Treatment scheme extraction and retrieval method and device, computer equipment and storage medium

Abstract

The invention provides a therapeutic scheme extraction and retrieval method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring medical literatures of a to-be-extracted therapeutic scheme; inputting the medical literature into a trained text classification model, and outputting a target statement describing a therapeutic schedule in the medical literature; and aiming at the target statement, respectively extracting a first treatment scheme independently existing in the target statement and / or a second treatment scheme associated with the target statement to obtain a treatment scheme in the medical literature. By adopting the method and the device, the extraction efficiency of treatment schemes for various medical literatures can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a treatment plan extraction and retrieval method, device, computer equipment and storage medium. Background Art

[0002] With the advancement of science and technology, the improvement of human health awareness and the increase in global medical needs, the pharmaceutical field is facing unprecedented development opportunities.

[0003] For medical professionals and researchers, it is very important to read and deeply analyze the treatment plans involved in various literature, because this in-depth analysis can help them understand the current treatment methods and drug efficacy, guide clinical practice and decision-making, and provide basic support for medical diagnosis and treatment. However, in-depth analysis often consumes a lot of manpower to interpret each article, and can only obtain a general treatment plan and some simple details, and the information reading efficiency is not high.

[0004] Therefore, the existing treatment plan extraction method has the technical problem of low extraction efficiency. Summary of the invention

[0005] The purpose of the present invention is to provide a treatment plan extraction and retrieval method, device, computer equipment and storage medium, which are used to reduce the treatment plan retrieval response time by improving the treatment plan extraction efficiency, while improving the treatment plan extraction accuracy and robustness.

[0006] In a first aspect, the present invention provides a method for extracting a treatment regimen, comprising: Obtain medical literature on treatment options to be extracted; Input medical literature into the trained text classification model and output the target sentences in the medical literature that describe treatment plans; For the target sentence, the first treatment plan that exists alone and / or the second treatment plan that exists in association therein are extracted respectively to obtain the treatment plans in the medical literature.

[0007] In some embodiments of the present invention, for a target sentence, a first treatment plan that exists alone and / or a second treatment plan that exists in association therein are extracted respectively to obtain treatment plans in medical literature, including: inputting the target sentence into a trained pointer model so that the pointer model extracts plans for the target sentence based on preset prompt information, and outputs a treatment plan corresponding to the prompt information; parsing the treatment plan corresponding to the prompt information to obtain a first treatment plan and / or a second treatment plan; and counting and cleaning the first treatment plan and / or the second treatment plan to obtain treatment plans in medical literature.

[0008] In some embodiments of the present invention, prompt information includes first prompt information and second prompt information, the first prompt information is used to prompt the output of a separate treatment plan, and the second prompt information is used to prompt the output of an associated treatment plan; wherein, parsing the treatment plan corresponding to the prompt information to obtain the first treatment plan and / or the second treatment plan includes: determining the treatment plan corresponding to the first prompt information as the first treatment plan; determining the treatment plan corresponding to the second prompt information as the second treatment plan.

[0009] In some embodiments of the present invention, before inputting medical literature into a trained text classification model, it also includes: constructing an initial text classification model; obtaining a sample data set, the sample data set including a plurality of positive sample sentences and a plurality of negative sample sentences, the positive sample sentences and the negative sample sentences being determined based on an analysis of clinical trial registration information; and using the sample data set to train the initial text classification model to obtain a trained text classification model.

[0010] In some embodiments of the present invention, obtaining a sample data set includes: obtaining sample medical literature containing a clinical trial registration number, and clinical trial registration information associated with the clinical trial registration number; encoding each sentence in the sample medical literature to obtain a first feature vector; encoding the clinical trial registration information to obtain a second feature vector; and respectively calculating the similarity between each first feature vector and the second feature vector, so as to use sentences with corresponding similarities greater than or equal to a preset similarity threshold as positive sample sentences, and use the remaining sentences except the positive sample sentences as negative sample sentences.

[0011] In a second aspect, the present invention provides a treatment plan retrieval method, comprising: Get the target search term; Determine a target treatment plan that matches the target search term through a treatment plan library, wherein the treatment plan library is constructed based on treatment plans determined by any of the above treatment plan extraction methods, and the target treatment plan includes a first treatment plan and / or a second treatment plan; Presentation of targeted treatment options.

[0012] In a third aspect, the present invention provides a treatment plan extraction device, comprising: The literature acquisition module is used to obtain medical literature on the treatment plans to be extracted; The literature analysis module is used to input medical literature into the trained text classification model and output the target sentences in the medical literature that describe the treatment plan; The scheme extraction module is used to extract the first treatment scheme that exists alone and / or the second treatment scheme that exists in association with the target sentence, and obtain the treatment scheme in the medical literature.

