Research On Personalized Physical Examination Item Recommendation Method Based On Physical Examination Data

Physical examination is a medical service that provides medical examination and evaluation for individuals and groups without obvious symptoms,aiming to screen out sub healthy populations at risk of chronic non communicable diseases,thereby achieving disease prevention,early diagnosis and treatment.However,due to the wide variety of chronic diseases and atypical early symptoms,it is difficult to fully cover all chronic diseases and their risk factors in physical examination.Currently,physical examination packages are often used to address this issue in physical examination practice.The physical examination institution classifies the subjects based on basic characteristics such as age and gender,and conducts physical examinations based on the problems that each type of subject is more likely to develop diseases.However,there are significant differences in personal health,life history,and family history among subjects of the same age and gender,and simple feature classification alone cannot provide accurate physical examination project recommendations.In recent years,with the development of medical informatization,physical examination data has been continuously accumulated.A series of studies based on historical physical examination data have emerged to evaluate the health status of examinees,predict chronic disease risks,and provide personalized physical examination recommendations.However,due to the temporal nature of physical examination data,previous studies have overlooked the impact of sequence patterns on disease risk.In response to this issue,this study will start from physical examination data,and based on this,study the time series characteristics of various test indicators and diseases,excavate the frequent patterns between changes in physical examination items and chronic diseases.Before the examination,match the historical data of the subject with the disease risk pattern,in order to predict the possible disease risk of the subject and recommend physical examination items accordingly.The specific research content of this paper is as follows:(1)Conduct feature analysis on the abnormal data characteristics of diseased patients,and find that high-risk patients have highly similar sequence patterns,and use the data corresponding to these features as attributes;Standardize physical examination data,select minimum support(min＿sup=0.05),serialize data,and conduct time series mining research.Taking diabetes as a research case,we used Prefixspan algorithm to mine the time series of the processed physical examination data,introduced relevant risks to filter the risk pattern,and confirmed 9 optimal risk pattern.Finally,relevant literature was searched on Pub Med for evaluation,and the results showed that the optimal risk model was reasonable.(2)By reading literature and using crawler technology to obtain relevant knowledge about examination items related to diseases,and then using medical definitions and expert experience to fuse the acquired knowledge,a knowledge base of examination items for six types of diseases within the risk model is constructed.Using sequence matching algorithm,by matching the risk pattern with the user’s historical physical examination data,recommend physical examination items for diseases within the risk pattern for the successfully matched subjects,and complete the recommendation of physical examination items.(3)Based on the above personalized physical examination recommendation methods,we designed and developed a personalized physical examination package recommendation system,integrated data with the hospital information system,obtained the historical physical examination data of the examinee,recommended physical examination items using the historical physical examination data,and provided functions such as physical examination appointment of the examinee,personalized physical examination package recommendation,and physical examination report viewing.Finally,the performance of the system was evaluated through case analysis,and the recommended accuracy of the system was 0.84.The results showed that this system has certain feasibility.