Medical experiment report AI auxiliary reading system and method based on PTA scale, medium, program product and terminal

Abstract

The invention provides an AI auxiliary reviewing system and method for medical experiment reports based on a PTA scale, a medium, a program product and a terminal, integrates a multi-dimensional evaluation system based on the PTA scale and an artificial intelligence core algorithm, and aims to improve the reviewing efficiency and accuracy of the medical experiment reports. By constructing the data acquisition and preprocessing module, the intelligent analysis and comment module and the comprehensive scoring and comment module, the system realizes deep analysis of professional knowledge details in medical experiment reports, can flexibly process uncertain results in medical experiments, and provides accurate analysis and feedback. According to the invention, medical experiment report marking assistance can be efficiently and accurately provided for teachers, the workload of the teachers is greatly reduced, and the quality and evaluation efficiency of medical education teaching are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and educational information technology, and in particular to an AI-assisted review system, method, medium, program product and terminal for medical laboratory reports based on the PTA scale. Background Art

[0002] For a long time, in the teaching of basic medical experimental courses, the review of experimental reports has mainly relied on teachers to complete it manually. The traditional manual review method not only consumes a lot of time and energy, but also due to the subjective factors of teachers and time constraints, it is difficult to ensure the objectivity, consistency and timeliness of the review. With the gradual popularization of computer technology, some general homework and experimental report review system software have emerged on the market, using keyword retrieval technology to correct errors and conduct preliminary content analysis at the basic level of text grammar, vocabulary application, etc., and some universities have tried to introduce it into the review of medical experimental reports. Students upload the experimental report in Word or PDF document format, and teachers review it through computers, which can quickly locate keywords, copy and paste common comments, etc. However, the functions of this type of system are extremely limited. For the diverse, complex and uncertain experimental results in medical experiments, the software only stays at the level of text surface matching, and cannot intelligently judge whether students have properly applied professional knowledge, and cannot measure the quality of the report from the perspective of the overall experimental process, which is far from the requirements of efficient and accurate review. Summary of the invention

[0003] In view of the above-mentioned shortcomings of the prior art, the present invention provides an AI-assisted review system, method, medium, program product and terminal for medical laboratory reports based on the PTA scale, which are used to solve the problems of low review efficiency and low accuracy of laboratory report review system software for medical laboratory reports in the prior art.

[0004] To achieve the above-mentioned purpose and other related purposes, the first aspect of the present application provides an AI-assisted review system for medical laboratory reports based on the PTA scale, including: a data acquisition and preprocessing module, used to obtain medical laboratory reports to be reviewed and preprocess the medical laboratory reports to be reviewed; an intelligent analysis and comment module, used to process the preprocessed medical laboratory reports to be reviewed based on the review model constructed by the PTA scale method, so as to obtain the PTA scale scoring results and comment results of the medical laboratory reports to be reviewed; a comprehensive scoring and comment module, used to generate a comprehensive score based on the preset scoring rules according to the PTA scale scoring results, and use natural language generation technology to integrate the comment results to generate review comments, so as to obtain the review results of the medical laboratory reports to be reviewed.

[0005] In some embodiments of the first aspect of the present application, the construction process of the review model constructed based on the PTA scale method includes: obtaining a first medical laboratory report sample set and professional knowledge data to construct a medical knowledge base; performing self-supervised pre-training on a basic large language model based on the medical knowledge base to obtain a reward model and an initial review model; performing supervised fine-tuning on the reward model, and performing reinforcement learning on the initial review model based on the supervised fine-tuning reward model to obtain a review model.

[0006] In some embodiments of the first aspect of the present application, the process of supervised fine-tuning the reward model includes: obtaining a second medical laboratory report sample set, manually evaluating each sample based on the evaluation criteria of a preset PTA scale, and reviewing the quality of the manual evaluation results to generate a reward model data annotation sample set; based on the reward model data annotation sample set, supervised fine-tuning training of the reward model to obtain a supervised fine-tuned reward model.

[0007] In some embodiments of the first aspect of the present application, the process of performing reinforcement learning on the initial review model based on the supervised fine-tuning reward model includes: screening the reward model data annotation sample set based on the results of the review quality annotation in combination with the screening conditions, and supplementing the acquired third medical laboratory report sample set to generate a reinforcement learning sample data set; performing reinforcement learning training on the initial review model based on the supervised fine-tuning reward model and the reinforcement learning sample data set to obtain a review model.

[0008] In some embodiments of the first aspect of the present application, the preprocessing method includes: text cleaning, standardization processing and data structured conversion.

[0009] In some embodiments of the first aspect of the present application, the system further includes: a user interaction and result display module, which is used to provide a user interaction page and display the review results.

[0010] To achieve the above-mentioned purpose and other related purposes, the second aspect of the present application provides an AI-assisted review method for medical laboratory reports based on the PTA scale, including: obtaining a medical laboratory report to be reviewed, and preprocessing the medical laboratory report to be reviewed; processing the preprocessed medical laboratory report to be reviewed based on a review model constructed based on the PTA scale method to obtain the PTA scale scoring result and review result of the medical laboratory report to be reviewed; generating a comprehensive score based on the PTA scale scoring result based on preset scoring rules, and using natural language generation technology to integrate the review result to generate review comments to obtain the review result of the medical laboratory report to be reviewed.

