Interaction system for tumor gene detection report interpretation scene

Abstract

The invention provides an interaction system for a tumor gene detection report interpretation scene. The interaction system comprises a user instruction identification module, a multi-modal file analysis module, a knowledge base module, a language processing module and a result feedback module. After a user inputs an instruction and uploads personal information such as medical information and a gene detection report, the system performs semantic analysis through an instruction identification module, extracts key information through a file analysis module, matches related tumor gene data in combination with a knowledge base module, and extracts individual tumor gene information. And the language processing module converts the individual tumor gene information into natural language answers, and the natural language answers are returned to the user by the result feedback module. All the modules of the system work cooperatively, multiple rounds of interaction are supported, clearer instructions or more comprehensive information is obtained through guiding type inquiry and guiding type information supplement, the answer content is dynamically updated, and it is ensured that a personalized feedback result which is clear in output logic, easy to understand and capable of meeting user requirements is obtained.

Description

Technical Field

[0001] The present invention belongs to the field of healthcare informatics, and particularly relates to an interactive system for medical test report interpretation scenarios, and more particularly to an interactive system for tumor gene test report interpretation scenarios. Background Art

[0002] With the rapid development of precision medicine, tumor gene testing has become a core part of cancer diagnosis and treatment. Currently, the detection products on the market mainly analyze markers such as gene mutations (e.g., EGFR, KRAS mutations), tumor mutation burden (TMB), microsatellite instability (MSI), etc. through high-throughput sequencing (NGS) technology, providing a basis for targeted therapy, immunotherapy, and prognosis evaluation. However, the existing report interpretation systems for tumor gene test reports generally have the following problems.

[0003] 1. Traditional systems rely on manual or semi-automated processes to analyze gene test reports, and it is difficult to efficiently process multimodal data (such as reports in PDF, picture formats), resulting in incomplete or inaccurate extraction of key information. For example, information such as tumor mutation burden (TMB) and gene mutation sites in unstructured text needs to rely on manual secondary entry, which is inefficient and error-prone.

[0004] 2. Existing systems are mostly in a one-way Q&A mode. Traditional text input Q&A systems allow users to ask questions in text. Based on the user's one-time input, the system generates answers based on a general language model or rule engine, but there is a lack of a dynamic guidance mechanism in this process. When the user's input is vague or incomplete (such as only uploading a report but not clarifying the requirements), the system cannot actively guide the user to supplement necessary information (such as tumor stage, treatment history), resulting in low interaction efficiency.

[0005] 3. General large language models lack professionalism in the tumor field and are difficult to accurately integrate multi-source medical knowledge (such as drug sensitivity evidence, clinical trial data, etc.). Therefore, the accuracy and depth of answers are limited and cannot meet professional needs. For example, although the comparative document CN119493847A analyzes gene reports based on a large language model, it lacks a customized knowledge base for tumor scenarios, resulting in one-sided answers and lack of practicality.

[0006] 4. Traditional gene test reports highly rely on professionals to adjust and verify the report interpretation results, which is inefficient, costly, and has poor accessibility, and the results are easily affected subjectively. Summary of the Invention

[0007] In order to solve the problems of insufficient information processing ability, poor user interaction experience, lack of professionalism in the tumor field, and high dependence on professionals when interpreting tumor gene test reports in the prior art, the present invention provides an interactive system for tumor gene test report interpretation scenarios.

[0008] The technical solution adopted by the present invention is: an interactive system for interpreting tumor gene detection reports, including: A user instruction recognition module, which is used to perform semantic analysis on user instructions, determine the user's identity and obtain the user's intention, and generate a knowledge base matching logic based on the user's intention; A multimodal file parsing module, which is used to parse the personal information uploaded by the user and obtain the parsed content, and extract knowledge base matching information using the deployed large language model based on the parsed content; the personal information includes individual medical information and tumor gene detection report files; A knowledge base module, which is used to extract individual tumor gene information from the knowledge base according to the knowledge base matching logic and knowledge base matching information; the knowledge base includes a clinical case library, a drug database, a gene mutation database, a clinical trial database, and a medical literature library; A language processing module, which is used to generate answer data according to the individual tumor gene information, and use the large language model to perform logical reasoning and language polishing on the answer data to generate answer content; A result feedback module, including an inquiry unit, an information supplement unit, and a content return unit; the content return unit is used to stream the answer content back to the user; when the user instruction recognition module determines that the user instruction is a fuzzy semantic, the inquiry unit conducts a guided inquiry on the user and updates the user instruction; when the multimodal file parsing module determines that the obtained parsed content is an information-insufficient file, the information supplement unit conducts a guided information supplement on the user and updates the personal information; after one or more guided inquiries and / or guided information supplements, the language processing module updates the answer content according to the updated user instruction and / or updated personal information, and streams the updated answer content back to the user through the content return unit.