[0013] In a fourth aspect, the present invention provides a treatment plan retrieval device, comprising: A search term acquisition module, used to acquire target search terms; A treatment plan retrieval module is used to determine a target treatment plan that matches a target search term through a treatment plan library, wherein the treatment plan library is constructed based on treatment plans determined by any of the above treatment plan extraction methods, and the target treatment plan includes a first treatment plan and / or a second treatment plan; The plan display module is used to display the target treatment plan.

[0014] In a fifth aspect, the present invention further provides a computer device, comprising: one or more processors; A memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the above-mentioned treatment plan extraction method, or to implement the above-mentioned treatment plan retrieval method.

[0015] In a sixth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to execute the steps in the treatment plan extraction method or the steps in the treatment plan retrieval method.

[0016] In a seventh aspect, an embodiment of the present invention provides a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in the first aspect or the second aspect.

[0017] The above-mentioned treatment plan extraction and retrieval method, device, computer equipment and storage medium, the server obtains the medical literature of the treatment plan to be extracted, and inputs the medical literature into the trained text classification model, so as to obtain the target sentence describing the treatment plan in the medical literature, and then extracts the first treatment plan that exists alone and / or the second treatment plan that exists in association with the target sentence, and finally integrates them to obtain the treatment plan in the medical literature. This solution avoids high-cost manpower consumption both in terms of literature intensive reading and data annotation, and effectively improves the efficiency of information extraction for treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A schematic diagram of a scenario of a treatment plan extraction method in an embodiment of the present invention; Figure 2 is a flow chart of a treatment plan extraction method in an embodiment of the present invention; Figure 3 is a flowchart of a treatment plan retrieval method in an embodiment of the present invention; Figure 4 is a schematic diagram of the structure of a treatment plan extraction device in an embodiment of the present invention; Figure 5 is a schematic diagram of the structure of a treatment plan retrieval device in an embodiment of the present invention; Figure 6 It is a schematic diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0022] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known model structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0023] In the description of the present invention, the treatment plan extraction method and treatment plan retrieval method provided by the embodiment of the present invention mainly involve natural language processing (NLP) in artificial intelligence (AI). Among them, artificial intelligence (AI) is the theory, method, technology and application system of using digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can realize effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science and mathematics. Therefore, research in this field will involve natural language, that is, the language used by people in daily life, so it is closely related to the study of linguistics, but there are important differences. Natural language processing is not a general study of natural language, but the development of computer systems that can effectively realize natural language communication, especially the software system therein, so it is a part of computer science.

[0024] In the description of the present invention, a treatment plan refers to a series of medical measures and drug use plans formulated to treat a disease or condition. Its role is to promote the patient's rehabilitation and health recovery by providing effective treatment methods for specific diseases. A good treatment plan is an indispensable and important part of modern medical practice. It provides patients with the best treatment effect, improves the recovery rate, and improves the quality of life through systematic treatment strategies and personalized treatment plans. Therefore, mining and collating treatment plans recorded in various medical literature (such as scientific research papers, clinical trial reports, professional magazines, and academic conference records, etc.) is an indispensable step in the medical process. It not only helps doctors and researchers to develop more effective treatment plans and optimize existing treatment methods, but can even guide relevant personnel to discover new therapies.

[0025] However, currently, the information related to treatment plans in medical literature is generally obtained manually. Even if a small number of them use artificial intelligence to achieve automatic information extraction, it is still very expensive in terms of manpower and material resources. For example, in order to achieve automatic information extraction, data annotation needs to be completed in advance. However, it is very expensive to annotate the main treatment plans mentioned in the article from the title and abstract of a document. There are several reasons for this: 1. The descriptions of the plans involved in each article may be different. Even if it is the same plan, the representations in different articles may be different. Over-reliance on manual experience leads to low annotation efficiency; 2. The plans involved in the article may be mentioned everywhere in the article, and the semantic context is different; 3. There may be multiple plans in the article, but some plans are not the main plans that the article wants to show, but are just mentioned; 4. Some plans are joint plans, that is, the position of the complete plan in the article is jumpy; 5. Different annotators may have different strengths in defining the plan, which may lead to conflicts in data annotation rules; 6. Manual annotation sometimes requires understanding the entire text to figure out what the main plan described in the article is, which is very time-consuming.

[0026] In summary, although the neural network combined with deep learning methods can pre-learn this type of document text, and after learning is completed, a large amount of text can be input for automatic text recognition and information extraction, it is limited by the current deep learning algorithm design and computing power bottlenecks. Only a small number of algorithm technologies have been successfully implemented, and the adapted scenarios are very single. When encountering complex situations, such as multi-event extraction, entity overlap, etc., the existing information extraction algorithms cannot meet the requirements of accuracy and robustness for the extraction effect.

[0027] In this regard, this solution provides a new idea. By training a text binary classification model, the structured extraction of treatment plans in medical literature is realized. This structured extraction method is not only for single treatment plans, but also for multiple combined treatment plans, that is, to clarify their associations, thereby satisfying the user's efficient and accurate retrieval of single treatment plans or related treatment plans. The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] The embodiments of the present invention provide a treatment plan extraction and retrieval method, apparatus, computer equipment and storage medium, which are described in detail below.

