Peritoneal dialysis professional knowledge intelligent question and answer generation method based on large language model

Abstract

The invention discloses a peritoneal dialysis professional knowledge intelligent question and answer generation method based on a large language model, and relates to the technical field of artificial intelligence and medical information. Collaborative preprocessing is carried out on the target medical problem and the multi-modal data associated with the target medical problem, joint features are generated, and the target medical problem is a problem associated with peritoneal dialysis; inputting the joint features into a trained medical knowledge question-answer model, and generating an answer to the target medical question; and feeding back the answer to the target medical question to the user. In the invention, the multi-modal data associated with the target medical question provides more information, so that the trained medical knowledge question-answer model can analyze and understand the target medical question from multiple angles, and the accuracy and reliability of answers are improved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and medical information technology, and in particular to a method for generating intelligent questions and answers about peritoneal dialysis professional knowledge based on a large language model. Background Art

[0002] In the process of implementing intelligent question-answering of medical knowledge, problems are often described in a single modality (for example, text information or voice, etc.). The information in a single modality is often not comprehensive enough. When there is missing data, it is impossible to obtain information from data in other modalities to supplement the missing data, resulting in the generated answers being inaccurate. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and specifically provide a method for generating intelligent questions and answers about peritoneal dialysis professional knowledge based on a large language model, as follows:

[0004] 1) In the first aspect, the present invention provides a method for generating intelligent questions and answers about peritoneal dialysis professional knowledge based on a large language model. The specific technical solution is as follows:

[0005] When a user inputs a target medical question and multimodal data associated with the target medical question, the target medical question and the multimodal data associated with the target medical question are collaboratively preprocessed to generate a joint feature; wherein the multimodal data associated with the target medical question includes: at least two of text description information, images, and voice clips associated with the target medical question, and the target medical question is a question related to peritoneal dialysis;

[0006] Inputting the joint features into a trained medical knowledge question-answering model to generate an answer to the target medical question, wherein the trained medical knowledge question-answering model is obtained by training a preset large prediction model;

[0007] Feedback the answers to the targeted medical questions to the users.

[0008] The beneficial effects of the method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model provided by the present invention are as follows:

[0009] The multimodal data associated with the target medical question provides more information, and information from different modalities can complement each other, enabling the trained medical knowledge question-answering model to analyze and understand the target medical question from multiple perspectives, thereby improving the accuracy and reliability of the answer.

[0010] Based on the above solution, the method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model of the present invention can be further improved as follows.

[0011] Furthermore, it also includes:

[0012] After generating the answer to the target medical question, optimizing the answer to the target medical question;

[0013] Provide users with answers to targeted medical questions, including:

[0014] Provide users with optimized answers to target medical questions.

[0015] Furthermore, the process of acquiring the trained medical knowledge question-answering model includes:

[0016] Construct multiple sample question-answer pairs;

[0017] The preset large prediction model is trained based on multiple sample questions and answers to obtain a trained medical knowledge question and answer model.

[0018] Furthermore, the process of acquiring multimodal data associated with the target medical problem includes:

[0019] Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

[0020] 2) In a second aspect, the present invention further provides an intelligent question-answering system for peritoneal dialysis expertise based on a large language model. The specific technical solution is as follows:

[0021] Includes joint feature generation module, answer generation module and feedback module;

[0022] When a user inputs a target medical question and multimodal data associated with the target medical question, the target medical question and the multimodal data associated with the target medical question are collaboratively preprocessed to generate a joint feature; wherein the multimodal data associated with the target medical question includes: at least two of text description information, images, and voice clips associated with the target medical question, and the target medical question is a question related to peritoneal dialysis;

[0023] The answer generation module is used to: input the joint features into a trained medical knowledge question answering model to generate an answer to the target medical question, wherein the trained medical knowledge question answering model is obtained by training a preset large prediction model;

[0024] The feedback module is used to provide the user with the answer to the target medical question.

[0025] Based on the above solution, the intelligent question-answering system for generating peritoneal dialysis professional knowledge based on a large language model of the present invention can also be improved as follows.