[0009] Preferably, the user instruction recognition module generates the knowledge base matching logic through the following steps: Perform semantic analysis on the obtained user instructions to obtain the user's intention; Train a classification model based on the historical data of user instructions and a medical domain corpus to classify the user's intention; Determine the type of knowledge base to be called and the data priority according to the classification result, and generate a knowledge base matching logic corresponding to the user's intention.

[0010] Preferably, the multimodal file parsing module includes: A format conversion unit, which is used to convert the personal information in picture format or PDF format uploaded by the user into structured text; A key indicator extraction unit is used to extract key indicators from structured text through a large language model and store them in a database as knowledge base matching information; the key indicators include at least one of tumor type, tumor stage, tumor grade, treatment history, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, and minimal molecular residual status.

[0011] Preferably, the drug database includes at least one of the relationship between genes and drug sensitivity / resistance, the evidence level of the relationship between genes and drug sensitivity, drug instructions, medical guideline information, drug clinical trial results, and drug economics data.

[0012] Preferably, the drug economics data includes at least one of FDA approval information, NMPA approval information, and medical insurance coverage status; the medical insurance coverage status includes at least one of the approved drug manufacturer, approved reference price, approved cancer type, and approved target.

[0013] Preferably, the knowledge base module further includes a similar case matching unit, which is used to match similar cases from the clinical case library according to the extracted individual tumor gene information and embed the matching results into the answer data.

[0014] Preferably, the language processing module further includes a language optimization unit, which is used to adjust the expression mode of the answer content according to the user's identity, where the user's identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the language optimization unit converts the professional language in the answer content into popular language.

[0015] Preferably, the answer content includes gene information, targeted drug usage tips, and similar case information; The gene information includes at least one of the association information between genes and tumors, the evidence of the association between genes and tumors, and gene variant analysis information; The targeted drug usage tips include at least one of the relationship between genes and targeted drug sensitivity / resistance, the evidence level of the relationship between genes and targeted drug sensitivity, targeted drug instructions, targeted drug medical guideline information, targeted drug clinical trial results, and targeted drug economics data.

[0016] Preferably, the method for the result feedback module to provide guided information supplementation to the user includes: The multimodal file parsing module determines whether the knowledge base matching information extracted from the current parsed content meets a preset minimum information set; if not, the result feedback module sends a supplementary information request to the user, and the request includes uploading missing key indicators and / or confirming ambiguous key indicators.

[0017] Preferably, the minimum information set includes at least one key indicator, and the key indicators include tumor type, tumor stage, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, and minimal molecular residual status.

[0018] Preferably, the result feedback module further includes a prediction unit, which actively predicts and proposes a suggestive inquiry based on the user's intention and the parsed content, and embeds the suggestive inquiry into the guided inquiry.

[0019] Preferably, the result feedback module further includes an adaptation unit; it is used to adjust the expression mode of the guided inquiry and the guided information supplement according to the user's identity, where the user's identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the adaptation unit converts the professional language in the guided inquiry and the guided information supplement into popular language.

[0020] Advantages of the present invention: 1. The setting of the multi-modal document parsing module combines the OCR technology with the large language model to automatically parse the unstructured data in personal information in different formats such as pictures / PDFs, accurately extract key indicators such as tumor type and gene variation, reduce manual intervention, and improve the system's multi-modal data processing ability.

