An intelligent artificial intelligence medical technology value review method and system

Abstract

This invention relates to the field of artificial intelligence-based medical technology value assessment technology, and provides an intelligent AI-based medical technology value assessment method, comprising: S1: analyzing past cases and establishing an expert database; S2: training a Chinese RoBERTa pre-trained model, using health technology assessment reports from past cases as input, and assessment indicators and medical technology completion levels as output; S3: when assessing the value of new medical technologies, using the Chinese RoBERTa pre-trained model and health technology assessment reports to generate assessment indicators and medical technology completion levels for the new medical technologies; S4: obtaining a list of recommended experts from the expert database based on the assessment indicators of the new medical technologies and the type of medical technologies; S5: conducting a simulated assessment of the new medical technologies based on the assessment results of the experts in the recommended expert list on past cases. By relying on past cases to conduct simulated assessments of new medical technologies and proposing preliminary assessment results, this method improves assessment efficiency and saves time and manpower costs.

Description

Technical Field

[0001] This invention relates to the technical field of medical technology value assessment, and in particular to an intelligent artificial intelligence medical technology value assessment method and system. Background Technology

[0002] With the increasingly widespread application of artificial intelligence in the medical field, there is an urgent need for a comprehensive review of the clinical application value of AI medical technologies to support medical and health management decisions. This method was developed based on the research findings of the Clinical Application Guidelines for AI Medical Technologies from the Ministry of Science and Technology-supported study on the evaluation standards for the clinical application of medical AI technologies.

[0003] Valuation assessment of a new artificial intelligence medical technology is a crucial step and method for determining its clinical application value. The assessment of the clinical application value of a medical technology needs to consider multiple factors, including the following:

[0004] (1) Evaluation report quality: the correctness of research design, indicator setting, data quality, and statistical analysis.

[0005] (2) Evaluation report results: clinical needs, clinical use value (safety, effectiveness, efficiency, economy, suitability, accessibility, and innovation) and ethical rationality of the artificial intelligence medical technology to be evaluated.

[0006] When evaluating the clinical application value of artificial intelligence medical technology, a systematic approach and tools are needed to comprehensively consider the above-mentioned factors, as well as the actual situation and the needs and opinions of stakeholders. The expert review panel will evaluate the overall value of the technology in clinical application and make recommendations.

[0007] In existing technologies, the conventional methods for evaluating the value of artificial intelligence in healthcare typically include the following:

[0008] (1) After the health technology assessment team completes the health technology assessment report, a multidisciplinary expert review group (clinical, medical engineering, evidence-based medicine, health economics, etc.) is formed.

[0009] (2) The expert review panel selects appropriate evaluation indicators from a pre-established indicator library containing the correspondence between technology type and evaluation indicators based on the technical characteristics of the medical technology to be evaluated, constructs a multi-dimensional evaluation indicator framework for the medical technology to be evaluated (including multiple dimensions such as technical characteristics, clinical benefits, economy, suitability, and system impact), and determines the weight of the evaluation indicators.

[0010] (3) Each expert representative in the expert review panel scores each evaluation indicator and calculates the overall score based on the health technology assessment report (which includes a list of evidence and results for each evaluation indicator);

[0011] (4) The expert review panel will conduct a thorough discussion based on the health technology assessment report and the comprehensive score results, judge the value of the current medical technology to be evaluated, and finally form an expert opinion on whether to recommend the current medical technology for clinical use.

[0012] The existing methods for evaluating the value of technology require significant manpower and time to invite experts from various disciplines to conduct reviews each time a new medical technology emerges. Furthermore, if a review fails, subsequent improvements to the technology necessitate another review by experts from various disciplines, resulting in low efficiency and substantial waste of manpower and time. Summary of the Invention

[0013] To address the aforementioned problems, the present invention aims to provide an intelligent method and system for evaluating the value of artificial intelligence medical technologies. This system can rely on past cases of medical technology value evaluation to conduct expert simulation reviews of new medical technologies, propose preliminary review results and suggestions, improve review efficiency, and save review time and manpower costs.