[0011] To achieve the above-mentioned purpose and other related purposes, the third aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the AI-assisted review method of medical laboratory reports based on the PTA scale.

[0012] To achieve the above-mentioned purpose and other related purposes, the fourth aspect of the present application provides a computer program product, which includes a computer program code. When the computer program code is run on a computer, the computer implements the AI-assisted review method of medical laboratory reports based on the PTA scale.

[0013] To achieve the above-mentioned purpose and other related purposes, the fifth aspect of the present application provides an electronic terminal, including a memory, a processor and a computer program stored in the memory; the processor executes the computer program to implement the AI-assisted review method of medical laboratory reports based on the PTA scale.

[0014] As described above, the AI-assisted review system, method, medium, program product and terminal for medical laboratory reports based on the PTA scale provided by the present application have the following beneficial effects:

[0015] The present invention combines a multi-dimensional evaluation system based on the PTA scale with an artificial intelligence core algorithm, aiming to improve the efficiency and accuracy of reviewing medical laboratory reports. The present invention constructs a data acquisition and preprocessing module, an intelligent analysis and review module, and a comprehensive scoring and comment module, so that the system can achieve an in-depth analysis of the details of professional knowledge in medical laboratory reports, and can flexibly handle uncertain results in medical experiments, providing accurate analysis and feedback. The present invention can efficiently and accurately provide teachers with assistance in reviewing medical laboratory reports, greatly reducing the workload of teachers and improving the quality and evaluation efficiency of medical education and teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Shown is a structural schematic diagram of an AI-assisted review system for medical laboratory reports based on the PTA scale in one embodiment of the present application.

[0017] Figure 2 Shown is a specific embodiment diagram of the construction process of an evaluation model constructed based on the PTA scale method in one embodiment of the present application.

[0018] Figure 3 Shown is a flow chart of an AI-assisted review method for medical laboratory reports based on the PTA scale in one embodiment of the present application.

[0019] Figure 4 Shown is a schematic diagram of the structure of an electronic terminal in one embodiment of the present application. DETAILED DESCRIPTION

[0020] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0021] Before further describing the present invention in detail, the nouns and terms involved in the embodiments of the present invention are explained. The nouns and terms involved in the embodiments of the present invention are applicable to the following interpretations:

[0022] <1> Large Language Model (LLM): refers to a deep learning model trained with a large amount of text data that can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, and dialogue, and are an important path to artificial intelligence.

[0023] <2> Self-supervised learning: It is a form of unsupervised learning that mines its own supervisory information from large-scale unsupervised data by designing auxiliary tasks (pretexttasks). This method does not require manually annotated label information, but the algorithm automatically constructs supervisory information in large-scale unsupervised data for training. Self-supervised learning usually uses part of the data as a label and trains the model to predict the other part. This method is widely used in many fields such as computer vision, natural language processing, and audio processing. The advantage of self-supervised learning is that it can learn meaningful representations without the need for a large amount of labeled data, thereby reducing the dependence on a large amount of manually annotated data and greatly improving the performance of the model on a variety of tasks.

[0024] <3> Supervised learning: This method uses a large amount of labeled data to train the model. After the model's prediction and the real label of the data generate a loss, back propagation is performed. Through continuous learning, the ability to recognize new samples can eventually be acquired. In supervised learning, each training sample includes input and correct output, which allows the model to learn over time. The algorithm measures its accuracy through the loss function and adjusts until the error is fully minimized. Supervised learning is the most common solution to classification and regression problems. It uses existing training samples (that is, known data and its corresponding output) to train an optimal model, and then uses this model to map all inputs to corresponding outputs, and makes simple judgments on the outputs to achieve the purpose of classification.

[0025] <4> Reinforcement Learning (RL), also known as reinforcement learning, evaluation learning or enhanced learning, is one of the paradigms and methodologies of machine learning. It is used to describe and solve the problem of how an agent can maximize rewards or achieve specific goals through learning strategies during its interaction with the environment.

[0026] <5> The PPO (Proximal Policy Optimization) algorithm is a policy optimization algorithm widely used in the field of reinforcement learning. The core idea of ​​the PPO algorithm is to improve the stability and performance of the algorithm based on the policy gradient algorithm. It avoids training instability caused by excessive policy changes by limiting the amplitude of policy updates.

[0027] <6> SAC (Soft Actor-Critic) is a deep reinforcement learning algorithm based on policy gradients. It has the dual goals of maximizing rewards and maximizing entropy (exploration). SAC introduces an entropy regularization term to make the policy more random when making decisions, thereby improving exploration capabilities. The advantages of this algorithm include: strong exploration capabilities, which helps to cope with complex environments; high training stability, especially in high-dimensional action spaces; and the ability to handle high-dimensional states and high-dimensional action spaces.

[0028] <7> RLHF (Reinforcement Learning from Human Feedback): RLHF is a training technique that combines reinforcement learning algorithms in machine learning with human subjective judgment. By incorporating human feedback into the training process, it provides a natural, humane, interactive learning process for the machine. This approach aims to use human intuition and judgment to guide AI systems to learn more complex and more human-like behavior patterns, especially in areas such as natural language processing (NLP) and decision making.