[0029] See also Figure 1 , Figure 1 The schematic diagram of the scenario of the treatment plan extraction and retrieval method provided by the present invention can be applied to the treatment plan retrieval system. The treatment plan retrieval system includes a terminal 102 and a server 104. The terminal 102 can be a device including both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing two-way communication on a two-way communication link. Such a device may include: a cellular or other communication device having a single-line display or a multi-line display or a cellular or other communication device without a multi-line display.

[0030] Specifically, the terminal 102 can be a desktop terminal or a mobile terminal, and the terminal 102 can also be a mobile phone, a tablet computer, or a laptop computer. The server 104 can be an independent server, or a server network or a server cluster composed of servers, which includes but is not limited to a computer, a network host, a single network server, an edge server, a plurality of network server sets, or a cloud server composed of a plurality of servers. Among them, the cloud server is composed of a large number of computers or network servers based on cloud computing. In addition, a communication connection is established between the terminal 102 and the server 104 through a network, and the network can be any one of a wide area network, a local area network, and a metropolitan area network.

[0031] Those skilled in the art will understand that Figure 1 The application environment shown in the figure is only one application scenario applicable to the solution of the present invention and does not constitute a limitation on the application scenario of the solution of the present invention. Other application environments may also include Figure 1 More or less computer equipment as shown in Figure 1 Only one server 104 is shown in FIG. 1 . It is understandable that the treatment plan retrieval system may also include one or more other servers, which are not specifically limited here. Figure 1As shown, the treatment plan retrieval system may also include a memory for storing data, such as various medical documents of different formats, sources, and publication times, and / or treatment plans extracted from medical documents using the method described in the present invention.

[0032] It should be noted that Figure 1 The scenario diagram of the treatment plan retrieval system shown is only an example. The treatment plan retrieval system and scenario described in the embodiment of the present invention are intended to more clearly illustrate the technical solution of the embodiment of the present invention, and do not constitute a limitation on the technical solution provided by the embodiment of the present invention. Ordinary technicians in this field can know that with the evolution of the treatment plan retrieval system and the emergence of new business scenarios, the technical solution provided by the embodiment of the present invention is also applicable to similar technical problems.

[0033] See also Figure 2 The present invention provides a method for extracting a treatment plan. This embodiment mainly applies the method to the above Figure 1 Taking the server 104 in the example, the method includes steps S201 to S203, which are as follows: S201, obtaining medical literature of treatment plans to be extracted.

[0034] Among them, medical literature includes but is not limited to clinical guideline articles, review articles, original research papers, randomized controlled trial (RCT) research articles, some medical and pharmaceutical magazines or journals, academic conference abstracts and speeches, etc. It is understandable that some of the above articles or materials can be collected and stored in compliance with medical databases (such as PubMed, Medline, Embase, etc.).

[0035] In the specific implementation, this embodiment and subsequent embodiments will be explained by taking the medical literature included in the database PubMed as an example. Specifically, the server 104 can obtain medical literature in the following ways: 1. If in a common network structure, the server 104 can obtain medical literature from the terminal 102 or other cloud devices with a network connection; 2. If in a preset blockchain network, the server 104 can synchronously obtain medical literature from other terminal nodes or server nodes, and the blockchain network can be a public chain, a private chain, etc. In addition, since the text classification model will be used to determine whether there is a treatment plan in each medical document in the future, the currently obtained medical literature will not be screened for the time being, but if analysis and extraction of a specific type of treatment plan is required, further screening can be performed in this step.

[0036] For example, if a treatment plan related to lung cancer is needed, the server 104 may filter the collected medical literature by keywords after collecting the medical literature. The keywords may be preset to be associated with lung cancer, so that the filtered medical literature is the medical literature related to lung cancer.

[0037] S202, inputting medical literature into the trained text classification model, and outputting target sentences in the medical literature that describe treatment plans.

[0038] Among them, the text classification model selected in the embodiment of the present invention includes but is not limited to any of the following: Support Vector Machines (SVM), Logistic Regression, deep learning models: such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), etc., which learn text features and semantic information through multi-layer neural networks; pre-trained models: such as BERT (Bidirectional Encoder Representations from Transformers), which is pre-trained on a large-scale data set and then fine-tuned on a specific task, which can effectively perform text classification.

[0039] In a specific implementation, after the server 104 obtains the medical literature that meets the actual application requirements, it can call the trained text classification model to perform machine analysis on the currently obtained medical literature, so that the model can classify and annotate the medical literature with multiple labels based on the functions learned in the previous training, and then output the sentence describing the treatment plan in the literature as the target sentence. However, since the output target sentence may contain multiple treatment plans, the multiple treatment plans may have the same parts (generally referring to time, number of people, drug usage and dosage, etc.), or different parts (generally drug plans, usage and dosage, etc.), it is impossible to clarify whether the multiple treatment plans can be used in combination or only separately. Therefore, in order to further clarify the relationship between the multiple treatment plans and to accurately extract each treatment plan, it is necessary to further analyze the target sentence, and the specific analysis steps will be described in detail below.

