Medical Data Mining And Analysis Of Operation Anesthesia Information System

The Operation Anesthesia Information System(OAIS),as a subsystem of the hospital information management system,is responsible for accessing the patient’s diagnostic data during the operation,which is of great significance to the treatment and rehabilitation of patients.However,since the main data types of OAIS are diagnostic text and monitoring sequence data and its structure is heterogeneous,which is difficult to process.At present,the data is just collected,stored and viewed,and rarely mined and analyzed.If we can conduct in-depth mining and analysis of these data,it will definitely improve the level of hospital resource management,promote communication between different departments of the hospital,and ultimately improve the quality of medical care and provide better services for public health.Based on the heterogeneous surgical diagnosis data in the OAIS of a third-class hospital in Chongqing,the paper studies data preprocessing,feature extraction,mining model construction,mining platform construction and data fusion analysis,which has good academic research value and practical significance.The main research contents are as follows:(1)For heterogeneous data in OAIS,the advantages and disadvantages of the current feature extraction algorithms for text data and sequence data in surgical data are analyzed,which lays a foundation for subsequent improvement of the algorithm.(2)For the three characteristics("short","missing" and "technical")of text data,the HMM word segmentation algorithm based on THUOCL dictionary and custom dictionary was used.Then,three feature selection algorithms of chi-square statistics(CHI),mutual information(MI)and information gain(IG)were used to select representative words for these words;For the two characteristics("different" and "stable")of sequence data,using moving average algorithm for interpolation processing firstly,and then an improved PAA algorithm was used to feature extraction and representation,and the results can be directly used for subsequent modeling analysis.(3)Combining structured and text features to build a regression mining model.The experimental results show that the fusion feature regression model based on CHI-MLP is best.Its MAE,MSE and R2 are 0.8267,1.2813 and 0.4912 respectively.Case analysis shows that the mining model can better predict the time of surgery length;Combining sequence features to construct a clustering mining model,the experimental results show that the MI-SC-based fusion feature clustering model is best.Its Silhouette Coefficient,Calinski-Harabaz index and Davies-Bouldin index scores are 0.5740,1201 and 1.1890,respectively.The experiment shows that the mining model can predict the latent interdependence between departments.(4)Built a surgical data mining and analysis platform based on Python+Java Script.Platform architecture design,platform function implementation and functional verification are completed,and functions such as exploratory analysis of surgical data,SDOR prediction,clustering,and department relationship discovery are realized.Based on this platform,it is possible to predict the length of the operation duration of different patients and discover the degree of connection between different departments,so as to rationally allocate operating room resources,dynamically adjust the multi-department joint consultation team,improve medical quality,and provide better diagnosis and treatment services.