Research On Cognitive Assessment Method Of Alzheimer’s Disease Based On Machine Learning

Nowadays,with the deepening of the aging of the population,the number of people with Alzheimer’s disease is increasing rapidly,that is,the situation is becoming more and more serious.It not only brings great physical and mental pain to the patients themselves,but also has many impacts on their families and society.Dementia has become the seventh leading cause of death in the world.Unfortunately,there is no cure for AD yet,but early screening,early diagnosis and early intervention are of great significance to delay the progression of the disease,especially intervening individuals in the stage of Mild cognitive impairment(MCI)before the onset of AD,and even preventing the occurrence of the disease.However,it is estimated that less than 25%of people with dementia are diagnosed worldwide,and in some low-and middle-income countries,this proportion may even be less than 10%,that is,the vast majority of people with dementia go undetected,let alone receive timely intervention.In light of this,the World Health Organization is calling for increased screening rates for Alzheimer’s disease.According to the investigation,the existing AD diagnosis researches have:1)Most of them are based on data that are not easy to obtain and high cost,such as MRI,PET,genetic data,etc.;2)Most of the studies has been done for binary classification of AD,while multi-class classification has greater significance;3)Most of the models cannot explain the reasons behind their decisions,which makes it difficult for them to gain the trust of physicians,etc.As a result,existing AD classification models are not suitable for AD screening.In addition,it was also learned that cognitive scores play a major role in the diagnosis of Alzheimer’s disease and its effect is intuitive,easy to collect and low cost.In view of the above situation,the main work and innovations of the paper include:1.A solution for small sample size&multi-class&categoryimbalanced data sets is proposed-RN-SSASThis method can reliably evaluate the generalization ability of the model under extremely small sample size&multi-class&classimbalanced data dilemma.Futhermore,when the dataset situation is not very bad,in addition to addressing the class imbalance problem,1)Compared to Train/Test split,it can learn patterns in the data to a greater extent,thus achieving higher performance;2)Compared to CV,it has better robustness.2.Explore a convenient,reliable and interpretable screening method for Alzheimer’s disease,and establish a three-class classification(CN vs.MCI vs.AD)model for Alzheimer’s disease,an intelligent AD screening system is constructed accordinglyIn order to improve the screening rate of AD,based on the above solution and multiple cognitive scores,this paper establishes a three-class classification(CN vs.MCI vs.AD)model of AD,and provides single instance and global instance explanations based on Shapley values for each class of the model,so as to form an interpretable AD intelligent screening system,so that physicians can communicate with patients transparently on the premise of deeply understanding the reasons behind the decision.Ultimately,the model identify four cognitive scores,bringing it to an Fmeasure of 0.878.This result indicates that the model can be conveniently,reliably and interpretable for the early diagnosis of AD,thereby facilitating timely cognitive intervention for MCI patients and AD patients.3.Further explore the feasibility of conveniently,reliably and interpretably predicting potential AD patients,and establish a fourclass classification(CN vs.pMCI vs.sMCI vs.AD)model for Alzheimer’s disease,an intelligent AD screening and prediction system is constructed accordinglyTo further predict conversion in MCI patients,a four-class classification(CN vs.pMCI vs.sMCI vs.AD)model is established based on RN-SSAS and multiple cognitive scores.Similarly,single-instance and global-instance interpretations based on Shapley values are also equipped for each class of the model to give physicians insight into how decisions are made.At last,the model determine five different cognitive scores with a corresponding F-measure of 0.881.This means that the model can not only be conveniently,reliably and interpretable for AD screening,but also can be further used to predict whether MCI patients will develop AD patients within a few years,thus facilitating cognitive interventions to more effectively delay the development of Alzheimer’s disease and even prevent its occurrence.In short,this article in view of the existing AD research should not be used for screening,diagnosis and Alzheimer’s disease situation is becoming more and more serious,targeted intelligent AD screening system is established and intelligent AD screening prediction system,realize the convenient and reliable and can explain to screening and predict MCI AD conversion,thus to improve the screening rate of Alzheimer’s disease.During this period,it has been learned that large,small sample data sets exist today and often exhibit category imbalances,particularly in the medical field.In view of this,RN-SSAS is proposed in this paper to solve the problem that classification modeling cannot be carried out under a small sample size&multi-category&unbalanced data set.In short,this article in view of the existing AD research are not suitable for AD screening,and Alzheimer’s disease situation is becoming more and more serious,targeted intelligent AD screening system and intelligent AD screening prediction system is established,realize the convenient and reliable and can explain to screening and predict MCI AD conversion,thus to improve the screening rate of Azheimer’s disease.During this period,it has been learned that large,small sample data sets exist today and often exhibit category imbalances,particularly in the medical field.In view of this,RN-SSAS is proposed in this paper to solve the problem that classification modeling cannot be carried out under a small sample size&multi-category&unbalanced data set.In short,in view of the fact that the existing AD diagnosis research is not suitable for screening,and the situation of Alzheimer’s disease is becoming more and more serious,this paper establishes an intelligent AD screening system and an intelligent AD screening prediction system.The results show that the two systems can achieve convenient,reliable and interpretable screening for AD and predict MCI transition,which is beneficial to improve the screening rate of Alzheimer’s disease.In addition,in the early stage of the research,it was learned that there are a large number of small-sample datasets today and often appear to be class imbalanced,especially in the medical field.In view of this,this paper proposes a solution named RN-SSAS to solve the problem that classification modeling cannot be performed under small sample size&multi-class&class imbalance datasets.