[0021] 2. The result feedback module realizes dynamic interactive guidance by conducting guided inquiries and guided information supplements for users, triggers the guided answer mechanism when the user's input is vague or information is missing, actively asks the user to supplement necessary information such as tumor stage and treatment history, and optimizes the answer content through multiple interactions.

[0022] 3. The knowledge base module aggregates multi-source tumor domain knowledge bases (clinical cases, pharmacoeconomics, medical guidelines, etc.), and combines the logical reasoning and language optimization capabilities of the large language model arranged by the language processing module to automatically generate interpretation results that are both professional and popular.

[0023] 4. The language optimization unit in the language processing module supports adaptively adjusting the expression mode of the answer content according to the user's identity (doctor / patient), realizing the differential and personalized output of the answer content, and meeting the needs of different user groups. For example, converting the professional language expression of "EGFR p.L858R mutation is sensitive to osimertinib (Level A evidence)" into the popular language expression of "Your gene test results show that you are suitable for using osimertinib, and this drug has been recommended by the authoritative guidelines."

[0024] In summary, the interactive system provided by the present invention automatically helps users efficiently interpret tumor gene detection reports by enhancing the multi-modal data analysis ability, adding a dynamic interaction guidance mechanism, and integrating multi-source tumor domain knowledge bases, and generates clear and easy-to-understand response content through a multi-interaction response mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a structural diagram of an interactive system for tumor gene detection report interpretation scenarios provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following specific embodiments illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0027] This embodiment provides an interactive system for tumor gene detection report interpretation scenarios. The interactive system includes a user instruction recognition module, a multi-modal file parsing module, a knowledge base module, a language processing module, and a result feedback module. The user inputs an instruction and uploads personal information including individual medical information and a tumor gene detection report file. The user instruction is semantically analyzed by the user instruction recognition module, and the personal information is parsed by the multi-modal file parsing module to obtain the corresponding knowledge base matching logic and knowledge base matching information. The knowledge base module extracts individual tumor gene information from the knowledge base based on this, and generates corresponding response data and response content in natural language through the language processing module. Finally, the response content is streamed back to the user through the result feedback module. Among them, with the joint action of the user instruction recognition module, the multi-modal file parsing module, and the result feedback module, and the support of the knowledge base module and the language processing module, this embodiment also supports multiple interactions between the system and the user during the response process, conducts guided inquiries and guided information supplementation for the user, and updates the response content according to the supplemented personal information and more specific user instructions with the support of the knowledge base module and the language processing module to generate clear, easy-to-understand, and user identity-adapted response content, and then returns it to the user.

[0028] In a specific embodiment, the user instruction recognition module is used to perform semantic analysis on user instructions, determine the user identity and obtain the user intention, and generate a knowledge base matching logic based on the user intention. After performing semantic analysis on the user instructions, the user instruction recognition module believes that the user has weak knowledge of tumors, and thus determines that the user identity is a non-medical professional (such as a patient). The user instruction recognition module can generate a knowledge base matching logic through the following steps: perform semantic analysis on the obtained user instructions to obtain the user intention; train a classification model based on the historical data of the user instructions and the medical domain corpus to classify the user intention; determine the type of knowledge base to be called and the data priority according to the classification result, and generate a knowledge base matching logic corresponding to the user intention. However, when the user instruction recognition module performs semantic analysis on the user instructions and determines that the user instructions are of fuzzy semantics and it is difficult to clearly obtain the user intention, the result feedback module conducts a guided inquiry on the user and updates the user instructions. For example, the user instruction can be a simple "Help me interpret this NGS test report", or it can be a tendentious description, such as "Help me interpret this NGS report and recommend suitable drugs and corresponding clinical trials". When the user instruction is a fuzzy or incomplete question, the result feedback module is triggered to conduct a guided inquiry on the user, so as to obtain a more clear user intention, which helps to generate a knowledge base matching logic corresponding to the user intention.