[0040] S203, for the target sentence, extract the first treatment plan that exists alone and / or the second treatment plan that exists in association therein, to obtain the treatment plans in the medical literature.

[0041] Among them, the first treatment plan refers to the treatment plan mentioned separately in the target sentence; the second treatment plan refers to the treatment plan that can be used in combination with other treatment plans mentioned in the target sentence. For example, if a target sentence describes "treatment plan A + treatment plan C, treatment plan B + treatment plan C" or "treatment plan A or B, plus treatment plan C", then the first treatment plan includes "A" and "B", and the second treatment plan includes "C". For another example, if a target sentence describes "treatment plan A, treatment plan B + treatment plan C, treatment plan D + treatment plan C", then the first treatment plan includes "A", "B" and "D", and the second treatment plan includes "C".

[0042] In a specific implementation, in order to extract all treatment plans in medical literature, the embodiment of the present invention proposes that artificial intelligence technology can be used to analyze the current target sentence, that is, the server 104 can input the target sentence into a certain type of neural network model for machine analysis, so that the model can extract the first treatment plan that exists alone and / or the second treatment plan that exists in association under a supervised task setting, and finally count and clean all treatment plans in the medical literature. Preferably, the neural network model described above can use a prompt-based pointer network, and the specific analysis steps will be described in detail below.

[0043] In one embodiment, step S203 includes: inputting the target sentence into a trained pointer model so that the pointer model extracts a treatment plan for the target sentence based on preset prompt information, and outputs a treatment plan corresponding to the prompt information; parsing the treatment plan corresponding to the prompt information to obtain a first treatment plan and / or a second treatment plan; counting and cleaning the first treatment plan and / or the second treatment plan to obtain a treatment plan in medical literature.

[0044] Among them, the preset prompt information can be prompt information related to the treatment plan preset by the user, for example, the prompt information can be "What is the treatment plan mentioned separately in the sentence"; the preset prompt information can also be prompt information generated after the pointer model is learned, for example, "What is the common treatment plan combined with treatment plan A" and so on.

[0045] In a specific implementation, the trained pointer model can be called to analyze and extract the treatment plan in the medical literature, and the model training needs to be completed before calling the trained pointer model. At this time, since the cost of manual annotation is high but the amount of original data is also large, the embodiment of the present invention proposes that when constructing the data set for training the pointer model, a small part of it can be data annotated by experts, and most of the training data is constructed by the algorithm.

[0046] Specifically, in order to make full use of the advantages of a large amount of original data, we can combine the original unlabeled data and some expert-level labeled data to train a generative model based on a certain type of network architecture. After the model is trained, it can generate some pseudo-labeled data. Then, we combine a small amount of expert-level labeled data with a large amount of generated pseudo-labeled data for supervised model learning. The model structure uses a prompt-based pointer network, which inputs prompt information + original sentences, outputs the results of the specified task, and repeatedly iterates and updates to obtain a trained pointer model.

[0047] It should be noted that the prompt-based pointer network is a model architecture for generative text tasks, which is mainly used to handle text generation tasks, such as generating answers to questions, article summaries, etc. In the prompt-based pointer network, the input consists of two parts: one is the question or prompt (prompt), and the other is the target text to be generated. The goal of this model is to generate text content related to the given question or prompt by using it. The model usually uses an encoder-decoder architecture. The encoder encodes the input question or prompt and converts it into a fixed-length vector representing semantic information. The decoder then receives this vector and uses it as context information to start generating the target text. The decoder can use methods such as recurrent neural networks (RNNs) or attention mechanisms for text generation. The key innovation is to use the question or prompt as part of the input. By providing additional question or prompt information, the model can better understand and capture the context and semantics of the target text. This approach helps to alleviate ambiguity and confusion in the generation model and can more accurately output text results that meet expectations. The prompt-based pointer network has great potential in natural language processing tasks, especially in generative text tasks. It can generate various related texts according to different questions or prompts, improving the accuracy and coherence of automatically generated texts.

[0048] Furthermore, after the pointer model has the calling conditions, the server 104 can input the target sentence into the trained pointer model, so that the model can extract the scheme for the target sentence based on the prompt information, and then output the treatment scheme corresponding to the prompt information. At this time, the server 104 will obtain the first treatment scheme and the second treatment scheme output by the pointer model, integrate all the schemes and clean up the redundant and repetitive schemes, and then obtain all the treatment schemes in the medical literature.

[0049] In one embodiment, the prompt information includes a first prompt information and a second prompt information, the first prompt information is used to prompt the output of a separate treatment plan, and the second prompt information is used to prompt the output of an associated treatment plan; wherein, parsing the treatment plan corresponding to the prompt information to obtain the first treatment plan and / or the second treatment plan includes: determining the treatment plan corresponding to the first prompt information as the first treatment plan; determining the treatment plan corresponding to the second prompt information as the second treatment plan.

