherein the screening a plurality of target medical record segments from the plurality of candidate medical record segments for which the importance value meets a preset value based on the importance values corresponding to the plurality of candidate medical record segments, comprises: And inputting the importance values corresponding to the candidate medical record fragments, the candidate medical record fragments and the retrieval prompt text into a pre-trained fragment sequencing model to obtain a plurality of target medical record fragments corresponding to the retrieval prompt text with the highest matching degree.
11. The method of claim 1, wherein the displaying the plurality of structured medical record information in the medical record retrieval interface comprises: sorting the plurality of structured medical record information according to a preset imaging evaluation priority rule to obtain a plurality of sorted structured medical record information; and sequentially displaying the sequenced plurality of structured medical record information in the medical record retrieval interface.
12. An electronic device comprising a memory and a processor, the memory for storing a computer program, the processor coupled to the memory for executing the computer program to implement the steps in the electronic medical record retrieval method of any one of claims 1-11.

Description

Electronic medical record retrieval method and device Technical Field The application belongs to the technical field of computers, and particularly relates to a method and equipment for retrieving electronic medical records. Background In the medical imaging field, imaging doctors need to convert complex visual information such as CT, MRI, X-ray and the like into image description text, and generate reliable image reports in combination with patient clinical medical record information. The imaging description text generally contains a large amount of terms and specific image symptom information, and how to quickly extract key information from the terms and perform targeted electronic medical record retrieval is an important requirement for doctors and researchers. However, conventional electronic medical record retrieval relies on keyword matching, and language structures specific to imaging and specific image sign information in imaging descriptions are generally not understood, and insufficient accuracy and relevance of retrieval results cannot be identified. Disclosure of Invention In view of the above, the present application provides a method, apparatus and storage medium for retrieving electronic medical records, which solve or partially solve the above technical problems. In a first aspect, an embodiment of the present application provides a method for retrieving an electronic medical record, where the method includes: Responding to an imaging description text input by a user aiming at a target object in a medical record retrieval interface, and extracting an image symptom description from the imaging description text, wherein the image symptom description is used for describing an image representation corresponding to focus tissues; Analyzing the image symptom description and predicting the lesion type corresponding to the target object; Generating a retrieval prompt text based on the lesion type, and retrieving a plurality of candidate medical record fragments matched with the retrieval prompt text from the electronic medical record text corresponding to the target object; Determining an importance value corresponding to each candidate medical record segment, wherein the importance value is used for representing the clinical influence degree of the candidate medical record segment on the image problems described in the imaging description text; screening a plurality of target medical record fragments with importance values meeting preset values from the plurality of candidate medical record fragments based on the importance values corresponding to the plurality of candidate medical record fragments; And carrying out structuring treatment on the plurality of target medical record fragments to obtain a plurality of pieces of structured medical record information, and displaying the plurality of pieces of structured medical record information in the medical record retrieval interface. In a second aspect, an embodiment of the present application provides an electronic medical record retrieving apparatus, including: the extraction module is used for responding to an imaging description text input by a user aiming at a target object in a medical record retrieval interface, extracting an image symptom description from the imaging description text, wherein the image symptom description is used for describing an image representation corresponding to focus tissues; The prediction module is used for analyzing the image symptom description and predicting the lesion type corresponding to the target object; The generation module is used for generating a retrieval prompt text based on the lesion type and retrieving a plurality of candidate medical record fragments matched with the retrieval prompt text from the electronic medical record text corresponding to the target object; The determining module is used for determining an importance value corresponding to each candidate medical record segment, wherein the importance value is used for representing the clinical influence degree of the candidate medical record segment on the image problems described in the imaging description text; the screening module is used for screening a plurality of target medical record fragments with importance values meeting preset values from the plurality of candidate medical record fragments based on the importance values corresponding to the plurality of candidate medical record fragments; The processing module is used for carrying out structuring processing on the target medical record fragments to obtain a plurality of pieces of structured medical record information, and displaying the structured medical record information in the medical record retrieval interface. In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a communication interface, where the memory stores executable code, and when the executable code is executed by the processor, the processor is