[0014] The above-mentioned objective of this invention is achieved through the following technical solutions:

[0015] An intelligent method for evaluating the value of artificial intelligence-based medical technologies includes the following steps:

[0016] S1: Utilize user profiling technology to analyze past cases of medical technology value assessment, create expert review information profiles, and establish an expert database.

[0017] S2: Train the Chinese RoBERTa pre-trained model, which takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output;

[0018] S3: When evaluating the value of a new medical technology, the evaluation indicators and the completion rate of the new medical technology are generated by using the Chinese RoBERTa pre-trained model and the health technology evaluation report corresponding to the new medical technology.

[0019] S4: Based on the evaluation indicators of the new medical technology and the type of medical technology, obtain a list of recommended experts from the expert database;

[0020] S5: Based on the review results of past medical technology value assessment cases by experts in the expert recommendation list, conduct a simulated review of the new medical technology.

[0021] Further, in step S1, the user profiling technology is used to analyze past cases of medical technology value assessment, create expert review information profiles, and establish the expert database, specifically as follows:

[0022] The experts in the expert review panel reviewed past cases of medical technology value assessment.

[0023] Collect information on each expert, including basic information about the expert, information about the expert's academic background, information about the type of medical technology involved in the review, and information about the evaluation indicators in the value review.

[0024] The expert information is concatenated into a vector consisting of the expert's basic information, the expert's academic background information, the medical technology type information, and the weights, value assessment scores, and pass rates of the reviewed evaluation indicators, and then stored in the expert database.

[0025] Further, in step S2, the Chinese RoBERTa pre-trained model is trained, specifically as follows:

[0026] S21: Obtain a large number of health technology assessment reports from past medical technology value review cases, as well as the assessment indicators and medical technology completion rates corresponding to the health technology assessment reports, as training datasets, and clean and preprocess the text of the training datasets.

[0027] S22: Using Chinese word segmentation tools, including jieba, the text of the health technology assessment report is divided into word sequences;

[0028] S23: Sort the words using methods including word frequency or TF-IDF, and select the words related to the evaluation index and the completion of the medical technology to construct a word list;

[0029] S24: Convert the segmented words into an integer sequence in the vocabulary, and add tokens including CLS and SEP to mark the beginning and end of the sequence and sentence separators;

[0030] S25: Randomly mask some of the words in the sequence, and let the Chinese RoBERTa pre-trained model predict the masked words through the context;

[0031] S26: The Chinese RoBERTa pre-trained model uses a Transformer model containing several TransformerEncoder layers as the network structure, and uses a sequence of words that have been randomly masked as input. It is pre-trained on the Transformer model and optimizes the Chinese RoBERTa pre-trained model by maximizing the prediction probability of the masked words and the prediction probability of the next sentence.

[0032] S27: Using the evaluation metrics and medical technology completion data from the downstream training dataset, fine-tune the output data of the Chinese RoBERTa pre-trained model to finally form a natural language model that transforms the health technology assessment report into the evaluation metrics and medical technology completion data of the Chinese RoBERTa pre-trained model.

[0033] Further, in step S4, the expert recommendation list is obtained from the expert database based on the evaluation indicators of the new medical technology and the type of medical technology, specifically as follows:

[0034] S41: Retrieve all expert review profile information related to the evaluation indicators of the new medical technology from the expert database. Based on each expert in the retrieved expert review profile information, multiply the medical technology completion degree by the average score of the current expert's value review score to obtain the current expert's scoring expectation, and sort the experts in descending order according to the scoring expectation.

[0035] S42: Based on the medical technology type of the new medical technology, retrieve the expert database to obtain all experts related to the medical technology type, as well as the weights, value assessment scores, and pass rates of the evaluation indicators related to the medical technology type. Calculate the average weight, average value assessment score, and average pass rate of the different evaluation indicators of the new medical technology by averaging the weights, value assessment scores, and pass rates for each related evaluation indicator. Then, sort the evaluation indicators in descending order according to the average weights.