[0050] The first prompt information is used to prompt the output of a separate treatment plan, for example, "What is the treatment plan mentioned separately in the sentence?" The second prompt information is used to prompt the output of associated treatment plans, for example, "What is the common treatment plan combined with treatment plan B?" It should be noted that the second prompt information is usually determined based on the treatment plan corresponding to the first prompt information.

[0051] In a specific implementation, the server 104 may use the treatment plan corresponding to the first prompt information output as the first treatment plan, and use the treatment plan corresponding to the second prompt information output as the second treatment plan. For example, a sentence contains two treatment plans: A+C and B+C. After the sentence is input into the trained pointer model, if the prompt information "prompt" is: "What are the treatment plans mentioned separately in the sentence", then the output results of the pointer model are "A" and "B"; if the prompt information "prompt" is: "What are the common treatment plans combined with treatment plan B", then the output result of the pointer model is "C"; if the prompt information "prompt" is: "What are the common treatment plans combined with treatment plan A", then the output result of the pointer model is "C".

[0052] Therefore, the treatment plan extraction method proposed in the embodiment of the present invention can actually be applied to processing scenarios where there is only one entity in the text, but there are several entity texts, and these entity texts contain or cross-contain each other, and their respective associations cannot be clearly defined. It also solves the problem that an entity may be distributed in different parts of a sentence. For example, the content of a sentence is ABCDE, where A+D is an entity text and C+D is an entity text. The above-mentioned present invention can be used to clarify the text composition in a single entity.

[0053] In one embodiment, before step S202, the method further includes: constructing an initial text classification model; obtaining a sample data set, the sample data set including a plurality of positive sample sentences and a plurality of negative sample sentences, the positive sample sentences and the negative sample sentences being determined based on the analysis of clinical trial registration information; and using the sample data set to train the initial text classification model to obtain a trained text classification model.

[0054] Among them, positive sample sentences refer to sentences that meet certain specific conditions or styles, which are used to train machine learning models or verify the effectiveness of algorithms; negative sample sentences, correspondingly, refer to sentences that do not meet certain conditions or styles.

[0055] Among them, clinical trial registration information usually includes the following: clinical trial registration number (NCT number for short), trial design (such as the objectives of the trial, research questions, trial design type such as randomized controlled trials or non-randomized controlled trials), participant recruitment conditions, intervention measures, primary and secondary research results, expected effect indicators, trial progress and completion status.

[0056] In a specific implementation, the server 104 needs to train an initial text classification model so that it can be called in subsequent steps, and the sample data set used to train the text classification model can be constructed based on the analysis of clinical trial registration information, but not only based on the analysis of clinical trial registration information, the clinical trial registration information can be obtained in a certain compliant manner. That is to say, after the server 104 obtains the clinical trial registration information in compliance with regulations, the clinical trial registration information can be analyzed according to certain preset rules to obtain the positive sample sentences and negative sample sentences required for training, and then the initial text classification model is trained with the positive sample sentences and negative sample sentences to obtain a text classification model capable of outputting sentences containing treatment plans.

[0057] In one embodiment, obtaining a sample data set includes: obtaining sample medical literature containing a clinical trial registration number, and clinical trial registration information associated with the clinical trial registration number; encoding each sentence in the sample medical literature to obtain a first feature vector; encoding the clinical trial registration information to obtain a second feature vector; and respectively calculating the similarity between each first feature vector and the second feature vector, so as to use sentences with corresponding similarities greater than or equal to a preset similarity threshold as positive sample sentences, and use the remaining sentences except the positive sample sentences as negative sample sentences.

[0058] In a specific implementation, the sample medical literature containing the clinical trial registration number refers to the literature with the NCT number. Taking the PubMed article as an example, for the PubMed article with the NCT number, the server 104 can first use the NCT number to find the matching scheme information in the corresponding clinical trial registration information, and then integrate the scheme information, and use the existing embedding model to encode the scheme information and each sentence in the PubMed article into a 768-dimensional vector, and obtain the first eigenvector and the second eigenvector respectively.

[0059] Further, the similarity between the first feature vector and the second feature vector of all sentences is calculated, and the sentence with the largest similarity value is taken to determine whether it is greater than or equal to a preset similarity threshold. If it exceeds the threshold, the corresponding sentence should be the sentence representing the main treatment plan in the article, which can be used as a positive sample sentence, and the remaining sentences with similarity values ​​less than the similarity threshold are used as negative sample sentences. It can be understood that the similarity algorithm used in this embodiment can be any of the following: cosine similarity, Jaccard Similarity, TF-IDF (TermFrequency-Inverse Document Frequency).

[0060] Of course, those skilled in the art should know that the trained text classification model can be used not only for medical literature with NCT numbers, but also for medical literature without NCT numbers.