[0029] In a specific embodiment, the multimodal file parsing module is used to parse the personal information uploaded by the user and obtain the parsed content, and extract knowledge base matching information using the pre-arranged large language model based on the parsed content. The personal information includes individual medical information and tumor gene detection report files. The individual medical information generally includes basic information of the test sample (such as name, gender, age, family history, medical records, sample type, etc.) and test information (such as products, hospitals, institutions, projects), etc., so that key indicators including gender, age, genetic factors, tumor type, tumor stage, tumor grade, etc. can be extracted as knowledge base matching information. The tumor gene detection report file generally includes the corresponding gene information and the detection results of its mutations (such as nucleotide sequence information, amino acid sequence information, exon / intron, transcript, mutation type, mutation frequency, etc., TMB, MSI), etc., so that key indicators including gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, minimal molecular residual status, etc. can be extracted as knowledge base matching information. However, when the multimodal file parsing module parses the uploaded personal information and determines that the parsed content is an information-insufficient file and does not meet the minimum standard for personal information interpretation, the result feedback module guides the user to supplement information and updates the personal information. More specifically, when the multimodal file parsing module determines whether the knowledge base matching information extracted from the current parsed content meets the preset minimum information set; if not, the result feedback module sends a supplementary information request to the user, and the request includes uploading missing key indicators and / or confirming ambiguous key indicators. Among them, the minimum information set includes at least one key indicator, and the key indicators include tumor type, tumor stage, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, minimal molecular residual status. For example, when the user only uploads the tumor gene detection report file and does not upload the medical record information, key indicators including tumor type, tumor stage or tumor grade, etc. cannot be extracted from the parsed content, then the result feedback module is triggered to guide the user to supplement information, so as to obtain more comprehensive parsed content, which helps the large language model to extract more comprehensive knowledge base matching information from the parsed content.

[0030] In a preferred embodiment, the multimodal file parsing module includes a format conversion unit and a key index extraction unit. The format conversion unit is used to convert personal information in different formats (such as pictures, PDFs, tables, texts, etc.) uploaded by users into structured text, for example, by using ORC technology to identify files in picture or PDF format. The key index extraction unit is used to extract key indexes from the structured text through a large language model and store them in a database as knowledge base matching information. The key indexes include at least one of tumor type, tumor stage, tumor grade, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, minimal molecular residual status.

[0031] In a specific embodiment, the knowledge base module is used to extract individual tumor gene information from the knowledge base according to the knowledge base matching logic and the knowledge base matching information. The knowledge base includes a clinical case library, a drug database, a gene variation database, a clinical trial database, and a medical literature library.

[0032] In a preferred embodiment, the drug database includes at least one of the relationship between genes and drug sensitivity / resistance, the evidence level of the relationship between genes and drug sensitivity, drug instructions, medical guideline information, drug clinical trial results, and pharmacoeconomic data. The pharmacoeconomic data includes at least one of FDA approval information, NMPA approval information, and medical insurance coverage status; the medical insurance coverage status includes at least one of the approved drug manufacturers, approved reference prices, approved cancer types, and approved target indications. The addition of pharmacoeconomic data helps the interactive system, for example, to recommend treatment options with different costs to patients according to the drug economic level, and to compare the efficacy, safety, and ease of purchase for patients.

[0033] In a preferred embodiment, the knowledge base module further includes a similar case matching unit. The similar case matching unit is used to match similar cases from the clinical case library according to the extracted individual tumor gene information and embed the matching results into the answer data. For example, this unit can present similar cases matched from the clinical case library according to factors such as the patient's tumor type, tumor stage, tumor grade, gene variation information, and treatment history, etc., to provide reference for users.

[0034] In a specific embodiment, the language processing module is used to generate response data based on the individual tumor gene information, and use a large language model to perform logical reasoning and language polishing on the response data to generate response content. In a specific embodiment, the response content generally includes gene information, targeted drug usage tips, and similar case information. The gene information should include at least one of the association information between the gene and the tumor, the evidence of the association between the gene and the tumor, and the gene mutation analysis information. The targeted drug usage tips include at least one of the relationship between the gene and the sensitivity / resistance of the targeted drug, the evidence level of the relationship between the gene and the sensitivity of the targeted drug, the targeted drug instruction manual, the medical guidelines information of the targeted drug, the clinical trial results of the targeted drug, and the economic data of the targeted drug.