[0029] <8> Natural Language Processing (NLP): NLP is an interdisciplinary field in computer science, artificial intelligence and linguistics, which aims to enable computers to understand and process human natural language. Natural language is the language used by humans for communication, including spoken and written language, which is highly complex and diverse. The goal of NLP is to enable computers to "understand" human language so as to perform various text-based or voice-based tasks. NLP can be applied to machine translation, public opinion monitoring, automatic summarization, opinion extraction, text classification, question answering, text semantic comparison, speech recognition, Chinese OCR and other aspects.

[0030] To facilitate understanding of the embodiments of the present application, first Figure 1 Detailed description. Figure 1 The schematic diagram of the structure of an AI-assisted review system 100 for medical laboratory reports based on the PTA scale in an embodiment of the present invention is shown. The system structure in this embodiment includes: a data acquisition and preprocessing module 110, an intelligent analysis and review module 120, and a comprehensive scoring and comment module 130; wherein the data acquisition and preprocessing module 110 is connected to the intelligent analysis and review module 120, and the intelligent analysis and review module 120 is connected to the comprehensive scoring and comment module 130.

[0031] In one embodiment, the data acquisition and preprocessing module 110 is used to obtain the medical laboratory reports to be reviewed and preprocess the medical laboratory reports to be reviewed.

[0032] In some examples, the preprocessing method includes: text cleaning, standardization, and data structured conversion.

[0033] Specifically, obtain the medical laboratory reports that are currently waiting for review, and use regular expressions and lexical analysis tools in natural language processing technology to clean and standardize the text of medical laboratory reports, including removing garbled characters, removing interference from special symbols, unifying the expression of medical terms, correcting typos, and unifying the format of data such as dates and units of measurement in different formats. For example, the unified expression of medical terms means that "cesarean section" should be "cesarean section", "hemoglobin" should be "hemoglobin", etc., and common typos such as "myocardial infarction" should be corrected to "myocardial infarction". Preprocessing can effectively improve the accuracy and consistency of the text and provide strong support for subsequent analysis.

[0034] Furthermore, the medical experiment reports are converted into structured data. By parsing the structure of the medical experiment reports, they are divided into structured fields such as experimental purpose, experimental methods, experimental results, result analysis, and conclusions, and stored in XML or JSON format, which makes it convenient for subsequent review models to quickly locate key information blocks and improve processing efficiency.

[0035] In one embodiment, the intelligent analysis and comment module 120 is used to process the pre-processed medical laboratory report to be reviewed based on the review model constructed based on the PTA scale method to obtain the PTA scale score result and comment result of the medical laboratory report to be reviewed.

[0036] It should be noted that the customization and construction process of the PTA scale method is: in conjunction with medical education experts and professional teaching teams, the PTA scale is customized according to the experimental characteristics of each medical course, and the evaluation indicators under the experimental design dimension are subdivided to form a complete evaluation system. The PTA scale method has significant advantages in reviewing medical laboratory reports. By setting specific evaluation indicators for each key element of medical laboratory reports, such as experimental purpose, experimental methods, experimental results and conclusions, and refining the specific scoring rules one by one, the subjective differences that are prone to manual review are greatly reduced, ensuring the objectivity and fairness of the grade assessment, and accurately reflecting the students' real knowledge mastery.

[0037] Furthermore, the professional knowledge and skills involved in medical laboratory reports are more professional and complex, and the dimensions and indicators that need to be evaluated are more refined and specific. For example, in medical laboratory reports, it may be necessary to evaluate the scientific nature of the experimental design, the standardization of the experimental operation, the accuracy of data processing, and the reliability of the experimental results. Non-medical reports may focus more on the basic operating skills of the experiment and the presentation of the results, and the evaluation criteria are relatively simple.

[0038] In view of this, the embodiment of the present invention adds the following evaluation indicators on the basis of the conventional PTA scale: the scientificity of the experimental design, the standardization of the experimental operation, the accuracy of the data processing, and the reliability of the experimental results. In addition, specific evaluation criteria are formulated for each dimension and evaluation indicator. For example, the scientificity of the experimental design can be evaluated from the aspects of the innovation of the research problem and the rationality of the experimental design. The standardization of the experimental operation can be evaluated from the aspects of the proficiency of the operating technology, the control of the experimental conditions, and the integrity of the experimental records. The accuracy of data processing can be evaluated from the aspects of the comprehensiveness of data collection, the rationality of the data processing method, and the presentation of the data results. The reliability of the experimental results can be evaluated from the aspects of the repeatability of the results, the interpretability of the results, and the adaptability of the results. Through these specific evaluation indicators and evaluation criteria, the quality of medical experimental reports can be evaluated more comprehensively and meticulously to ensure the objectivity and fairness of the evaluation.

[0039] In some examples, the construction process of the review model based on the PTA scale method includes: obtaining a first medical laboratory report sample set and professional knowledge data to construct a medical knowledge base; performing self-supervised pre-training on a basic large language model based on the medical knowledge base to obtain a reward model and an initial review model; performing supervised fine-tuning on the reward model, and performing reinforcement learning on the initial review model based on the supervised fine-tuned reward model to obtain a review model.