[0036] S43: Iterate through each of the evaluation indicators related to the new medical technology in sequence, select the expert with the highest expected score corresponding to each evaluation indicator, form an expert group recommendation, and calculate the overall pass rate of the expert group in the form of weighted calculation of the pass rate, and give a set of expert recommendation results, namely, the expert group list and the expected pass rate;

[0037] S44: For each evaluation indicator, select the top three experts as optional experts, forming a total of cubic expert recommendation schemes for the number of evaluation indicators. From the expert recommendation schemes, select the three expert lists with the highest expected pass rate and output them as the expert recommendation list.

[0038] Furthermore, in step S5, based on the review results of past medical technology value assessment cases by experts in the expert recommendation list, a simulated review of the new medical technology is conducted, specifically as follows:

[0039] S51: Based on all experts in the expert recommendation list, calculate the new value assessment score P of the new medical technology for each of the evaluation indicators according to the average weight and the expected score.

[0040] S52: Based on all experts in the expert recommendation list, calculate the new pass rate R of the new medical technology for each of the evaluation indicators according to the average weight and the pass rate;

[0041] S53: When P>(1-R)×100, the new medical technology is deemed to have passed expert review and is recommended for clinical use.

[0042] Further, in step S51, based on all experts in the expert recommendation list, the new value assessment score P of the new medical technology is calculated for each of the evaluation indicators according to the average weight and the expected score, specifically as follows:

[0043] For each of the evaluation indicators, a weighted average is calculated based on the average weight ratio of the expected score for each evaluation indicator, wherein the expected score ranges from 0 to 100.

[0044] Further, in step S52, based on all experts in the expert recommendation list, the new pass rate R of the new medical technology is calculated for each evaluation indicator according to the average weight and the pass rate, specifically as follows:

[0045] For each of the evaluation indicators, a weighted average is calculated based on the proportion of the average weight, wherein the pass rate ranges from 0% to 100%.

[0046] An intelligent AI medical technology value assessment system for performing the intelligent AI medical technology value assessment method as described above includes:

[0047] The expert database creation module is used to analyze past cases of medical technology value assessment using user profiling technology, create expert review information profiles, and build an expert database.

[0048] The RoBERTa model training module is used to train a Chinese RoBERTa pre-trained model. The Chinese RoBERTa pre-trained model takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output.

[0049] The indicator and technology acquisition module is used to generate the evaluation indicators and the completion degree of a new medical technology by using the Chinese RoBERTa pre-trained model and the health technology evaluation report corresponding to the new medical technology when conducting a value assessment of a new medical technology.

[0050] The expert recommendation list acquisition module is used to acquire an expert recommendation list from the expert database based on the evaluation indicators of the new medical technology and the type of medical technology.

[0051] The simulated review module is used to conduct a simulated review of the new medical technology based on the review results of past medical technology value review cases by experts in the expert recommendation list.

[0052] A computer device includes a memory and one or more processors, the memory storing computer code that, when executed by the one or more processors, causes the one or more processors to perform the method described above.

[0053] A computer-readable storage medium storing computer code that, when executed, performs the method described above.

[0054] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0055] (1) An intelligent AI-based method for evaluating the value of medical technologies is provided, comprising: S1: using user profiling technology to analyze past cases of medical technology value evaluation, creating expert review information profiles, and establishing an expert database; S2: training a Chinese RoBERTa pre-training model, wherein the Chinese RoBERTa pre-training model takes health technology assessment reports from past cases of medical technology value evaluation as input, and assessment indicators and medical technology completion as output; S3: when evaluating the value of a new medical technology, the Chinese RoBERTa pre-training model generates the assessment indicators and medical technology completion of the new medical technology using the health technology assessment report corresponding to the new medical technology; S4: obtaining a list of recommended experts from the expert database based on the assessment indicators of the new medical technology and the type of the medical technology; S5: performing a simulated evaluation of the new medical technology based on the evaluation results of the experts in the recommended expert list on past cases of medical technology value evaluation. The above technical solution relies on past cases of medical technology value assessment to conduct expert simulation reviews of new medical technologies, propose preliminary review results, improve review efficiency, and save review time and manpower costs.

[0056] (2) For the proposer of a new medical technology, a mock review can be conducted using the review method of this invention before a formal expert review. The new medical technology can then be continuously improved based on the results of the mock review.