[0026] Furthermore, an optimization module is included, which is used to: after generating an answer to the target medical question, optimize the answer to the target medical question;

[0027] The feedback module is specifically used to provide the optimized answer to the target medical question to the user.

[0028] Furthermore, a model training module is included, which is used to:

[0029] Construct multiple sample question-answer pairs;

[0030] The preset large prediction model is trained based on multiple sample questions and answers to obtain a trained medical knowledge question and answer model.

[0031] Furthermore, a data acquisition module is included, which is used to:

[0032] Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

[0033] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor coupled to a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor, so that the electronic device implements any of the above-mentioned methods for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model.

[0034] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for generating intelligent question and answer information about peritoneal dialysis expertise based on a large language model.

[0035] It should be noted that the beneficial effects achieved by the technical solutions of the second to fourth aspects of the present invention and the corresponding possible implementation methods can be found in the above-mentioned technical effects of the first aspect and its corresponding possible implementation methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments of the present invention:

[0037] Figure 1 Schematic diagram of a process for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of the structure of an intelligent question-answering system for peritoneal dialysis professional knowledge based on a large language model according to an embodiment of the present invention;

[0039] Figure 3 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0041] The following describes in detail the technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems using specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following embodiments of the present invention are described in conjunction with the accompanying drawings.

[0042] like Figure 1 As shown, a method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model according to an embodiment of the present invention includes the following steps:

[0043] S1. When a user inputs a target medical question and multimodal data associated with the target medical question, collaboratively preprocessing the target medical question and the multimodal data associated with the target medical question to generate a joint feature; wherein the multimodal data associated with the target medical question includes: at least two of text description information, images, and voice clips associated with the target medical question, and the target medical question is a question related to peritoneal dialysis;

[0044] The target medical question can be input via text or voice. The text description information associated with the target medical question includes: content specifically explaining the target medical question, descriptions of symptoms associated with the target medical question, and text-based diagnostic report documents. The images associated with the target medical question may include CT images and electronic versions of diagnostic report documents. The voice clips associated with the target medical question include: audio describing the symptoms. Depending on the target medical question, the text description information, images, and voice clips associated with the target medical question can be set according to actual circumstances. For example, if the target medical question is "Determine whether it is peritonitis," the text description information may include: "Sudden onset of abdominal pain, fever, nausea, vomiting, etc." The images may include: abdominal CT images and abdominal plain films. The voice clips may include: audio of a patient stating, "Doctor, I've been undergoing peritoneal dialysis recently. I've suddenly had severe stomach pain, a fever, and nausea. The dialysis fluid has also become cloudy. Please help me see if it's peritonitis."

[0045] The target medical problem and the multimodal data associated with the target medical problem are collaboratively preprocessed to generate joint features. The specific implementation process is as follows:

[0046] ① Extract the text semantic vector of the target medical question and record it as the first text semantic vector;

[0047] The target medical question is segmented to obtain multiple word units. Each word unit is converted into a corresponding word vector through the embedding layer of a pre-trained language model (such as BERT or RoBERTa). Self-attention calculation is performed on each word vector through a multi-layer Transformer to capture the long-range dependencies between different word units. The [CLS] labeled vector output by the last layer of Transformer is used as the first text semantic vector. Alternatively, the word vectors corresponding to all word units are average pooled to obtain the first text semantic vector.

[0048] In the existing technology, TF-IDF or Word2Vec is often used to obtain text semantic vectors, but TF-IDF or Word2Vec cannot capture context dependencies. The present invention uses multi-layer Transformer to perform deep context perception, solving the problem of being unable to capture context dependencies. Moreover, through segmented coding (generating word vectors through a pre-trained language model, and then generating the first text semantic vector through a multi-layer Transformer), it can effectively avoid the problem of computing resource explosion caused by the target medical problem and the excessive length of text description information.