[0035] In a preferred embodiment, the language processing module further includes a language optimization unit. The language optimization unit is used to adjust the expression mode of the response content according to the user identity, where the user identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the language optimization unit converts the professional language in the response content into popular language. For example, after the user instruction recognition module performs semantic analysis on the user instruction and believes that the user has weak tumor knowledge, it determines that the user identity is a non-medical professional (such as a patient). Then, the language optimization unit uses the large language model to perform logical reasoning and language adjustment on the response data to generate a response content in popular language. For example, it converts the professional language expression of "EGFR p.L858R mutation is sensitive to osimertinib (Level A evidence)" into the popular language expression of "Your gene test results show that you are suitable for using osimertinib, and this drug has been recommended by the authoritative guidelines", and provides suggestions suitable for the actual situation of the patient, such as recommending treatment plans with different costs for the patient according to the drug economy level, and making comparisons of efficacy, safety, and ease of purchase for the patient.

[0036] In a specific embodiment, the result feedback module includes an inquiry unit, an information supplement unit, and a content return unit. The content return unit is used to stream the response content back to the user; when the user instruction recognition module determines that the user instruction is a fuzzy semantics, the inquiry unit conducts a guided inquiry on the user and updates the user instruction; when the multi-modal file parsing module determines that the obtained parsing content is an information-deficient file, the information supplement unit conducts a guided information supplement on the user and updates the personal information; after one or more guided inquiries and / or guided information supplements, the language processing module updates the response content according to the updated user instruction and / or the updated personal information, and streams the updated response content back to the user through the content return unit.

[0037] In a preferred embodiment, the method for the result feedback module to provide guided information supplementation to the user includes: the multimodal file parsing module determines whether the knowledge base matching information extracted from the current parsed content meets a preset minimum information set; if not, the result feedback module sends a supplementary information request to the user, and the request includes uploading missing key indicators and / or confirming ambiguous key indicators. Among them, the minimum information set includes at least one key indicator, and the key indicators include tumor type, tumor stage, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, minimal molecular residual status.

[0038] In a preferred embodiment, the result feedback module further includes a prediction unit and an adaptation unit. The prediction unit is used to actively predict and propose a suggestive inquiry based on the user's intention and the parsed content. The prediction unit predicts the user's high-frequency needs and generates a suggestive inquiry based on the statistical data and pre-trained model in the knowledge base, and proposes potential questions or suggestions, so as to obtain a more explicit user instruction and enrich the answer content. For example, if the report shows that a mutation is suitable for a certain type of targeted drug, the system automatically asks "Do you want to know about the clinical trials of the corresponding targeted drug?" The adaptation unit is used to determine the user's identity based on the user's instruction and adjust the expression mode of the guided inquiry and the guided information supplementation, where the user's identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the adaptation unit converts the professional language in the guided inquiry and the guided information supplementation into popular language, so as to adjust the comprehensibility of the inquiry question. For example, if the user's answer is less professional, the system uses a large model to polish a simpler inquiry method, such as "Do you know what stage the tumor is?"

[0039] For example, in a specific embodiment, the user uploads an incomplete gene test report, and the input user instruction is "How to treat cancer?". After semantic analysis of the user instruction by the user instruction recognition module, it is determined that the user instruction is a fuzzy semantics, and after parsing the gene test report by the multimodal file parsing module, it is determined that the key indicators of tumor type and gene mutation type are missing. Therefore, the knowledge base matching information extracted from the current parsed content does not meet the preset minimum information set, and this parsed content is an information-insufficient file. Therefore, the inquiry unit and the information supplementation unit of the result feedback module are triggered to conduct a guided inquiry and guided information supplementation for the user. At the same time, the adaptation unit of the result feedback module determines that the user's identity is a non-medical professional based on the user's instruction, and therefore converts the professional language in the guided inquiry and the guided information supplementation into popular language.

[0040] Therefore, the system first used plain language to guide the user to supplement information: Hello! It is recognized that you have uploaded a gene test report. To better interpret your tumor gene test report, I need some information. I did not extract the information about the tumor type from your report. Please first tell me what type of tumor your report is about. For example, is it lung cancer, breast cancer, or other types?