[0057] (3) Generally, new medical technologies must undergo a formal expert review before entering clinical practice. Before the formal review, the review method of this invention is used to conduct a review, provide results and suggestions, and the formal expert review is conducted only after the simulated review of this invention is passed, thereby improving the pass rate of the formal expert review and avoiding the occurrence of multiple formal expert reviews. Attached Figure Description

[0058] Figure 1 This is an overall flowchart of the intelligent artificial intelligence medical technology value assessment method of the present invention;

[0059] Figure 2 This is an overall structural diagram of the intelligent artificial intelligence medical technology value assessment system of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0062] First Embodiment

[0063] like Figure 1 As shown, this embodiment provides an intelligent method for evaluating the value of artificial intelligence medical technology, including the following steps:

[0064] S1: Utilize user profiling technology to analyze past cases of medical technology value assessment, create expert review information profiles, and establish an expert database.

[0065] In this embodiment, it is necessary to conduct expert simulation reviews of a new medical technology. The first step is to analyze the expert review results from past cases of actual expert evaluations of medical technologies to establish an expert database for conducting expert simulation reviews of new medical technologies. The specific method for establishing the expert database is as follows:

[0066] S11: Experts in the expert review panel who have reviewed past cases of medical technology value assessment.

[0067] S12: Collect expert information for each expert, including basic expert information, expert discipline background information, information on the type of medical technology involved in the review, and information related to the evaluation indicators in the value review.

[0068] S13: The expert information is concatenated into a vector consisting of the expert's basic information, the expert's academic background information, the medical technology type information, and the weights, value assessment scores, and pass rates of the reviewed evaluation indicators, and stored in the expert database.

[0069] The aforementioned establishment of an expert database specifically utilizes user profiling technology. User profiling technology refers to collecting and analyzing expert data from expert review panels in past medical technology value assessment cases. This data, combined with basic expert information, disciplinary background information, information on the types of medical technologies reviewed, and information related to assessment indicators in the value assessment, establishes a characteristic model of the expert, forming an expert review information profile. Specifically, the user profiling technology of this invention includes the following steps:

[0070] (1) Data collection: Experts in the expert review group who have reviewed a large number of past medical technology value review cases collect expert information for each expert, including basic expert information, expert discipline background information, medical technology type information, and evaluation indicator information in the value review.

[0071] (2) Data cleaning: The collected expert information is cleaned, deduplicated, and filtered to ensure the quality and accuracy of the expert information data.

[0072] (3) Data analysis: The expert information data is analyzed and processed through data mining, machine learning and other methods to discover the characteristics of the expert information data, including the basic information of the expert, the background information of the expert's discipline, the type of medical technology, as well as the weight of the reviewed evaluation indicators, the value review score, and the pass rate.

[0073] (4) User modeling: Based on the features obtained from the analysis, construct a user profile model and form user feature tags.

[0074] S2: Train the Chinese RoBERTa pre-trained model. The Chinese RoBERTa pre-trained model takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output.

[0075] Specifically, in order to review new medical technologies, it is necessary to find evaluation indicators for reviewing new medical technologies and the degree of completion of medical technologies.

[0076] The evaluation indicators are the standards for reviewing a new medical technology. These indicators may include, but are not limited to, the following:

[0077] Clinical efficacy refers to the therapeutic effect of medical technology on patients' diseases or symptoms in clinical applications, including indicators such as cure rate, effectiveness rate, recurrence rate, and survival rate.

[0078] Safety: refers to the safety of medical technologies during clinical application, including indicators such as adverse reactions, complications, and mortality.

[0079] Economic benefits: refers to the economic performance of medical technology, including indicators such as cost, cost-benefit ratio, and savings in medical resources.

[0080] Quality of life: refers to the impact of medical technology on patients' quality of life, including indicators such as physical function, mental health, and social function.

[0081] Feasibility: refers to the feasibility of medical technology in clinical practice, including indicators such as technical feasibility, operational difficulty, and equipment usability.

[0082] Innovation: refers to the innovativeness of medical technology, including indicators such as technological innovation and treatment method innovation.