[0040] It should be noted that the collection of medical laboratory report data: through multiple channels, medical laboratory report samples of students covering different grades, majors, and courses are collected to form the first medical laboratory report sample set to ensure the diversity and comprehensiveness of the data. Among them, the GAN network (Generative Adversarial Networks) is used to generate medical laboratory reports for rare diseases, and the generated medical laboratory reports for rare diseases are also stored in the constructed medical knowledge base. It should be understood that this type of laboratory report focuses on the diagnosis, treatment and research of rare diseases. Due to the large number of rare diseases and the small and scattered number of patients with a single disease, the teaching-related laboratory reports are relatively scarce. In order to make the collected data more comprehensive, the GAN network is used to generate simulated medical laboratory reports for rare diseases.

[0041] At the same time, we also collect professional knowledge data such as relevant medical textbooks, academic literature, and industry standards and specifications. Among them, medical textbooks are basic materials in the medical field. They cover a wide range of content from basic medicine to clinical medicine. The textbooks are usually strictly reviewed and revised by a team of experts to ensure the accuracy and authority of the content. By using medical textbooks as training data, the model can learn the core knowledge and basic concepts in the medical field. Academic literature is the cutting-edge achievement of medical research, containing the latest research results and theoretical progress. The data, experiments, and case studies in academic literature provide the model with rich practical cases, which helps the model better understand and deal with complex medical problems. Industry standards and specifications are operating guidelines and standard procedures in the medical field. They ensure the standardization and safety of medical activities. By incorporating industry standards and specifications into model training, it can ensure that the model follows the correct operating specifications when dealing with medical problems, which helps to improve the reliability and safety of the model in practical applications.

[0042] As an auxiliary knowledge source for model training, professional knowledge data can not only provide rich training materials for the model, but also significantly improve the professionalism and accuracy of the model. After collecting the first medical laboratory report sample set and professional knowledge data, these data also need to be preprocessed. The preprocessing method is text cleaning, standardization processing and data structured conversion, and then a medical knowledge base is constructed.

[0043] In some examples, the basic large language model includes but is not limited to: Bert, LLama, Bloom, internLM, T5, GPT, Qwen2, chatGLM, etc. The basic large language model is self-supervised learning using data from a medical knowledge base, and multiple rounds of iterative training are performed to gradually master the professional knowledge in the medical field, and a reward model and an initial review model can be generated respectively, which can accurately reflect the accuracy and practicality of medical knowledge.

[0044] In some examples, the process of supervised fine-tuning the reward model includes: obtaining a second medical laboratory report sample set, manually evaluating each sample based on the evaluation criteria of a preset PTA scale, and reviewing the quality of the manual evaluation results to generate a reward model data annotation sample set; based on the reward model data annotation sample set, supervised fine-tuning training of the reward model to obtain a supervised fine-tuned reward model.

[0045] It should be noted that obtaining the second medical laboratory report sample set means collecting multiple medical laboratory report samples covering a wide range. Then, each medical laboratory report sample is manually evaluated based on the evaluation criteria of the preset PTA scale, which means that teachers review (scoring and commenting) according to the evaluation indicators in the preset PTA scale. For example, the experimental purpose in the current laboratory report sample is scored and commented on, and the experimental principle in the current laboratory report sample is scored and commented on. Furthermore, the manual evaluation results are reviewed and labeled for quality, specifically, professional review of the teacher's scores and comments, evaluation of the quality of the teacher's scores and comments, and finally generating a reward model data labeling sample set.

[0046] The experimental report, PTA scale and teacher review are used as the input of the reward model, and the review quality annotation is used as the output of the reward model. The reward model is fine-tuned and trained to obtain the reward model after supervised fine-tuning, which can be used to guide the reward signal in the reinforcement learning process. It should be noted that by continuously accumulating high-quality reward model data, using supervised fine-tuning (SFT) and human feedback (RLHF) methods, the stability and scoring consistency of the supervised fine-tuned reward model can be ensured.

[0047] In some examples, the process of performing reinforcement learning on the initial review model based on the supervised fine-tuned reward model includes: screening the reward model data annotation sample set based on the results of the review quality annotation in combination with the screening conditions, and supplementing the acquired third medical laboratory report sample set to generate a reinforcement learning sample data set; performing reinforcement learning training on the initial review model based on the supervised fine-tuned reward model and the reinforcement learning sample data set to obtain a review model.

[0048] It should be noted that the screening condition is preferably: in the sample set screened out from the reward model data annotated sample set, samples with review quality annotations within a preset high score threshold account for a first proportion, and samples with review quality annotations within a preset low score threshold account for a second proportion, and the first proportion is greater than the second proportion. Preferably, samples with review quality annotations within the preset high score threshold account for 80%, and samples with review quality annotations within the preset low score threshold account for 20%. After the samples are screened out, a third medical laboratory report sample set is additionally supplemented. The third medical laboratory report sample set is other medical laboratory report data that is different from the reward model data annotated sample set but has not been annotated with review quality, which can avoid overfitting in subsequent model training. The final sample set obtained is a reinforcement learning sample data set.