[0049] ② Extract the text semantic vector of the text description information, recorded as the second text semantic vector, specifically:

[0050] The text description information is segmented to obtain multiple word units. Each word unit is converted into a corresponding word vector through the embedding layer of a pre-trained language model (such as BERT or RoBERTa). Self-attention calculation is performed on each word vector through a multi-layer Transformer to capture the long-range dependencies between different word units. The [CLS] labeled vector output by the last layer of Transformer is used as the second text semantic vector. Alternatively, the word vectors corresponding to all word units are average pooled to obtain the second text semantic vector.

[0051] ③ Obtain the visual feature matrix of the image through the image feature extraction network, specifically:

[0052] After the image is preprocessed (such as normalization and scaling to a fixed size), it is input into a convolutional neural network (CNN). The local feature map is extracted through the convolutional layer of the convolutional neural network. The local feature map is subjected to global average pooling to obtain a visual feature matrix. Alternatively, after the image is preprocessed (such as normalization and scaling to a fixed size), it is input into the ViT visual model. The ViT visual model divides the preprocessed image into sequences and then generates a visual feature matrix through a self-attention mechanism.

[0053] ④ Perform speech-to-text transcription on the speech segment to obtain speech transcription text, extract the text semantic vector from the speech transcription text, and record it as the third text semantic vector. Specifically:

[0054] The speech transcription text is segmented to obtain multiple word units. Each word unit is converted into a corresponding word vector through the embedding layer of a pre-trained language model (such as BERT or RoBERTa). Self-attention calculation is performed on each word vector through a multi-layer Transformer to capture the long-range dependencies between different word units. The [CLS] labeled vector output by the last layer of Transformer is used as the third text semantic vector. Alternatively, the word vectors corresponding to all word units are average pooled to obtain the third text semantic vector.

[0055] ④ Fuse the first text semantic vector, the second text semantic vector, the third text semantic vector and the visual feature matrix to generate joint features. Specifically:

[0056] a. Concatenate the first, second, and third text semantic vectors to obtain a concatenated vector. Then, based on the cross-attention mechanism, calculate the attention weight matrix of the concatenated vector for the image. Generate text-guided visual features based on the attention weight matrix of the concatenated vector for the image. Calculate the attention weight matrix of the image for the concatenated vector. Generate visually guided text features based on the attention weight matrix of the image for the concatenated vector.

[0057] Among them, the attention weight matrix of the splicing vector to the image is calculated, and the text-guided visual features are generated according to the attention weight matrix of the splicing vector to the image. The specific implementation process is as follows:

[0058] b. Map the splicing vector to the Query vector through the linear layer, flatten the visual feature matrix, and then generate the Key matrix through the Key generation linear layer (the Key generation linear layer is also a linear transformation layer, which is used to linearly transform the input feature vector to generate the Key matrix) for the flattened visual feature matrix, and generate the Value matrix through the Value generation linear layer (the Value generation linear layer is a linear transformation layer, which is used to linearly transform the input feature vector to generate the Value matrix) after the flattened visual feature matrix. Then, generate the Key matrix and the Value matrix respectively through two linear layers, calculate the dot product of the Query vector and each Key value in the Key matrix, and obtain multiple products. Each product is input into the Softmax function to obtain the weight corresponding to each key value, thereby obtaining the attention weight matrix of the splicing vector to the image, and perform weighted summation of the weight corresponding to each key value with the Value value at the same position in the Value matrix to generate text-guided visual features.

[0059] Among them, the attention weight matrix of the image to the splicing vector is calculated, and the visually guided text features are generated according to the attention weight matrix of the image to the splicing vector. The specific implementation process is as follows:

[0060] The visual feature matrix is mapped to the Query vector through a linear layer, the splicing vector is passed through the Key generation linear layer to generate a Key matrix, and the splicing vector is passed through the Value generation linear layer to generate a Value matrix. The dot product of the Query vector and each Key value in the Key matrix is calculated to obtain multiple products. Each product is input into the Softmax function to obtain the weight corresponding to each key value, thereby obtaining the image's attention weight matrix for the splicing vector. The weight corresponding to each key value is weightedly summed with the Value value at the same position in the Value matrix to generate visually guided text features.

