Research And Development Of Artificial Intelligence Assisted Diagnosis And Treatment System For Coronary Heart Disease

BackgroundCoronary heart disease(CHD)is a major global public health problem and a major risk factor for death,occupying the first place as a single cause of death in some countries.The prevalence and mortality of CHD in China has been on the rise in recent years.A search of the Web of Science database shows that CHD has been a hot research area worldwide for the past 30 years.Chinese medicine is a treasure of Chinese civilization and has played an important role in the prevention and treatment of common diseases,multi-morbidity and major epidemics.CHD is classified as "chest paralysis" and "true heart pain" in traditional Chinese medicine(TCM),and Chinese medicine has been practised for thousands of years in the prevention and treatment of CHD,and has certain treatment characteristics and advantages.The level of treatment of famous Chinese medicine practitioners nationwide represents the first echelon of Chinese medicine,and the treatment of CHD by famous Chinese medicine practitioners has remarkable efficacy.However,due to the limitations of Chinese medicine,the promotion and popularisation of the treatment advantages of famous TCM practitioners is faced with problems such as long training period for TCM personnel and difficulties in passing on the experience of famous TCM practitioners,which make it difficult to meet the increasing demand of patients for treatment by famous TCM practitioners.Therefore,there is an urgent need for new technologies to simulate,replicate and promote the treatment of CHD by famous TCM doctors,so as to assist the training of TCM talents and the transmission of the experience of famous TCM doctors,and benefit the majority of patients.Artificial intelligence(AI)provides a pathway for profound change in the field of medicine.The research will be an effective solution to these problems by simulating the clinical treatment of CHD by famous TCM doctors through new AI technologies and realising the AI treatment for CHD in TCM.ObjectiveTo address the above problems,this study explores the method of simulating the treatment of CHD by Professor Wang Jie,a famous Chinese medicine practitioner,through new AI technology,and forms a Chinese medicine AI-assisted diagnosis and treatment system for CHD,to scientifically evaluate the intelligence and Chinese medicine thinking of the diagnosis and treatment system,and to demonstrate the molecular composition and target regulation mechanism of Chinese medicine prescribed by the diagnosis and treatment system.The system will provide an effective solution for the TCM AI diagnosis and treatment of CHD,and provide a technical pathway for the batch replication and promotion of the treatment model of famous TCM doctors,the inheritance of famous TCM doctors’ treatment experience and the cultivation of TCM talents.Methods1.Clarify research proposals based on literature researchThrough qualitative and quantitative analysis of the literature on the intersection of AI and medicine research,we obtain its research lineage,hotspots and prospective directions,and sort out the main applications and algorithms of AI in the medical field,etc.Through a systematic analysis of the intersection of Chinese medicine and AI research,an in-depth analysis of the reasons limiting the deep integration of Chinese medicine and AI is conducted.The above is combined to clarify the research proposal for this study.2.Dismantling functional modules based on the research proposalFrom the perspective of top-level design,the AI-assisted diagnosis and treatment system for CHD is disassembled by functional modules to carry out research in modules and refine the technical route.3.Data set constructionCollect clinical cases of CHD treated by Professor Wang Jie,ancient and modern famous doctors,Chinese medicine treatment guidelines and classical Chinese medicine prescriptions,etc.,and construct a basic Chinese medicine dataset for CHD after data pre-processing and data mining.4.Natural language processing module Follow up the research progress of natural language processing,select the natural language processing models suitable for the treatment of CHD in TCM,and carry out model screening and evaluation.5.Artificial intelligence diagnosis and treatment moduleBased on the results of the literature review,we use knowledge mapping,expert augmented machine learning and deep learning methods to build and train AI diagnosis and treatment models based on the underlying data set and conduct testing and evaluation.6.Assistive function moduleSimulating the real treatment mode and process of TCM experts,we design and develop auxiliary functions for AI treatment of CHD in TCM,in order to integrate the functions and models built in this research,and form a convenient and easy-to-use AI auxiliary treatment system for CHD in TCM.7.Turing TestThe AI Turing test was designed for the AI aided diagnosis and treatment system for CHD in TCM.The prescriptions of Professor Wang Jie’s medical case and the prescriptions issued by the AI-aided treatment system according to the corresponding medical case were randomly blinded,and clinical experts in the field of TCM cardiovascular medicine were invited to identify the above two prescriptions,and the incorrect classification rate of the clinical experts was counted.Results1.Based on the construction method of clinical assisted decision-making system,AI technologies such as artificial neural network,natural language processing,machine learning,deep learning and knowledge mapping are selected as the technical means to study the