[0083] Practicality: refers to the performance of medical technology in terms of practicality, including indicators such as technology promotion and application, and scope of use.

[0084] The above indicators can be adjusted and supplemented according to different medical technologies and assessment requirements, but they basically cover the main aspects of clinical application value assessment. During the assessment process, it is necessary to select appropriate indicators for evaluation based on the specific technology and clinical application, comprehensively consider the weight of each indicator, and draw an assessment conclusion on the application value of the medical technology.

[0085] Completion rate is a crucial indicator for evaluating the application of medical technologies. It refers to the ratio between the actual workload completed and the planned workload during clinical application. Assessing completion rate helps evaluators understand whether the technology achieves its intended effects and whether there are issues such as implementation difficulties or poor results. Furthermore, it provides a reference for subsequent optimization of the technology. Various methods can be used to assess completion rate, such as calculating the ratio by comparing actual completion with the expected workload, or calculating completion rate based on data such as actual operation time and workload. Additionally, other factors, such as implementation difficulty and actual results, can be considered to comprehensively evaluate completion rate. In the clinical application value review of medical technologies, completion rate, as one of the evaluation indicators, provides valuable reference for the actual application of medical technologies and helps evaluators conduct a comprehensive assessment and analysis of the application effects.

[0086] Evaluation metrics and the degree of completion of medical technology were obtained using a pre-trained Chinese RoBERTa model. The training of the Chinese RoBERTa pre-trained model was as follows:

[0087] S21: Obtain a large number of health technology assessment reports from past medical technology value review cases, as well as the assessment indicators and medical technology completion levels corresponding to the health technology assessment reports, as training datasets, and clean and preprocess the text of the training datasets.

[0088] S22: Using Chinese word segmentation tools, including jieba, the text of the health technology assessment report is divided into word sequences.

[0089] S23: Sort the words using methods including word frequency or TF-IDF, and select the words related to the evaluation index and the completion of the medical technology to construct a word list.

[0090] S24: Convert the segmented words into an integer sequence in the vocabulary, and add tokens including CLS and SEP to mark the beginning and end of the sequence and sentence separators.

[0091] S25: Randomly mask some of the words in the sequence, and let the Chinese RoBERTa pre-trained model predict the masked words through the context.

[0092] S26: The Chinese RoBERTa pre-trained model uses a Transformer model containing several TransformerEncoder layers as the network structure, and uses a sequence of words that have been randomly masked as input. It is pre-trained on the Transformer model, and the Chinese RoBERTa pre-trained model is optimized by maximizing the prediction probability of the masked words and the prediction probability of the next sentence.

[0093] S27: Using the evaluation metrics and medical technology completion data from the downstream training dataset, fine-tune the output data of the Chinese RoBERTa pre-trained model to finally form a natural language model that transforms the health technology assessment report into the evaluation metrics and medical technology completion data of the Chinese RoBERTa pre-trained model.

[0094] S3: When evaluating the value of a new medical technology, the evaluation indicators and the completion level of the new medical technology are generated using the Chinese RoBERTa pre-trained model and the corresponding health technology evaluation report.

[0095] S4: Based on the evaluation indicators of the new medical technology and the type of medical technology, obtain a list of recommended experts from the expert database.

[0096] When reviewing new medical technologies, it is necessary to find the most suitable list of experts recommended by the expert database for subsequent review of the new medical technologies. The specific steps include:

[0097] S41: Retrieve all expert review profile information related to the evaluation indicators of the new medical technology from the expert database. Based on each expert in the retrieved expert review profile information, multiply the medical technology completion degree by the average score of the current expert's value review score to obtain the current expert's scoring expectation, and sort the experts in descending order according to the scoring expectation.

[0098] Specifically, the process begins by obtaining expert review profiles related to all evaluation indicators relevant to the new medical technology under review, as outlined in the steps above. Generally, because evaluation indicators encompass multiple disciplines including clinical efficacy, safety, economic benefits, quality of life, feasibility, innovativeness, and practicality, and each expert has their own area of ​​expertise, in most cases, an expert will only review one evaluation indicator. Therefore, the value review score, weight, and pass rate corresponding to one expert record represent only one evaluation indicator. However, it's possible for the same expert to be proficient in multiple disciplines. In such cases, when an expert is involved in reviewing multiple evaluation indicators, we can split the expert's data into multiple separate expert records corresponding to different evaluation indicators to simplify the complexity of the calculation and analysis.