[0061] In the prior art, multimodal fusion methods (such as feature splicing and simple weighted averaging) often ignore the semantic alignment between modalities, resulting in a disconnect between the image and the text description. In the present invention, after obtaining the attention weight matrix of the splicing vector to the image, the weight corresponding to each key value is weightedly summed up for the Value value at the same position in the Value matrix, and, after obtaining the attention weight matrix of the image to the splicing vector, the weight corresponding to each key value is weightedly summed up for the Value value at the same position in the Value matrix. This can effectively achieve semantic alignment between modalities and prevent the disconnect between the image and the text description.

[0062] c. Concatenate the visual features and text features, and then input them into the fully connected layer to obtain the joint features.

[0063] S2. Inputting the joint features into a trained medical knowledge question-answering model to generate an answer to the target medical question, wherein the trained medical knowledge question-answering model is obtained by training a preset large prediction model;

[0064] Optionally, in the above technical solution, before inputting the joint features into the trained medical knowledge question-answering model, the method further includes: generating a contextual semantic vector of the historical conversation records using a temporal modeling network based on the user's historical conversation records; and modifying the joint features based on the contextual semantic vector to obtain a modified joint feature. Then:

[0065] S2 specifically includes: inputting the modified joint features into the trained medical knowledge question-answering model to generate answers to target medical questions.

[0066] Among them, based on the user's historical conversation records, the temporal modeling network is used to generate the contextual semantic vector of the historical conversation records; based on the contextual semantic vector, the joint features are modified to obtain the modified joint features. The specific implementation process is as follows:

[0067] ① Construct the user's historical conversation records into a time series D, D = [u1,u2...u T ],u t Represents the content of the tth round of dialogue, including the medical question input by the user, the multimodal data associated with the medical question, and the answer.

[0068] ② Use a text encoder (such as BERT) to extract the semantic vector of each round of dialogue and obtain a semantic vector sequence;

[0069] ③ Use a bidirectional LSTM or Transformer encoder to process the semantic vector sequence and extract the contextual semantic vector.

[0070] ④ Map the context semantic vector and the joint feature to the same feature space, and generate the dynamic weight matrix g using the following formula:

[0071] g=σ(W g ·[c;f joint ]+b g )

[0072] Among them, c represents: context semantic vector, f joint Represents: Joint features, W g and b g are coefficients, W g The value of b g The value of can be obtained through multiple data experiments, [c; f joint ] represents: context semantic vector c and joint feature f joint Perform splicing operation, σ(W g ·[c;f joint ]+b g ) means: (W g ·[c;f joint ]+b g ) is brought into the Sigmoid function.

[0073] ⑤ According to the dynamic weight matrix g, the context semantic vector and the joint feature are weighted and modified to generate the modified joint feature f d , which is specifically achieved through the following formula:

[0074] f d =g⊙f joint +(1-g)⊙c

[0075] Among them, ⊙ represents element-by-element multiplication, that is, the elements of corresponding positions of two vectors of the same dimension (representing the context semantic vector and the joint feature mapped to the same feature space) are multiplied.

[0076] Dynamically adjust the joint feature f through the context semantic vector c joint The weight of the data is used to strengthen the information related to the user's long-term intention, reduce the deviation of the target medical problem, and improve the accuracy.

[0077] S3. Feedback the answer to the target medical question to the user.

[0078] Optionally, in the above technical solution, the following is further included:

[0079] After generating the answer to the target medical question, the answer to the target medical question is optimized. The specific optimization process is as follows:

[0080] Entities in the generated answers to target medical questions are extracted and semantically matched with the pre-established medical knowledge graph to obtain matching results. The matching results are then checked for logical conflicts. If so, the answers to target medical questions are optimized based on the logical conflicts.

[0081] The process of extracting entities from the generated answers to the target medical questions is as follows:

[0082] Use a BiLSTM-CRF or BERT-based sequence labeling model to identify entities from the generated answers to target medical questions. The identified entities include names of people, places, and drugs.