[0061] It should be noted that the positive sample sentences mentioned in this embodiment may be sentences indicating the primary treatment plan, rather than sentences indicating the secondary treatment plan. The sentences indicating the secondary treatment plan are sentences that simply mention a treatment plan as an introduction, background or reference and other secondary information. In order to make the extracted treatment plan cover more complete information, the embodiment of the present invention proposes that the information completeness of the obtained treatment plan can be controlled by adjusting the similarity threshold, that is, the higher the similarity threshold is set, the sentences with the corresponding similarity value greater than or equal to the threshold are closer to the sentences indicating the primary treatment plan, thereby filtering out the sentences indicating the secondary treatment plan, and the retained treatment plan is more valuable for reference.

[0062] In the treatment plan extraction method in the above embodiment, the server obtains the medical literature of the treatment plan to be extracted and inputs the medical literature into the trained text classification model to obtain the target sentence describing the treatment plan in the medical literature, and then extracts the first treatment plan that exists alone and / or the second treatment plan that exists in association with the target sentence, and finally integrates to obtain the treatment plan in the medical literature. Since this solution does not require a lot of manpower and is suitable for structured and precise extraction of multiple overlapping entities, it can not only effectively improve the efficiency of information extraction for treatment plans, but also improve the accuracy of treatment plan extraction, providing high-quality data support for subsequent treatment plan retrieval. See also Figure 3 The present invention provides a treatment plan retrieval method. This embodiment mainly applies this method to the above Figure 1 Taking the terminal 102 in the example, the method includes steps S301 to S303, which are as follows: S301, obtaining a target search term; S302, determining a target treatment plan that matches the target search term through a treatment plan library, where the treatment plan library is constructed based on treatment plans determined by any of the above treatment plan extraction methods, and the target treatment plan includes a first treatment plan and / or a second treatment plan; S303, display the target treatment plan.

[0063] Among them, the target search terms can be medical field vocabulary related to the treatment plan, including but not limited to drug names (such as "patuzumab", "recombinant human hyaluronidase", etc.), target names (such as "PD1", "5T4", etc.), drug types (such as "antibody-drug conjugates", "mRNA vaccines", etc.), and drug categories (such as "compound", "peptide", "antibody", etc.).

[0064] In a specific implementation, according to the treatment plan extraction method in the above embodiment, after obtaining the treatment plans contained in the medical literature, a highly reliable treatment plan library can be constructed to facilitate users to quickly search and find the target treatment plan. It is understandable that when constructing a treatment plan, the combination and combined therapy can be associated and stored in the treatment plan library, so as to facilitate the rapid location of the combination and combined therapy plans of the target treatment plan during search and query.

[0065] In addition, the user can input the target search term through the terminal 102, and send the target search term to the server 104 for retrieval. Here, the target search term can be the complete content of the target treatment plan, or it can be a partial content of the target treatment plan, and the embodiment of the present invention does not specifically limit this. The user can input the target search term by directly inputting the target search term, or by uploading a file containing the target search term, and the embodiment of the present invention does not specifically limit this.

[0066] In the treatment plan retrieval method in the above embodiment, after the server builds a high-reliability treatment plan library based on the treatment plans extracted from medical literature, it can provide users with a quick search and query function for treatment plans through the terminal. That is, when users need to understand the current treatment methods and drug efficacy to guide clinical practice and decision-making, they can quickly search and query known treatment plans and their effects based on the treatment plan library, thereby meeting user search needs. It should be understood that although Figure 2 or Figure 3 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 or Figure 3At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps. In order to better implement the treatment plan extraction method provided in the embodiment of the present invention, on the basis of the treatment plan extraction method proposed in the embodiment of the present invention, the embodiment of the present invention further provides a treatment plan extraction device, which can be deployed in Figure 1 In the server 104 shown, Figure 4 As shown, the treatment plan extraction device 400 includes: The literature acquisition module 410 is used to acquire medical literature of the treatment plan to be extracted; The document analysis module 420 is used to input the medical documents into the trained text classification model and output the target sentences describing the treatment plan in the medical documents; The solution extraction module 430 is used to extract the first treatment solution that exists alone and / or the second treatment solution that exists in association with the target sentence, and obtain the treatment solution in the medical literature.

[0067] In one embodiment, the solution extraction module 430 is also used to input the target sentence into a trained pointer model so that the pointer model extracts the solution for the target sentence based on preset prompt information, and outputs a treatment plan corresponding to the prompt information; parses the treatment plan corresponding to the prompt information to obtain a first treatment plan and / or a second treatment plan; counts and cleans the first treatment plan and / or the second treatment plan to obtain the treatment plan in the medical literature.

[0068] In one embodiment, the prompt information includes first prompt information and second prompt information, the first prompt information is used to prompt the output of a separate treatment plan, and the second prompt information is used to prompt the output of an associated treatment plan; the plan extraction module 430 is also used to determine the treatment plan corresponding to the first prompt information as the first treatment plan; and determine the treatment plan corresponding to the second prompt information as the second treatment plan.