[0041] The user replied: It is a report about lung cancer.

[0042] The system made a guided information supplement: Thank you for your answer! You mentioned lung cancer. May I ask if the report mentions specific gene mutation information, such as mutations in EGFR, KRAS, or other genes. If so, please tell me the specific mutation type.

[0043] The user replied: I checked and it is an EGFR L858R mutation.

[0044] The system added "Tumor type - Lung cancer" and "Gene mutation - EGFR L858R mutation" to the personal information, which helps to better extract the information matching the knowledge base.

[0045] In addition, the inquiry unit of the result feedback module used plain language to guide the inquiry to the user to obtain a more specific user intention. At the same time, the prediction unit of the result feedback module actively predicted based on the user intention and the parsed content, and embedded the proposed suggested inquiry into the guided inquiry: Thank you for your answer! You mentioned how to treat cancer. Do you need me to focus on recommending drug-related information to you?

[0046] The user replied: Yes.

[0047] The system thus obtained a clear user intention, which helps to better generate the knowledge base matching logic and answer: Thank you for your answer! I will make a detailed report interpretation for you and focus on drug-related recommendations. Please wait a moment.

[0048] According to the updated user instructions and the updated personal information, the corresponding knowledge base matching logic (conventional report interpretation, focusing on drug information), knowledge base matching information (tumor information, gene mutation, drug instructions, pharmacoeconomics, medical guidelines, clinical cases) are obtained. The knowledge base module extracts individual tumor gene information from the knowledge base according to the knowledge base matching logic and knowledge base matching information (including the relationship between the specific mutated gene corresponding to the tumor and cancer, the relationship between the gene and sensitivity / resistance, medication tips, pharmacoeconomics, medication instructions, medical guidelines, and drugs confirmed in clinical case reports) and focuses on drug-related recommendations. The specific information extracted is as follows: 1. Relationship between genes and cancer: The mutation rate of EGFR in non-small cell lung cancer is 49.5%. EGFR is a cancer-promoting gene, and its mutation leads to signal activation, promoting the development of various tumors. The signal activation caused by EGFR gene variation appears in the development process of many tumors, including non-small cell lung cancer, head and neck tumors, esophageal cancer, gastric cancer, colorectal cancer, liver cancer, breast cancer, etc. The EGFRL858R variation is common, which can lead to continuous activation of kinases and enhance the invasion ability of cancer cells. EGFRL858R belongs to a sensitive mutation and is sensitive to a variety of EGFR-TKI drugs.

[0049] 2. Relationship between genes and sensitivity / resistance: For this cancer type, the NMPA / FDA or guidelines have an evidence level of A. In non-small cell lung cancer, the sensitive mutations of the gene EGFR are sensitive to gefitinib, furmonertinib, osimertinib, erlotinib, gefitinib, icotinib, befunolimab, afatinib, bevacizumab + erlotinib, erlotinib + ramucirumab. At the same time, provide the corresponding links to specific clinical trial results, drug brochures, NCCN guidelines, and CSCO guidelines. The evidence of small clinical studies for this cancer type has an evidence level of C. EAI045 is an investigational fourth-generation EGFR allosteric inhibitor that can target EGFR resistance mutations. At the same time, provide specific clinical trial results and relevant clinical trial recruitment information.

[0050] 3. Part of the pharmacoeconomics involved: According to the drugs mentioned above, extract the corresponding pharmacoeconomic information, as shown in Table 1 specifically.