AI diagnosis and treatment system for Chinese medicine,taking into account the characteristics of Chinese medicine.The research plan of the TCM AI-assisted diagnosis and treatment system for CHD is formed by simulating the mode and method of treatment of CHD by Professor Wang Jie,a famous Chinese medicine practitioner,through AI.2.Based on the top-level design and functional architecture of the AI diagnosis and treatment system,the dismantling plan of the sub-module research is determined,and the specific implementation of the five main modules of the CHD Chinese medicine AI-assisted diagnosis and treatment system is refined as follows:Module 1 is the symptom and sign acquisition module.Module 1 is the symptom and sign acquisition module,which implements speech recognition and intelligent processing of patient’s oral symptoms and signs.Module 2 is the Al diagnosis and treatment model module.The module uses knowledge mapping,expert augmented machine learning and deep learning methods to build an AI diagnosis and treatment model.Module 3 is a module for intelligent review of AI prescriptions,manual revision by physicians and analysis of prescription mechanisms.Module 4 is the system database module,which supports the management and maintenance of data and provides support for model calculation and data presentation.Module 5 is the front-end display module,which realizes the interaction with the physician’s operation interface,AI treatment model,data analysis and display interface and the back-end database.3.The basic data set of CHD was constructed.582 clinical cases of CHD treated by Professor Wang Jie,2055 cases of CHD treated by ancient and modern famous doctors,10,140 Chinese medicine treatment guidelines and classical Chinese medicine prescriptions for CHD are collected as data sources to build the basic data set.The data set is used to support the construction and training of natural language processing models and AI treatment models.4.A pre-training module for the natural language processing model of TCM for CHD was constructed using a pre-training large model similar to ChatGPT.The optimal model was selected as a natural language processing model with Bert-base-Chinese as the pre-training model and a fully connected neural network model with 768dimensional input layer,32-dimensional middle layer and 109-dimensional output layer as the fine-tuned model.The accuracy of the optimal model was tested to be 0.9373.5.For the evaluation of the three models constructed,the accuracy of the AI treatment model based on knowledge graph was 71.17%for the main prescription and 90%for the additive and subtractive prescriptions;the prescription of the AI treatment model based on expert augmented machine learning was 73.65%in accordance with Professor Wang Jie’s prescription;the prescription of the AI treatment model based on deep learning was 84.85%in accordance with Professor Wang Jie’s prescription.6.The design and development of the prescription recommendation index function,medication addition and reduction function,AI prescription dose initial assignment function,prescription intelligent review function,intelligent prescription regulation of disease molecular and target mechanism analysis function and knowledge map visualization display function.7.A total of five clinical experts in the field of TCM cardiovascular medicine were invited as interviewed experts,and the misclassification rates of the five experts in identifying the two prescriptions were 55%,60%,55%,60%and 55%respectively,all of which exceeded 50%and were close to random guesses.The Turing test was passed by the preliminary assessment of the AI-aided diagnosis and treatment system for CHD in TCM.Conclusions1.Based on the clinical treatment cases and experience of famous Chinese medicine practitioners,data pre-processing and data mining methods are used to build an autonomous dataset for CHD in Chinese medicine,and a natural language processing model for CHD in Chinese medicine is built by combining pre-training models and training fine-tuning models based on the autonomous dataset,which can effectively realize the recognition and processing of patients’ oral symptoms by the treatment system.2.Based on the basic dataset for CHD in TCM,the AI diagnosis and treatment models were constructed by using knowledge graph,expert augmented machine learning and deep learning methods to realize the function of outputting TCM prescriptions by the system after voice input or input of symptoms and signs to the diagnosis and treatment system.All three AI models have good results.3.The AI-assisted diagnosis and treatment system for CHD in TCM has passed the AI Turing test after preliminary evaluation.The process of distinguishing machine intelligent prescription from Professor Wang Jie’s prescription by clinical experts in the field of TCM cardiovascular is close to a random guessing process.The AI diagnosis and treatment system has a certain "intelligence",that is,it has a certain TCM thinking.4.After the integration of functional modules and auxiliary functions,the AI assisted diagnosis and treatment system for CHD in TCM can provide intelligent and convenient clinical assisted decision support for the clinical diagnosis and treatment of CHD and assist in the transmission of experience of famous doctors and the cultivation of clinical thinking of doctors.The method of constructing the AI-assisted diagnosis and treatment system for CHD in TCM in this study was initially validated to be feasible and scientifically sound.It provides a reference for other TCM experts to study AI-based diagnosis and treatment systems based on new AI technologies.