[0049] In this embodiment, the reinforcement learning method adopts the PPO or SAC method. Taking the PPO method as an example, the output of the initial review model training process is scored based on the reward model after supervision and fine-tuning, and the iteration direction of the initial review model is guided according to the scoring results of the reward model after supervision and fine-tuning, and the initial review model is continuously optimized to make the output results of the initial review model more in line with the scoring criteria of the PTA scale until the ideal accuracy is reached, and finally a professional review model that can be used to assist in the review of medical laboratory reports is generated. During the reinforcement learning training process, continuous iterative updating of model parameters can ensure that the model performance continues to improve with the accumulation of data.

[0050] It should be noted that the review model in this embodiment can deeply analyze the semantics of medical laboratory reports. It not only understands the surface meaning of the text, but also captures the correlation between medical concepts, such as identifying the causal relationship between abnormal fluctuations in a physiological indicator in the experimental results and operational errors in the experimental method. At the same time, combined with knowledge graph technology, it conducts knowledge reasoning around medical entities (diseases, drugs, physiological processes, etc.) to determine whether the experimental conclusions are in line with the principles of medical science and the logical structure of the knowledge system.

[0051] The review model in the intelligent analysis and review module 120 processes the pre-processed medical laboratory reports to be reviewed and performs specific review tasks. That is, based on the evaluation criteria of the preset PTA scale, the text structure in the medical laboratory report is reviewed and scored. For example, the experimental purpose is predicted based on the corresponding evaluation metrics in the PTA scale to obtain the corresponding score and comment; the experimental principle is predicted based on the corresponding evaluation metrics in the PTA scale to obtain the corresponding score and comment, etc.; the PTA scale scoring result and comment result of the current laboratory report are obtained after the corresponding scores and comments are collected.

[0052] In one embodiment, the comprehensive scoring and comment module is used to generate a comprehensive score based on the PTA scale scoring result based on the preset scoring rules, and integrate the review results using natural language generation technology to generate review comments to obtain the review results of the medical laboratory report to be reviewed. The preset scoring rule is to calculate the comprehensive score based on the weight corresponding to each evaluation indicator in the PTA scale.

[0053] Based on the scores of each evaluation indicator in the PTA scale scoring results predicted by the review model, the comprehensive score of the experimental report is automatically calculated according to the weight rules in the scale. At the same time, natural language generation technology is used, and according to the review results corresponding to each evaluation indicator, reference is made to the built-in review comment template library (customized for different score ranges, common error types, etc.), to generate personalized and targeted review comments for students. The review comments include affirmation of advantages, analysis of shortcomings and improvement suggestions, which help students improve their learning. The comprehensive score and review comments obtained are the review results of the current experimental report.

[0054] It should be noted that there is also a dynamic adaptation mechanism in this embodiment, that is, the system also provides a feedback module to regularly collect teachers' adjustment opinions during the actual marking process and new dynamic information on medical education, so as to dynamically update and adjust the evaluation indicator weights in the PTA scale, add or delete specific evaluation indicator items, and ensure that the PTA scale is always in line with teaching reality, so as to facilitate the continuous improvement of teaching quality.

[0055] In some examples, the system further includes: a user interaction and result display module for providing a user interaction page and displaying the review results. The user interaction and result display module includes teacher-side interaction design, student-side feedback presentation, and result data statistics and visualization.

[0056] Specifically, the teacher-side interactive design creates a simple and intuitive operating interface for teachers, showing a list of medical lab reports to be reviewed, a summary of key information for each lab report, and preliminary scores and comments given by the AI ​​system. Teachers can view detailed analysis reports and manually adjust scores and comments with one click. The system also records teachers' operation records in real time, which can be used for subsequent data analysis and model optimization. The teacher-side interactive design also provides quick functions such as batch review of lab reports and screening of lab reports with specific problems (such as low-scoring lab reports and suspected plagiarism lab reports), which facilitates teachers to efficiently manage review tasks.

[0057] After logging into the system, middle school students can view the AI-assisted review results (review results) of their own lab reports, including detailed scores of various evaluation indicators, detailed comments, and comparisons with examples of similar excellent lab reports, which can promote self-reflection among students and stimulate their own learning motivation and competitive awareness. At the same time, the student side also opens a Q&A interactive area, where students can ask teachers questions about the review results in real time, ensuring smooth communication and enhancing interactivity, forming a virtuous cycle and continuous promotion of mutual learning in teaching, which not only improves students' learning outcomes, but also promotes the continuous improvement and innovation of teachers' teaching methods.

[0058] The result data statistics and visualization refers to the in-depth mining of the review data of laboratory reports from multiple semesters and classes. The statistics include the overall class evaluation and the longitudinal evaluation statistics of individuals from the beginning to the end of the course, and the generation of visual charts, such as the distribution trend chart of laboratory scores for each course, the growth curve of individual student scores, and the pie chart of the proportion of common error types. It can assist teaching management departments and teachers to gain a deeper insight into teaching effectiveness and discover weak links in teaching, thereby providing data support for the adjustment of teaching strategies.