[0099] This step also introduces the concept of medical technology completion, multiplied by the average value assessment score of the expert corresponding to the current evaluation indicator in multiple previous case reviews, to obtain the expert's expected score for the current evaluation indicator. The advantage of this is that formal expert reviews are generally conducted based on a comparison of completion levels. Introducing the concept of completion allows for continuous simulated reviews during the research process of new medical technologies, obtaining simulated review results corresponding to the current progress.

[0100] S42: Based on the medical technology type of the new medical technology, retrieve the expert database to obtain all experts related to the medical technology type, as well as the weights, value assessment scores, and pass rates of the evaluation indicators related to the medical technology type. Calculate the average weight, average value assessment score, and average pass rate of the different evaluation indicators of the new medical technology by averaging the weights, value assessment scores, and pass rates for each related evaluation indicator. Then, sort the evaluation indicators in descending order according to the average weights.

[0101] S43: Following the order calculated in step S42, traverse each of the evaluation indicators related to the new medical technology, select the expert with the highest expected score corresponding to each evaluation indicator, form an expert group recommendation, and calculate the overall pass rate of the expert group in the form of weighted average, and give a set of expert recommendation results, namely the expert group list and the expected pass rate.

[0102] Specifically, in this step, based on the evaluation indicators in descending order, an expert with the highest expected score is found for each evaluation indicator to form an expert group recommendation. The overall pass rate is calculated by weighted averaging the pass rates of the experts with the highest expected scores for each evaluation indicator, based on the weight of each evaluation indicator, thus forming the expected pass rate.

[0103] S44: For each evaluation indicator, select the top three experts as optional experts, forming a total of cubic expert recommendation schemes for the number of evaluation indicators. From the expert recommendation schemes, select the three expert lists with the highest expected pass rate and output them as the expert recommendation list.

[0104] Specifically, in actual simulated reviews, in order to make the review results more accurate, three expert lists are usually selected as expert recommendation lists to review new medical technologies.

[0105] For each evaluation indicator, the top three experts are selected as potential experts. This results in a total of three possible expert recommendation schemes. For these expert recommendation schemes, the expected pass rate is calculated using the same method as in step S43. Finally, the three expert lists with the highest expected pass rates are selected as the output expert recommendation list.

[0106] S5: Based on the review results of past medical technology value assessment cases by experts in the expert recommendation list, a simulated review of the new medical technology is conducted, specifically as follows:

[0107] S51: Based on all experts in the expert recommendation list, calculate the new value assessment score P of the new medical technology for each of the evaluation indicators according to the average weight and the expected score.

[0108] Specifically, based on each evaluation indicator, the average weight of all corresponding experts and their expected scores are first calculated. Then, a weighted average is calculated for the expected scores of each evaluation indicator according to the proportion of the average weight of the evaluation indicators, to obtain the overall new value assessment score P for the new medical technology. The expected score ranges from 0 to 100, and the calculated new value assessment score P will theoretically also be between 0 and 100. However, this scoring method is not unique; this embodiment is only used as an example.

[0109] S52: Based on all experts in the expert recommendation list, calculate the new pass rate R of the new medical technology for each of the evaluation indicators according to the average weight and the pass rate.

[0110] Specifically, for each evaluation indicator, a weighted average is calculated based on the pass rate of each indicator according to its average weight, to obtain the overall new pass rate R of the new medical technology. The pass rate ranges from 0% to 100%, and the calculated new pass rate R will theoretically also be between 0% and 100%. However, this method of calculating the pass rate is not unique; this embodiment is merely an example.

[0111] S53: When P>(1-R)×100, the new medical technology is deemed to have passed expert review and is recommended for clinical use.

[0112] Second Embodiment

[0113] like Figure 2 As shown, this embodiment provides an intelligent artificial intelligence medical technology value assessment system for performing the intelligent artificial intelligence medical technology value assessment method as described in the first embodiment, comprising:

[0114] Module 1, the Expert Database Establishment Module, is used to analyze past cases of medical technology value assessment using user profiling technology, create expert review information profiles, and establish an expert database.