[0083] The process of establishing a medical knowledge graph is as follows:

[0084] Extract entity-relationship triplets (e.g., drug A, treatment, disease B) from medical databases (e.g., medical literature repositories). Crawl medical texts, such as medical papers and user reviews, to obtain crawling results. Clean the obtained entity-relationship triplets and crawling results to obtain cleaned data. Use the BERT-BiLSTM-CRF model to extract entities from the cleaned data, including diseases and drugs. Use an information extraction model (e.g., OpenIE) to extract relationships between different entities from the cleaned data. Also, extract numerical attributes (e.g., drug dosage: "100 mg each time") and categorical attributes (e.g., drug type: "antibiotic") for different entities from the cleaned data. Define the hierarchy between entities and set relationship types (predicates) and constraints (e.g., a "contraindications" relationship only allows connections between "drug" and "disease" nodes). Use Neo4j or Amazon Neptune to store nodes, edges, numerical attributes, and categorical attributes. Create a node for each entity, each containing a unique ID, numerical attributes, and categorical attributes. Create edges based on the relationships between different entities to generate a medical knowledge graph.

[0085] Traditional medical knowledge bases rely on manual construction (such as UMLS), which are updated laggingly and have difficulty covering the latest research results, and lack multimodal data integration. This invention can efficiently and automatically construct a medical knowledge graph.

[0086] The entities in the generated answers to the target medical questions are semantically matched with the pre-established medical knowledge graph. The specific implementation is as follows:

[0087] Use algorithms such as TransE and GraphSAGE to map the nodes in the medical knowledge graph into low-dimensional vectors, calculate the cosine similarity between the vector corresponding to the entity in the answer (specifically, it can be obtained through the acquisition process of the first text semantic vector, the second text semantic vector, or the third text semantic vector) and the vector of the node in the medical knowledge graph, and screen the candidate nodes with a similarity higher than the preset similarity threshold to obtain the matching results. If the attributes (categorical attributes and numerical attributes) corresponding to the entity in the answer conflict with the attributes of the candidate node, it is determined that there is a logical conflict. The answer to the target medical question is optimized based on the logical conflict. Specifically, the cause of the logical conflict and the suggested modification content can be highlighted to facilitate manual correction.

[0088] Provide users with answers to targeted medical questions, including:

[0089] Provide users with optimized answers to target medical questions.

[0090] Optionally, in the above technical solution, the process of obtaining the trained medical knowledge question-answering model includes:

[0091] Construct multiple sample question-answer pairs;

[0092] The preset large prediction model is trained based on multiple sample questions and answers to obtain a trained medical knowledge question and answer model.

[0093] Each sample question-and-answer pair includes a medical question and a corresponding answer. The medical question is associated with at least one modality of data. After processing the medical question and at least one modality of data associated with the medical question, the corresponding joint features are obtained. Then, based on each joint feature and the corresponding answer, the preset large prediction model is trained to obtain a trained medical knowledge question-and-answer model.

[0094] The preset large prediction model can be a Transformer architecture, an autoregressive model, a GPT series language large model, or a multimodal large model, etc., and can be set according to actual conditions.

[0095] Optionally, in the above technical solution, the process of acquiring multimodal data associated with the target medical problem includes:

[0096] Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

[0097] The present invention is described by way of another embodiment.

[0098] Peritoneal dialysis, a commonly used renal replacement therapy, requires patients to strictly follow their doctor's instructions in their daily lives, monitoring fluid balance, electrolyte changes, and complication risks. Existing peritoneal dialysis management systems often rely on fixed procedures and preset parameters, lacking real-time, intelligent answers to individualized patient questions. Furthermore, physicians face challenges such as the sheer volume of information and rapid updates in professional knowledge when answering patient questions and adjusting treatment plans.