[0059] It should be emphasized that the present invention combines a multi-dimensional evaluation system based on the PTA scale with an artificial intelligence core algorithm, aiming to improve the efficiency and accuracy of reviewing medical laboratory reports. The present invention constructs a data acquisition and preprocessing module, an intelligent analysis and review module, and a comprehensive scoring and comment module, so that the system can achieve an in-depth analysis of the details of professional knowledge in medical laboratory reports, and can flexibly handle uncertain results in medical experiments, providing accurate analysis and feedback. The present invention can efficiently and accurately provide teachers with assistance in reviewing medical laboratory reports, greatly reducing the workload of teachers and improving the quality and evaluation efficiency of medical education and teaching.

[0060] In order to facilitate the description of the construction process of the review model based on the PTA scale method of the present invention, Figure 2 , the following specific examples are provided for illustration.

[0061] Construct a medical knowledge base based on professional knowledge data such as medical laboratory report sample sets and laboratory course textbooks;

[0062] Based on the medical knowledge base combined with the basic large language model (Qwen2), self-supervised learning is performed to generate a reward model (Qwen2-Reward) and an initial review model (Qwen2-CHAT);

[0063] Obtain a reward model data annotation sample set, conduct teacher review based on a preset PTA scale in the sample set, then annotate the teacher review results for review quality, and supervise fine-tune the reward model (Qwen2-Reward) based on the reward model data annotation sample set to obtain a supervised fine-tuned reward model (Qwen2-Reward);

[0064] The reward model data annotation sample set is screened to obtain the screened sample set, and the experimental report sample set (the sample set without review quality annotation) is supplemented to generate a reinforcement learning sample data set. Based on the fine-tuned reward model and reinforcement learning sample data set, the initial review model (Qwen2-CHAT) is reinforced to learn, and the final review model (Fine-tuned Qwen2-CHAT) is obtained.

[0065] The obtained review model (Fine-tuned Qwen2-CHAT) inputs the experimental report to be reviewed, and the AI ​​review result of the experimental report can be output;

[0066] If there are experimental reports with low review quality, the teacher will manually review them and store the experimental report samples until the preset number is stored, and then integrate them into tuning data. The review model (Fine-tuned Qwen2-CHAT) is trained based on the tuning data, and the review model is iteratively updated to obtain a review model that better meets the evaluation quality requirements of the PTA scale.

[0067] In order to facilitate the description of the AI-assisted review system for medical laboratory reports based on the PTA scale of the present invention, the following specific embodiments are provided for illustration.

[0068] Step 1: Data collection and annotation. Obtain no less than 2,000 experimental report assignment samples for each experimental course. The sample sources include reviewed assignment documents of basic medical experimental courses such as human function, morphology, and molecular biology. A detailed and standardized PTA scale annotation guide is formulated by a team of senior teachers and experts to clarify the annotation details, scoring standards, and comment writing specifications for each structure and element. The annotation team carefully annotates the samples according to the guidelines. During the annotation process, a multi-person cross-review mechanism is adopted, and regular quality inspections and calibrations are carried out to ensure the accuracy, consistency, and reliability of the annotated data, and ultimately generate a high-quality pre-training data set.

[0069] Step 2: Review model training. Select the Qwen-2 basic large language model, and determine the model's hyperparameter configuration based on the hardware computing resources (such as GPU model and quantity, memory capacity, etc.) and the scale and characteristics of the pre-trained data set, including key parameters such as the number of neural network layers, number of neurons, learning rate, and batch size. In the supervised fine-tuning (SFT) stage of the supervised learning layer, the labeled pre-trained data set is divided into a training set, a validation set, and a test set. The model is trained for multiple rounds according to the set hyperparameters and training algorithms. After each round of training, the validation set is used to evaluate the model performance. The hyperparameters (such as learning rate decay strategy, regularization coefficient, etc.) are adjusted according to the evaluation results until the model's key indicators such as accuracy and recall rate on the test set reach the preset threshold. In the reinforcement learning process, teachers are used as human feedback providers to build a human-computer interaction environment. Strategy optimization algorithms (such as the PPO algorithm) are used to allow the model to generate review actions (such as scoring, writing comments, etc.) based on the current state. Human feedback providers evaluate the model's actions and give reward signals. The model continuously adjusts its own strategy based on the reward signals. After multiple rounds of iterative training, the stability of the reward model and the consistency of the scoring are ensured, so that the review model can generate high-quality review results that meet professional requirements.

[0070] Step 3: System application and maintenance. The artificial intelligence-assisted experimental report review system of the present invention is fully deployed in the basic medical experimental teaching environment. Students submit experimental reports through the experimental report assignment system module, and teachers use the review system for intelligent review. The system makes intelligent statistical analysis of the points lost by individual students and the class as a whole. Teachers view the analysis results, adjust the teaching focus, optimize the experimental course design, and carry out personalized tutoring and other activities through the background. Students obtain review feedback in a timely manner, review and consolidate basic medical knowledge in a targeted manner, and improve subsequent experimental report writing skills, such as strengthening the understanding and explanation of experimental principles, standardizing the description of experimental steps, and in-depth analysis of experimental results, so as to comprehensively improve their comprehensive quality. The system regularly updates the experimental report samples in the data acquisition module, and retrains the model to adapt to the update of teaching content, changes in experimental requirements, and the evolution of students' writing style, to ensure that the system always maintains efficient and accurate review capabilities.