[0115] RoBERTa model training module 2 is used to train a Chinese RoBERTa pre-trained model. The Chinese RoBERTa pre-trained model takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output.

[0116] The indicator and technology acquisition module 3 is used to generate the evaluation indicators and the completion degree of the new medical technology by using the Chinese RoBERTa pre-trained model and the health technology evaluation report corresponding to the new medical technology when conducting a value assessment of a new medical technology.

[0117] The expert recommendation list acquisition module 4 is used to acquire an expert recommendation list from the expert database based on the evaluation indicators of the new medical technology and the type of medical technology.

[0118] The simulated review module 5 is used to conduct a simulated review of the new medical technology based on the review results of past medical technology value review cases by experts in the expert recommendation list.

[0119] A computer-readable storage medium stores computer code that, when executed, performs the methods described above. Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0120] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

[0121] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0122] It should be noted that the above embodiments can be freely combined as needed. The above description is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent method for evaluating the value of artificial intelligence medical technology, characterized in that, Includes the following steps: S1: Using user profiling technology, analyze past cases of medical technology value assessment to create expert review information profiles and establish an expert database; the expert database contains a vector of information including basic expert information, expert academic background information, information on the types of medical technologies reviewed, as well as the weights of the reviewed evaluation indicators, value assessment scores, and pass rates. S2: Train the Chinese RoBERTa pre-trained model, which takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output; S3: When evaluating the value of a new medical technology, the evaluation indicators and the completion rate of the new medical technology are generated by using the Chinese RoBERTa pre-trained model and the health technology evaluation report corresponding to the new medical technology. S4: Based on the evaluation indicators of the new medical technology and the type of medical technology, obtain a list of recommended experts from the expert database, specifically as follows: S41: Retrieve all expert review profile information related to the evaluation indicators of the new medical technology from the expert database. Based on each expert in the retrieved expert review profile information, multiply the medical technology completion degree by the average score of the current expert's value review score to obtain the current expert's scoring expectation, and sort the experts in descending order according to the scoring expectation. S42: Based on the medical technology type of the new medical technology, retrieve the expert database to obtain all experts related to the medical technology type, as well as the weights, value assessment scores, and pass rates of the evaluation indicators related to the medical technology type. Calculate the average weight, average value assessment score, and average pass rate of the different evaluation indicators of the new medical technology by averaging the weights, value assessment scores, and pass rates for each related evaluation indicator. Then, sort the evaluation indicators in descending order according to the average weights. S43: Iterate through each of the evaluation indicators related to the new medical technology in sequence, select the expert with the highest expected score corresponding to each evaluation indicator, form an expert group recommendation, and calculate the overall pass rate of the expert group in the form of weighted calculation of the pass rate, and give a set of expert recommendation results, namely, the expert group list and the expected pass rate; S44: Select the top three experts for each evaluation indicator as optional experts, forming a total of cubic expert recommendation schemes for the number of evaluation indicators. Select the three expert lists with the highest expected pass rate from the expert recommendation schemes and output them as the expert recommendation list. S5: Based on the review results of past medical technology value assessment cases by experts in the expert recommendation list, conduct a simulated review of the new medical technology.

2. The intelligent artificial intelligence medical technology value assessment method according to claim 1, characterized in that, In step S1, the user profiling technology is used to analyze past cases of medical technology value assessment, create expert review information profiles, and establish the expert database. Specifically: The experts in the expert review panel reviewed past cases of medical technology value assessment. Collect information on each expert, including basic information about the expert, information about the expert's academic background, information about the type of medical technology involved in the review, and information about the evaluation indicators in the value review. The expert information is concatenated into a vector consisting of the expert's basic information, the expert's academic background information, the medical technology type information, and the weights, value assessment scores, and pass rates of the reviewed evaluation indicators, and then stored in the expert database.