[0051] Table 1. Part of the pharmacoeconomics involved Drug Name Trade Name Manufacturing Pharmaceutical Company Administration Method Specification Price Monthly Average Dosage Gefitinib Tablets Yiruike Qilu Pharmaceutical (Hainan) Co., Ltd. Oral Tablets 0.25g * 10 tablets 420 yuan Monthly average dosage is 3 boxes, total cost is 1260 yuan Gefitinib Tablets Jizhi Zhengda Tianqing Pharmaceutical Group Co., Ltd. Oral Tablets 0.25g * 10 tablets 450 yuan Monthly average dosage is 3 boxes, total cost is 1350 yuan Gefitinib Tablets Iressa AstraZeneca Pharmaceuticals Co., Ltd. Oral Tablets 0.25g * 10 tablets 1596 yuan Monthly average dosage is 3 boxes, total cost is 4788 yuan. Gefitinib Tablets Aixingkang Jiangsu Hengrui Medicine Co., Ltd. Tablets 0.25g * 10 tablets 498 yuan One - month dosage is 3 boxes, total price is 1494 yuan Fumeitinib Mesylate Tablets Aifusha Shanghai Alis Oncology Co., Ltd. Oral Tablets 40mg * 28 tablets 2682.4 yuan Monthly average is 2 - 3 boxes, cost is about 5364.8 - 8047.2 yuan. Osimertinib Mesylate Tablets Tagrisso AstraZeneca Pharmaceuticals Co., Ltd. Oral Tablets 80mg * 30 tablets 4966.2 yuan 1 box per month, total cost is 4966.2 yuan Erlotinib Hydrochloride Tablets Luoruite Shanghai Chuangnuo Pharmaceutical Co., Ltd. Oral Tablets 0.15g * 7 tablets 497 yuan Unknown Erlotinib Hydrochloride Tablets Tarceva Shanghai Roche Pharmaceuticals Co., Ltd. Oral Tablets 150mg * 7 tablets 567 yuan One - month dosage is 4 boxes, total purchase price is 2268 yuan 4. Similar cases: In the first case, the patient was a male with a smoking history, suffering from advanced non-small cell lung cancer and carrying EGFR L858R and E746_A750del mutations. He received targeted therapy with gefitinib, and his symptoms were relieved; the situation of the patient in the second case was slightly more complicated. He carried both HER2 amplification and EGFR L858R mutations. Through the initial treatment with afatinib, his condition had partial remission, and the curative effect lasted for 21 months; in the third case, the patient had different mutations - EGFR G719A and T725M. Although the situation was relatively rare, after treatment with erlotinib, the disease remained stable for 169 days.

[0052] After comprehensively extracting the individual tumor gene information above, the language processing module uses a large language model to perform logical reasoning on the answer data to generate the answer content. And the language optimization unit in it determines that the user's identity is a non-medical professional according to the user's instructions, and converts the professional language in the generated answer content into popular language. The following is the answer content returned to the user: Combined with the drug evidence level and the recommendations of the CSCO guidelines, you can choose osimertinib, amitinib, furmonertinib, afatinib, dacomitinib, gefitinib, erlotinib, and icotinib. Considering different economic conditions and drug efficacy, you can choose a suitable EGFR-TKI for treatment according to your own situation. For patients with better economic conditions, third-generation epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs), such as osimertinib, furmonertinib, and amitinib, are recommended. These drugs show good efficacy in inhibiting EGFR sensitive mutations and T790M drug-resistant mutations and have high safety. Taking osimertinib as an example, the price after medical insurance is about 510 yuan per 80mg tablet. For patients with average economic conditions, first-generation EGFR-TKIs, such as gefitinib, erlotinib, and icotinib, can be considered. These drugs are relatively economical in price and can also effectively inhibit EGFR sensitive mutations. For example, the prices of gefitinib and erlotinib after medical insurance are about 195 yuan per 150mg tablet, and icotinib has been included in the medical insurance scope, which helps to reduce the economic burden on patients.

[0053] The embodiments described above are only used to describe the preferred embodiments of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope of the present invention.

Claims

1. An interactive system for the interpretation scenario of tumor gene detection reports, characterized in that, Including: A user instruction recognition module, which is used to perform semantic analysis on user instructions, judge the user identity and obtain the user intention, and generate a knowledge base matching logic based on the user intention; A multimodal file parsing module, which is used to parse the personal information uploaded by the user and obtain the parsing content, and extract knowledge base matching information based on the parsing content using the deployed large language model; A knowledge base module, which is used to extract individual tumor gene information from the knowledge base according to the knowledge base matching logic and knowledge base matching information; A language processing module, which is used to generate answer data according to the individual tumor gene information, and use the large language model to perform logical reasoning and language polishing on the answer data to generate answer content; A result feedback module, including an inquiry unit, an information supplement unit, and a content return unit; the content return unit is used to return the answer content to the user; when the user instruction recognition module determines that the user instruction is a fuzzy semantics, the inquiry unit conducts a guided inquiry on the user and updates the user instruction; when the multimodal file parsing module determines that the obtained parsing content is an information-insufficient file, the information supplement unit conducts a guided information supplement on the user and updates the personal information; after one or more guided inquiries and / or guided information supplements, the language processing module updates the answer content according to the updated user instruction and / or updated personal information, and streams the updated answer content back to the user through the content return unit.