[0099] In recent years, with the rapid development of large language models (such as Deepseek, LLaMA, and the GPT series), they have demonstrated outstanding performance in natural language generation and understanding, enabling high-quality semantic parsing and human-computer dialogue. Combining large language models with peritoneal dialysis management systems to provide real-time, personalized medical consultations through intelligent question-and-answer (Q&A) not only improves patient self-management but also supports clinical decision-making, thereby improving treatment outcomes and reducing the risk of complications. Therefore, we propose a method for generating intelligent Q&A on peritoneal dialysis expertise based on large language models. The method includes the following steps:

[0100] 1) Automatically or semi-automatically collect data such as literature, clinical guidelines, expert consensus, case reports, etc. in the field of peritoneal dialysis to build a domain knowledge base.

[0101] 2) Clean, format, annotate and segment the domain knowledge base to ensure data quality and facilitate subsequent use.

[0102] 3) Utilize a pre-trained large language model and fine-tune it specifically for peritoneal dialysis, medical knowledge, and question-and-answer scenarios to build a dedicated medical knowledge base.

[0103] 4) Receive natural language input from patients or medical staff, and after semantic analysis and context understanding, work together with the large language model to generate accurate and personalized answers.

[0104] 5) Combine the professional knowledge base and preset rules to verify and correct the initially generated answers to ensure the accuracy and authority of the content.

[0105] 6) Provide a friendly human-computer interaction interface, support multiple input methods such as text, voice, and pictures, and feedback the final answer. At the same time, record user feedback for subsequent system optimization.

[0106] The application scenarios are as follows:

[0107] 1) Complications Q&A scenario:

[0108] User input: "What should I do if I have abdominal pain after dialysis?";

[0109] Execution process:

[0110] Entity recognition: Extract "abdominal pain" and "post-dialysis" as key entities.

[0111] Search enhancement: Match related entries of "peritonitis" and "catheter infection" from the knowledge base.

[0112] The generated answer is as follows:

[0113] Initial advice: "Abdominal pain may be a symptom of peritonitis. Please check immediately whether the dialysate is turbid...etc.";

[0114] Source annotation: "Refer to Section xxx of the Peritoneal Dialysis Guidelines".

[0115] Personalized supplement: If the user's history shows multiple infections, an additional prompt will be added: "It is recommended to contact the attending physician to review the catheter."

[0116] 2) Multimodal input processing scenario:

[0117] A user uploaded a picture of their dialysis record sheet and asked, "Is it normal for the ultrafiltration volume to decrease this week?"

[0118] Execution process:

[0119] OCR extracts the ultrafiltration volume data in the table and, combined with the patient's weight trend, generates an analysis: "The ultrafiltration volume decreased by 10% this week. It is recommended to adjust the dialysate glucose concentration to 2.5%...etc."

[0120] The beneficial effects of this embodiment are as follows:

[0121] 1) Real-time and intelligent: Leveraging the natural language processing advantages of large language models, this system provides real-time and accurate answers to peritoneal dialysis-related questions, improving patient satisfaction and safety.

[0122] 2) Reduce the burden on medical staff: The system can automatically handle a large number of common problems and provide auxiliary reference for medical staff, allowing them to focus more on difficult clinical decision-making and emergency response.

[0123] 3) Wide range of applications: The system is suitable for various scenarios such as home self-service peritoneal dialysis, hospital remote monitoring and clinical decision support, and has broad market application prospects.

[0124] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the scope of protection of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0125] like Figure 2 As shown, an intelligent question-answering system 200 for peritoneal dialysis professional knowledge based on a large language model according to an embodiment of the present invention includes a joint feature generation module 201, an answer generation module 202, and a feedback module 203;

[0126] The joint feature generation module 201 is configured to: when a user inputs a target medical question and multimodal data associated with the target medical question, perform collaborative preprocessing on the target medical question and the multimodal data associated with the target medical question to generate a joint feature; wherein the multimodal data associated with the target medical question includes at least two of text description information, images, and voice clips associated with the target medical question, and the target medical question is a question related to peritoneal dialysis;

[0127] The answer generation module 202 is used to: input the joint features into a trained medical knowledge question answering model to generate an answer to the target medical question, wherein the trained medical knowledge question answering model is obtained by training a preset large prediction model;

[0128] The feedback module 203 is used to feed back the answer to the target medical question to the user.