3. The intelligent artificial intelligence medical technology value assessment method according to claim 1, characterized in that, In step S2, the Chinese RoBERTa pre-trained model is trained, specifically as follows: S21: Obtain a large number of health technology assessment reports from past medical technology value review cases, as well as the assessment indicators and medical technology completion rates corresponding to the health technology assessment reports, as training datasets, and clean and preprocess the text of the training datasets. S22: Using Chinese word segmentation tools, including jieba, the text of the health technology assessment report is divided into word sequences; S23: Sort the words using methods including word frequency or TF-IDF, and select the words related to the evaluation index and the completion of the medical technology to construct a word list; S24: Convert the segmented words into an integer sequence in the vocabulary, and add tokens including CLS and SEP to mark the beginning and end of the sequence and sentence separators; S25: Randomly mask some of the words in the sequence, and let the Chinese RoBERTa pre-trained model predict the masked words through the context; S26: The Chinese RoBERTa pre-trained model uses a Transformer model containing several Transformer Encoder layers as the network structure, and uses a sequence of words that have been randomly masked as input. It is pre-trained on the Transformer model and optimizes the Chinese RoBERTa pre-trained model by maximizing the prediction probability of the masked words and the prediction probability of the next sentence. S27: Using the evaluation metrics and medical technology completion data from the downstream training dataset, fine-tune the output data of the Chinese RoBERTa pre-trained model to finally form a natural language model that transforms the health technology assessment report into the evaluation metrics and medical technology completion data of the Chinese RoBERTa pre-trained model.

4. The intelligent artificial intelligence medical technology value assessment method according to claim 1, characterized in that, In step S5, based on the review results of past medical technology value assessment cases by experts in the expert recommendation list, a simulated review of the new medical technology is conducted, specifically as follows: S51: Based on all experts in the expert recommendation list, calculate the new value assessment score P of the new medical technology for each of the evaluation indicators according to the average weight and the expected score. S52: Based on all experts in the expert recommendation list, calculate the new pass rate R of the new medical technology for each of the evaluation indicators according to the average weight and the pass rate; S53: When When a new medical technology is deemed acceptable by expert review, its clinical use can be recommended.

5. The intelligent artificial intelligence medical technology value assessment method according to claim 1, characterized in that, In step S51, based on all experts in the expert recommendation list, the new value assessment score P of the new medical technology is calculated for each evaluation indicator according to the average weight and the expected score, specifically as follows: For each of the evaluation indicators, a weighted average is calculated based on the average weight ratio of the expected score for each evaluation indicator, wherein the expected score ranges from 0 to 100.

6. The intelligent artificial intelligence medical technology value assessment method according to claim 1, characterized in that, In step S52, based on all experts in the expert recommendation list, the new pass rate R of the new medical technology is calculated for each evaluation indicator according to the average weight and the pass rate, specifically as follows: For each of the evaluation indicators, a weighted average is calculated based on the proportion of the average weight, wherein the pass rate ranges from 0% to 100%.

7. An intelligent artificial intelligence medical technology value assessment system for executing the intelligent artificial intelligence medical technology value assessment method as described in any one of claims 1-6, characterized in that, include: The expert database creation module is used to analyze past cases of medical technology value assessment using user profiling technology, create expert review information profiles, and build an expert database. The RoBERTa model training module is used to train a Chinese RoBERTa pre-trained model. The Chinese RoBERTa pre-trained model takes health technology assessment reports from past medical technology value review cases as input and assessment indicators and medical technology completion as output. The indicator and technology acquisition module is used to generate the evaluation indicators and the completion degree of a new medical technology by using the Chinese RoBERTa pre-trained model and the health technology evaluation report corresponding to the new medical technology when conducting a value assessment of a new medical technology. The expert recommendation list acquisition module is used to acquire an expert recommendation list from the expert database based on the evaluation indicators of the new medical technology and the type of medical technology. The simulated review module is used to conduct a simulated review of the new medical technology based on the review results of past medical technology value review cases by experts in the expert recommendation list.

8. A computer device comprising a memory and one or more processors, the memory storing computer code that, when executed by the one or more processors, causes the one or more processors to perform the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing computer code, wherein when the computer code is executed, the method of any one of claims 1 to 6 is performed.

Images