[0071] In the embodiments of the present application, words such as "first" and "second" are used to distinguish the same or similar items with substantially the same functions and effects, and do not limit their order. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different.

[0072] It should be noted that in the embodiments of the present application, words such as "exemplary" or "for example" represent examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0073] In the embodiments of the present application, "at least one" refers to one or more, and "plurality" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc or abc, where a, b, c can be single or multiple.

[0074] Figure 3 is a schematic block diagram of an AI-assisted review method for medical laboratory reports based on the PTA scale provided in an embodiment of the present application. Figure 3 As shown, the method includes:

[0075] Step S301: obtaining a medical laboratory report to be reviewed, and preprocessing the medical laboratory report to be reviewed;

[0076] Step S302: The review model constructed based on the PTA scale method processes the pre-processed medical laboratory report to be reviewed to obtain the PTA scale score result and comment result of the medical laboratory report to be reviewed;

[0077] Step S303: Generate a comprehensive score based on the PTA scale scoring result based on the preset scoring rules, and use natural language generation technology to integrate the review results to generate review comments to obtain the review results of the medical laboratory report to be reviewed.

[0078] It should be noted that the AI-assisted review method for medical laboratory reports based on the PTA scale provided in this application can be specifically applied to controllers such as ARM (Advanced RISC Machines), FPGA (Field Programmable Gate Array), SoC (System on Chip), DSP (Digital Signal Processing), or MCU (Microcontroller Unit); or can be applied to computers including components such as memory, storage controller, one or more processing units (CPU), peripheral interface, RF circuit, audio circuit, speaker, microphone, input / output (I / O) subsystem, display, other output or control devices, and external ports; the computer includes but is not limited to personal computers such as desktop computers, laptops, tablet computers, smart phones, smart TVs, personal digital assistants (PDA for short); it can also be applied to servers, which can be arranged on one or more physical servers according to various factors such as function and load, or can be composed of distributed or centralized server clusters.

[0079] It should be understood that the specific process of executing the above corresponding steps has been described in detail in the above embodiments, and for the sake of brevity, it will not be repeated here.

[0080] It should also be understood that the division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present application may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0081] Figure 4 is a schematic block diagram of an electronic terminal provided in an embodiment of the present application. Figure 4 As shown, the electronic terminal includes: at least one processor 401, a memory 402, at least one network interface 403 and a user interface 405. The various components in the device are coupled together through a bus system 404. It can be understood that the bus system 404 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 404 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 4 In the specification, various buses are labeled as bus systems.

[0082] The user interface 405 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0083] It is understood that the memory 402 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM). The memory described in the embodiments of the present invention is intended to include but is not limited to these and any other suitable categories of memory.

[0084] The memory 402 in the embodiment of the present invention is used to store various categories of data to support the operation of the electronic terminal 400. Examples of these data include: any executable program for operating on the electronic terminal 400, such as an operating system 4021 and an application 4022; the operating system 4021 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 4022 may include various applications, such as a media player (Media Player), a browser (Browser), etc., for implementing various application services. The AI-assisted review method for medical laboratory reports based on the PTA scale provided in the embodiment of the present invention may be included in the application 4022.

[0085] The method disclosed in the above embodiment of the present invention can be applied to the processor 401, or implemented by the processor 401. The processor 401 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 401 or the instruction in the form of software. The above processor 401 may be a general processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 401 can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general processor 401 may be a microprocessor or any conventional processor, etc. In combination with the steps of the accessory optimization method provided in the embodiment of the present invention, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium, which is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0086] In an exemplary embodiment, the electronic terminal 400 may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD) to execute the aforementioned method.

[0087] According to the method provided in the embodiments of the present application, the present application also provides a computer program product, which includes: computer program code, when the computer program code is run on a computer, the computer executes the AI-assisted review method of medical laboratory reports based on the PTA scale in any of the embodiments shown.

[0088] According to the method provided in the embodiments of the present application, the present application also provides a computer-readable storage medium, which stores a program code. When the program code runs on a computer, the computer executes the AI-assisted review method of medical laboratory reports based on the PTA scale in any of the embodiments shown.

[0089] The terms "component", "module", "system", etc. used in this specification are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program and / or a computer. By way of illustration, both applications running on a computing device and a computing device can be components. One or more components may reside in a process and / or an execution thread, and a component may be located on a computer and / or distributed between two or more computers. In addition, these components may be executed from various computer-readable media having various data structures stored thereon. Components may, for example, communicate through local and / or remote processes according to signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system and / or a network, such as the Internet interacting with other systems through signals).

[0090] Those of ordinary skill in the art will appreciate that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0092] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0094] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0095] In the above embodiments, the functions of each functional unit can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When loading and executing computer program instructions (programs) on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media integrations. Available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., high-density digital video discs (DVDs), or semiconductor media (e.g., solid state disks (SSDs), etc.).