[0129] Optionally, the above technical solution further includes an optimization module, which is used to: after generating the answer to the target medical question, optimize the answer to the target medical question;

[0130] The feedback module 203 is specifically used to feed back the optimized answer to the target medical question to the user.

[0131] Optionally, the above technical solution further includes a model training module, which is used to:

[0132] Construct multiple sample question-answer pairs;

[0133] The preset large prediction model is trained based on multiple sample questions and answers to obtain a trained medical knowledge question and answer model.

[0134] Optionally, the above technical solution further includes a data acquisition module, which is used to:

[0135] Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

[0136] In another embodiment, comprising:

[0137] Data collection module: Automatically or semi-automatically collects literature, clinical guidelines, expert consensus, case reports and other data in the field of peritoneal dialysis to build a domain knowledge base.

[0138] Data preprocessing module: cleans, formats, annotates and segments the domain knowledge base to ensure data quality and facilitate subsequent use.

[0139] Large language model module: Utilizes a pre-trained large language model and fine-tunes it specifically for peritoneal dialysis, medical knowledge, and question-and-answer scenarios to build a dedicated medical knowledge base.

[0140] Intelligent question-answering engine: Receives natural language input from patients or medical staff, performs semantic analysis and contextual understanding, and then works with a large language model to generate accurate and personalized answers.

[0141] Knowledge verification module: Combines professional knowledge base and preset rules to verify and correct the initially generated answers to ensure the accuracy and authority of the content.

[0142] User interaction module: provides a friendly human-computer interaction interface, supports multiple input methods such as text, voice, and pictures, and provides feedback on the final answer. It also records user feedback for subsequent system optimization.

[0143] It should be noted that the beneficial effects of the peritoneal dialysis professional knowledge intelligent question and answer generation system 200 based on a large language model provided in the above embodiment are the same as the beneficial effects of the above-mentioned peritoneal dialysis professional knowledge intelligent question and answer generation method based on a large language model, which will not be repeated here. In addition, when the system provided in the above embodiment realizes its functions, it only uses the division of the above-mentioned functional modules as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0144] Among them, the peritoneal dialysis professional knowledge intelligent question and answer generation system based on a large language model of the present invention can be a computer program (including program code) running on a computer device. For example, the peritoneal dialysis professional knowledge intelligent question and answer generation system based on a large language model of the present invention is an application software that can be used to execute the corresponding steps of the peritoneal dialysis professional knowledge intelligent question and answer generation method based on a large language model of the present invention.

[0145] In some embodiments, the peritoneal dialysis professional knowledge intelligent question and answer generation system based on a large language model of the present invention can be implemented in a combination of software and hardware. As an example, the peritoneal dialysis professional knowledge intelligent question and answer generation system based on a large language model of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the peritoneal dialysis professional knowledge intelligent question and answer generation method based on a large language model of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic components.

[0146] The modules described in the embodiments of the present invention may be implemented in software or hardware, and the name of a module does not necessarily limit the module itself.

[0147] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any of the above-mentioned methods for intelligent question-answering generation of peritoneal dialysis professional knowledge based on a large language model is implemented. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the method for intelligent question-answering generation of peritoneal dialysis professional knowledge based on a large language model shown in any embodiment of the present invention by calling the computer program.

[0148] In an alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data exchange between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the number of transceivers 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0149] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0150] Bus 4002 may include a path for transmitting information between the above components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 In the figure, only one thick line is used to represent the bus 4002, but this does not mean that there is only one bus or one type of bus.

[0151] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0152] The memory 4003 is used to store application code (computer program) for executing the solution of the present invention, and is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0153] Among them, the electronic device can also be a terminal device, and the terminal device can be any device that can install applications, including at least one of a smartphone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0154] It should be noted that Figure 3 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0155] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements any of the above-mentioned methods for generating intelligent question and answer on peritoneal dialysis professional knowledge based on a large language model.