[0069] In one embodiment, the treatment plan extraction device 400 also includes a model training module for constructing an initial text classification model; obtaining a sample data set, the sample data set includes multiple positive sample sentences and multiple negative sample sentences, the positive sample sentences and the negative sample sentences are determined based on the analysis of clinical trial registration information; using the sample data set to train the initial text classification model to obtain a trained text classification model.

[0070] In one embodiment, the model training module is also used to obtain sample medical documents containing clinical trial registration numbers, and clinical trial registration information associated with the clinical trial registration number; encode each sentence in the sample medical document to obtain a first feature vector; encode the clinical trial registration information to obtain a second feature vector; and respectively calculate the similarity between each first feature vector and the second feature vector, so as to use sentences with corresponding similarities greater than or equal to a preset similarity threshold as positive sample sentences, and use the remaining sentences except the positive sample sentences as negative sample sentences.

[0071] In the above embodiment, the server obtains the medical literature of the treatment plan to be extracted and inputs the medical literature into the trained text classification model to obtain the target sentence describing the treatment plan in the medical literature, and then extracts the first treatment plan that exists alone and / or the second treatment plan that exists in association with the target sentence, and finally integrates the treatment plan in the medical literature. Since this solution does not require a lot of manpower and is suitable for structured and precise extraction of multiple overlapping entities, it can not only effectively improve the efficiency of information extraction for treatment plans, but also improve the accuracy of treatment plan extraction, providing high-quality data support for subsequent treatment plan retrieval. In order to better implement the treatment plan retrieval method provided in the embodiment of the present invention, on the basis of the treatment plan retrieval method proposed in the embodiment of the present invention, the embodiment of the present invention further provides a treatment plan retrieval device, which can be deployed in Figure 1 In the terminal 102 shown in FIG. Figure 5 As shown, the treatment plan retrieval device 500 includes: A search term acquisition module 510 is used to acquire a target search term; A treatment plan retrieval module 520 is used to determine a target treatment plan that matches a target search term through a treatment plan library, where the treatment plan library is constructed based on treatment plans determined by any of the above treatment plan extraction methods, and the target treatment plan includes a first treatment plan and / or a second treatment plan; The plan display module 530 is used to display the target treatment plan.

[0072] In the above embodiment, since the server has built a highly reliable treatment plan library based on the treatment plans extracted from medical literature, when the user needs to understand the current treatment methods and drug efficacy to guide clinical practice and decision-making, the terminal can quickly search for known treatment plans and their effects, thereby meeting the user's search needs. It should be noted that, for the specific limitations of the treatment plan extraction device, please refer to the limitations of the treatment plan extraction method above, which will not be repeated here. For the specific limitations of the treatment plan retrieval device, please refer to the limitations of the treatment plan retrieval method above, which will not be repeated here. Each module in the above-mentioned treatment plan extraction device or treatment plan retrieval device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. In some embodiments of the present invention, the treatment plan extraction device 400 can be implemented in the form of a computer program. The computer program can be used in Figure 6 The computer device shown in the figure is run on the computer device. The memory of the computer device can store various program modules constituting the treatment plan extraction device 400, for example, Figure 4 The document acquisition module 410, document analysis module 420 and solution extraction module 430 shown in the figure; the computer program composed of each program module enables the processor to execute the steps of the treatment solution extraction method of each embodiment of the present invention described in this specification. For example, Figure 6 The computer device shown can be Figure 4 The document acquisition module 410 in the treatment plan extraction device 400 shown executes step S201. The computer device can execute step S202 through the document analysis module 420. The computer device can execute step S203 through the plan extraction module 430. Among them, the computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external computer device through a network connection. When the computer program is executed by the processor, a treatment plan extraction method is implemented.

[0073] In some embodiments of the present invention, the treatment plan retrieval device 500 may also be implemented in the form of a computer program, and the computer program may also be implemented in the form of a computer program. Figure 6 The computer device shown in the figure is run. The memory of the computer device can store various program modules constituting the treatment plan retrieval device 500, for example, Figure 5 The search term acquisition module 510, the treatment plan retrieval module 520 and the treatment plan display module 530 shown in the figure; the computer program composed of each program module enables the processor to execute the steps of the treatment plan retrieval method of each embodiment of the present invention described in this specification. For example, Figure 6 The computer device shown can be Figure 5 The search term acquisition module 510 in the treatment plan retrieval device 500 shown executes step S201. The computer device can execute step S202 through the plan retrieval module 520. The computer device can execute step S203 through the plan display module 530. Among them, the computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory and a display screen. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external computer device through a network connection. When the computer program is executed by the processor, a treatment plan retrieval method is implemented.