[0096] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0097] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0098] In summary, the present application provides an AI-assisted review system, method, medium, program product and terminal for medical laboratory reports based on the PTA scale, including: a data acquisition and preprocessing module, used to obtain medical laboratory reports to be reviewed and preprocess the medical laboratory reports to be reviewed; an intelligent analysis and comment module, used to process the preprocessed medical laboratory reports to be reviewed based on the review model constructed based on the PTA scale method, so as to obtain the PTA scale scoring results and comment results of the medical laboratory reports to be reviewed; a comprehensive scoring and comment module, used to generate a comprehensive score based on the preset scoring rules according to the PTA scale scoring results, and use natural language generation technology to integrate the comment results to generate review comments, so as to obtain the review results of the medical laboratory reports to be reviewed.

[0099] The present invention integrates a multi-dimensional evaluation system based on the PTA scale and an artificial intelligence core algorithm, aiming to improve the review efficiency and accuracy of medical laboratory reports. The present invention constructs a data acquisition and preprocessing module, an intelligent analysis and review module, and a comprehensive scoring and comment module, so that the system can achieve an in-depth analysis of the details of professional knowledge in medical laboratory reports, and can flexibly handle uncertain results in medical experiments, and provide accurate analysis and feedback. The present invention can efficiently and accurately provide teachers with medical laboratory report review assistance, greatly reducing the workload of teachers and improving the quality and evaluation efficiency of medical education and teaching. Therefore, the present application effectively overcomes the various shortcomings in the prior art and has a high industrial utilization value.

[0100] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.

Claims

1. An AI-assisted review system for medical laboratory reports based on the PTA scale, characterized in that: include: A data acquisition and preprocessing module, used to obtain medical laboratory reports to be reviewed and preprocess the medical laboratory reports to be reviewed; An intelligent analysis and comment module, which is used to process the pre-processed medical laboratory reports to be reviewed based on the review model constructed by the PTA scale method, so as to obtain the PTA scale score results and comment results of the medical laboratory reports to be reviewed; The comprehensive scoring and comment module is used to generate a comprehensive score based on the PTA scale scoring results based on preset scoring rules, and use natural language generation technology to integrate the review results to generate review comments to obtain the review results of the medical laboratory report to be reviewed.

2. The AI-assisted review system for medical laboratory reports based on the PTA scale according to claim 1 is characterized in that: The construction process of the review model based on the PTA scale method includes: Obtain the first medical laboratory report sample set and professional knowledge data to build a medical knowledge base; Performing self-supervised pre-training on a basic large language model based on the medical knowledge base to obtain a reward model and an initial review model; The reward model is fine-tuned through supervision, and reinforcement learning is performed on the initial review model based on the reward model after the supervision fine-tuning to obtain a review model.

3. The AI-assisted review system for medical laboratory reports based on the PTA scale according to claim 2 is characterized in that: The process of supervised fine-tuning the reward model includes: Obtain a sample set of laboratory reports of the second medical category, manually evaluate each sample based on the evaluation criteria of the preset PTA scale, and review and annotate the quality of the manual evaluation results to generate a sample set of data annotations for the reward model; The reward model is supervised fine-tuned based on the reward model data labeled sample set to obtain a supervised fine-tuned reward model.

4. The AI-assisted review system for medical laboratory reports based on the PTA scale according to claim 3 is characterized in that: The process of performing reinforcement learning on the initial review model based on the supervised fine-tuned reward model includes: Based on the result of the review quality annotation and the screening condition, the reward model data annotation sample set is screened, and the acquired third medical laboratory report sample set is supplemented to generate a reinforcement learning sample data set; The initial review model is trained by reinforcement learning based on the supervised fine-tuned reward model and the reinforcement learning sample dataset to obtain the review model.

5. The AI-assisted review system for medical laboratory reports based on the PTA scale according to claim 1 is characterized in that: The preprocessing method includes: text cleaning, standardization processing and data structured conversion.

6. The AI-assisted review system for medical laboratory reports based on the PTA scale according to claim 1 is characterized in that: The system also includes: a user interaction and result display module, which is used to provide a user interaction page and display the review results.

7. An AI-assisted review method for medical laboratory reports based on the PTA scale, characterized in that: An AI-assisted review system for medical laboratory reports based on the PTA scale as described in any one of claims 1 to 6, the method comprising: Obtaining medical laboratory reports to be reviewed, and preprocessing the medical laboratory reports to be reviewed; The review model constructed based on the PTA scale method processes the pre-processed medical laboratory report to be reviewed to obtain the PTA scale score result and comment result of the medical laboratory report to be reviewed; A comprehensive score is generated based on the PTA scale scoring results and preset scoring rules, and the review results are integrated with natural language generation technology to generate review comments to obtain the review results of the medical laboratory report to be reviewed.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the AI-assisted review method of medical laboratory reports based on the PTA scale as described in claim 7 is implemented.

9. A computer program product, characterized in that The computer program product includes computer program code, and when the computer program code is executed on a computer, the computer implements the AI-assisted review method for medical laboratory reports based on the PTA scale as described in claim 7.

10. An electronic terminal comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the AI-assisted review method for medical laboratory reports based on the PTA scale as described in claim 7.

Images