[0156] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0157] In an exemplary embodiment, a computer program product or computer program is also provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the aforementioned methods for generating intelligent questions and answers about peritoneal dialysis expertise based on a large language model.

[0158] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0159] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0160] The computer-readable storage medium provided in the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EEPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or component.

[0161] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0162] The above description is merely a preferred embodiment of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also includes other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

[0163] It should be noted that the terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects and to define a specific order or precedence. Where appropriate, the order used for similar objects may be interchanged, such that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0164] Those skilled in the art will appreciate that the present invention may be implemented as a system, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the present invention may be implemented in the form of a computer program product embodied in one or more computer-readable media containing computer-readable program code.

[0165] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model, characterized in that: include: When a user inputs a target medical question and multimodal data associated with the target medical question, the target medical question and the multimodal data associated with the target medical question are collaboratively preprocessed to generate a joint feature; wherein the multimodal data associated with the target medical question includes: at least two of text description information, images, and voice clips associated with the target medical question, and the target medical question is a question related to peritoneal dialysis; Inputting the joint features into a trained medical knowledge question-answering model to generate an answer to the target medical question, wherein the trained medical knowledge question-answering model is obtained by training a preset large prediction model; The answer to the target medical question is fed back to the user.

2. The method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model according to claim 1, characterized in that: Also includes: After generating the answer to the target medical question, optimizing the answer to the target medical question; Providing the user with an answer to the target medical question, comprising: The optimized answer to the target medical question is provided to the user.

3. The method for generating intelligent question and answer about peritoneal dialysis professional knowledge based on a large language model according to claim 1 or 2, characterized in that: The process of acquiring a trained medical knowledge question-answering model includes: Construct multiple sample question-answer pairs; The preset large prediction model is trained based on multiple sample questions and answers to obtain the trained medical knowledge question and answer model.

4. The method for generating intelligent question and answer of peritoneal dialysis professional knowledge based on a large language model according to claim 1 or 2, characterized in that: The process of acquiring multimodal data associated with the target medical problem includes: Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

5. An intelligent question-answering system for peritoneal dialysis professional knowledge based on a large language model, characterized by: Includes joint feature generation module, answer generation module and feedback module; The joint feature generation module is configured to: when a user inputs a target medical problem and multimodal data associated with the target medical problem, perform collaborative preprocessing on the target medical problem and the multimodal data associated with the target medical problem to generate a joint feature; wherein the multimodal data associated with the target medical problem includes: at least two of text description information, images, and voice clips associated with the target medical problem, and the target medical problem is a problem associated with peritoneal dialysis; The answer generation module is used to: input the joint features into a trained medical knowledge question answering model to generate an answer to the target medical question, wherein the trained medical knowledge question answering model is obtained by training a preset large prediction model; The feedback module is used to feed back the answer to the target medical question to the user.

6. The peritoneal dialysis professional knowledge intelligent question-answering system based on a large language model according to claim 5, characterized in that: The invention also includes an optimization module, wherein the optimization module is used to: after generating the answer to the target medical question, optimize the answer to the target medical question; The feedback module is specifically configured to feed back the optimized answer to the target medical question to the user.

7. The peritoneal dialysis professional knowledge intelligent question-answering system based on a large language model according to claim 5 or 6, characterized in that: It also includes a model training module, which is used to: Construct multiple sample question-answer pairs; The preset large prediction model is trained based on multiple sample questions and answers to obtain the trained medical knowledge question and answer model.

8. The peritoneal dialysis professional knowledge intelligent question-answering system based on a large language model according to claim 5 or 6, characterized in that: It also includes a data acquisition module, which is used to: Receive multimodal data associated with a target medical problem input by a user through a human-computer interaction interface.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for generating intelligent question and answer of peritoneal dialysis professional knowledge based on a large language model as described in any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for generating intelligent questions and answers about peritoneal dialysis professional knowledge based on a large language model as described in any one of claims 1 to 4.

Images