2. The interactive system according to claim 1, wherein The user instruction recognition module generates the knowledge base matching logic through the following steps: Perform semantic analysis on the obtained user instruction to obtain the user intention; Train a classification model based on the historical data of the user instruction and the medical domain corpus to classify the user intention; Determine the type of knowledge base to be called and the data priority according to the classification result, and generate a knowledge base matching logic corresponding to the user intention.

3. The interactive system according to claim 1, wherein The personal information includes individual medical information and a tumor gene detection report file; The multimodal file parsing module includes: A format conversion unit, which is used to convert the personal information uploaded by the user into structured text; A key index extraction unit, which is used to extract key indexes from the structured text through the large language model and store them in the database as knowledge base matching information; the key indexes include at least one of tumor type, tumor stage, tumor grade, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, and minimal molecular residual status.

4. The interactive system according to claim 1, wherein The knowledge base includes: a clinical case library, a drug database, a gene variation database, a clinical trial database, and a medical literature library; The drug database includes at least one of the relationship between genes and drug sensitivity / resistance, the evidence level of the relationship between genes and drug sensitivity, drug instructions, medical guideline information, drug clinical trial results, and drug economics data.

5. The interactive system according to claim 4, characterized in that, The pharmacoeconomic data includes at least one of FDA approval information, NMPA approval information, and medical insurance coverage status; the medical insurance coverage status includes at least one of the approved drug manufacturer, approved reference price, approved cancer type, and approved target.

6. The interactive system according to claim 4, wherein The knowledge base module further includes: A similar case matching unit, configured to match similar cases from the clinical case base according to the extracted individual tumor gene information, and embed the matching results into the answer data.

7. The interactive system according to claim 1, characterized in that, The language processing module further includes: A language optimization unit, configured to adjust the expression mode of the answer content according to the user identity, where the user identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the language optimization unit converts the professional language in the answer content into popular language.

8. The interactive system according to claim 1 or 6, characterized in that, The answer content includes: gene information, targeted drug usage tips, and similar case information; The gene information includes at least one of the association information between genes and tumors, the evidence of the association between genes and tumors, and gene mutation analysis information; The targeted drug usage tips include at least one of the relationship between genes and targeted drug sensitivity / resistance, the evidence level of the relationship between genes and targeted drug sensitivity, targeted drug instructions, targeted drug medical guidelines information, targeted drug clinical trial results, and targeted drug pharmacoeconomic data.

9. The interactive system according to claim 1, wherein The method for the information supplement unit to conduct guided information supplement for users includes: The multimodal file parsing module determines whether the knowledge base matching information extracted from the current parsed content meets a preset minimum information set; if not, the information supplement unit sends a supplementary information request to the user, and the request includes uploading missing key indicators and / or confirming ambiguous key indicators; The minimum information set includes at least one key indicator, and the key indicators include tumor type, tumor stage, gene information, gene mutation type, tumor mutation burden, programmed death ligand expression status, microsatellite instability status, and minimal molecular residual status.

10. The interactive system according to claim 1, wherein The result feedback module further includes a prediction unit and / or an adaptation unit; The prediction unit is configured to actively predict and propose a suggestive inquiry according to the user intention and the parsed content, and embed the suggestive inquiry into the guided inquiry; The adaptation unit is configured to adjust the expression modes of the guided inquiry and the guided information supplement according to the user identity, where the user identity includes medical professionals or non-medical professionals; when the user is a non-medical professional, the adaptation unit converts the professional language in the guided inquiry and the guided information supplement into popular language.

Images