The Research And Implementation Of Intelligent Diagnosis And Severity Assessment Algorithm For Dementia

Dementia is a kind of Neurological syndrome caused by brain disease,which has the characteristic of chronic and progressive and will affect the patient’s higher cortical functions,such as memory,thinking,orientation,comprehension,calculation,learning ability,language and judgment etc.At present,about half of the dementia cases in our country are thought as natural aging phenomenon by mistake,and only less than 20% of the patients can get the corresponding diagnosis and treatment.With the development of smart medical care,the artificial intelligence can be used to help doctors in decision-making and diagnosis,greatly improve the work efficiency of doctors,and has great significance for the timely prevention and treatment of dementia.Just because of that,in this thesis,an algorithm based on ensemble and incremental learning,which can classify the dementia’s stage and severity,is proposed,a telemedicine diagnosis server system,which integrates the algorithm,is built,it would greatly help the doctors to diagnose the dementia successfully.The content of this thesis is as follows:(1)Since traditional classifiers have different classification effects on different medical record features,this thesis uses the XGBoost,LGBM,and GBDT algorithms as the base model for feature extraction based on the idea of Stacking algorithm.The GBRT algorithm is used as the meta-model to classify and output,and the parameters of the algorithm are adjusted.Compared with the traditional algorithm,the improved algorithm has improved various indicators.(2)Based on the classification of dementia stage,this thesis classifies the severity of dementia.In response to the problem of poor classification of mild dementia and questionable dementia in the current literature,this thesis improves the traditional one-dimensional convolutional neural network.Use SVM to replace the Softmax layer of the original network,and combines the improved network with the Stacking algorithm is fused,and finally adjust the parameters of the algorithm.The improved algorithm can effectively solve the problem of poor classification of mild dementia and questionable dementia.Compared with the unimproved algorithm model,the accuracy is also improved to a certain extent.(3)In order to improve the accuracy of real medical record data prediction,this thesis proposes an incremental learning-based ILSCE algorithm.This thesis collects and processes160 real hospital patient medical records,and on the basis of the first two algorithms,incremental learning is performed on the training set in the hospital real medical records.This algorithm can retain the characteristics of the old data and at the same time learn the data characteristics of the real medical records.Compared with the model without incremental learning,the algorithm based on incremental learning has a certain improvement in the prediction of real medical record data.(4)For better application of the above algorithm,this thesis builds a telemedicine diagnosis server system to collect information such as patients’ medical history,symptoms,and medical examination results.The above algorithm is integrated into the system,and the algorithm is used to diagnose and classify the severity of dementia in patients,and finally achieve the purpose of remote diagnosis and treatment.