[0074] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0075] In some embodiments of the present invention, a computer device is provided, comprising one or more processors; a memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to execute the steps of the above-mentioned treatment plan extraction method or the steps of the above-mentioned treatment plan retrieval method by the processor. The steps of the treatment plan extraction method here may be the steps of the treatment plan extraction method of each of the above-mentioned embodiments. The steps of the treatment plan retrieval method here may be the steps of the treatment plan retrieval method of each of the above-mentioned embodiments.

[0076] In some embodiments of the present invention, a computer-readable storage medium is provided, storing a computer program, and the computer program is loaded by a processor, so that the processor executes the steps of the above-mentioned treatment plan extraction method, or executes the steps of the above-mentioned treatment plan retrieval method. The steps of the treatment plan extraction method here may be the steps of the treatment plan extraction method of each of the above-mentioned embodiments. The steps of the treatment plan retrieval method here may be the steps of the treatment plan retrieval method of each of the above-mentioned embodiments.

[0077] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Any reference to memory, storage, database or other media used in the embodiments provided by the present invention may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0078] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0079] The above is a detailed introduction to a treatment plan extraction and retrieval method, device, computer equipment and storage medium provided in an embodiment of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for extracting a treatment plan, characterized in that: include: Obtain medical literature on treatment options to be extracted; Inputting the medical literature into a trained text classification model, and outputting target sentences in the medical literature that describe treatment plans; For the target sentence, the first treatment plan that exists alone and / or the second treatment plan that exists in association therein are extracted respectively to obtain the treatment plans in the medical literature.

2. The method according to claim 1, characterized in that The step of extracting the first treatment plan that exists alone and / or the second treatment plan that exists in association with the target sentence to obtain the treatment plan in the medical literature includes: Inputting the target sentence into a trained pointer model, so that the pointer model extracts a treatment plan for the target sentence based on preset prompt information, and outputs a treatment plan corresponding to the prompt information; parsing the treatment plan corresponding to the prompt information to obtain the first treatment plan and / or the second treatment plan; The first treatment plan and / or the second treatment plan are counted and cleaned to obtain the treatment plan in the medical literature.

3. The method according to claim 2, characterized in that The prompt information includes first prompt information and second prompt information, the first prompt information is used to prompt the output of a separate treatment plan, and the second prompt information is used to prompt the output of an associated treatment plan; The step of analyzing the treatment plan corresponding to the prompt information to obtain the first treatment plan and / or the second treatment plan includes: determining a treatment plan corresponding to the first prompt information as the first treatment plan; Determine a treatment plan corresponding to the second prompt information as the second treatment plan.

4. The method according to any one of claims 1 to 3, characterized in that Before inputting the medical literature into the trained text classification model, the method further includes: Build an initial text classification model; Acquire a sample data set, wherein the sample data set includes a plurality of positive sample sentences and a plurality of negative sample sentences, wherein the positive sample sentences and the negative sample sentences are determined based on analysis of clinical trial registration information; The initial text classification model is trained using the sample data set to obtain a trained text classification model.

5. The method according to claim 4, characterized in that The obtaining of the sample data set comprises: Obtaining sample medical literature containing a clinical trial registration number and clinical trial registration information associated with the clinical trial registration number; Encoding each sentence in the sample medical document to obtain a first feature vector; Encoding the clinical trial registration information to obtain a second feature vector; The similarities between the first feature vectors and the second feature vectors are calculated respectively, so as to take the sentences corresponding to the similarities greater than or equal to the preset similarity threshold as positive sample sentences, and take the remaining sentences except the positive sample sentences as negative sample sentences.

6. A treatment plan retrieval method, characterized in that: include: Get the target search term; Determine a target treatment plan that matches the target search term through a treatment plan library, wherein the treatment plan library is constructed based on treatment plans determined by the treatment plan extraction method according to any one of claims 1 to 5, and the target treatment plan includes a first treatment plan and / or a second treatment plan; Present the target treatment regimen.

7. A treatment plan extraction device, characterized in that: include: The literature acquisition module is used to obtain medical literature on the treatment plans to be extracted; A document analysis module, used for inputting the medical document into a trained text classification model and outputting target sentences describing treatment plans in the medical document; The scheme extraction module is used to extract the first treatment scheme that exists alone and / or the second treatment scheme that exists in association with the target sentence, and obtain the treatment scheme in the medical literature.

8. A treatment plan retrieval device, characterized in that: include: A search term acquisition module, used to acquire target search terms; A regimen retrieval module, used to determine a target treatment regimen matching the target search term through a treatment regimen library, wherein the treatment regimen library is constructed based on a treatment regimen determined by the treatment regimen extraction method according to any one of claims 1 to 5, and the target treatment regimen includes a first treatment regimen and / or a second treatment regimen; A plan display module is used to display the target treatment plan.

9. A computer device, characterized in that: The computer device comprises: one or more processors; A memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the treatment plan extraction method according to any one of claims 1 to 5, or to implement the treatment plan retrieval method according to claim 6.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps of the treatment plan extraction method described in any one of claims 1 to 5, or to execute the steps of the treatment plan retrieval method described in claim 6.

Images